Transarterial Radioembolization for Hepatocellular Carcinoma and Hepatic Metastases: Clinical Aspects and Dosimetry Models

Abstract

Transarterial radioembolization (TARE) with Yttrium-90 (⁹⁰Y) microspheres is a liver-directed therapy for primary and metastatic disease. This manuscript provides a review of the clinical literature on TARE indications and efficacy with overviews of patient-selection and toxicity. Current dosimetry models used in practice are safe, relatively simple, and easy for clinicians to use. Planning currently relies on the imperfect surrogate, ^99mTc macroaggregated albumin. Post-therapy quantitative imaging (⁹⁰Y SPECT/CT or ⁹⁰Y PET/CT) of microspheres can be used to calculate the macroscopic in vivo absorbed dose distribution. Similar to the evolution of other brachytherapy dose calculations, TARE is moving toward more patient-specific dosimetry that includes calculating and reporting nonuniform dose distributions throughout tumors and normal uninvolved liver.

Introduction

Patients with liver dominant disease from either primary liver cancer or metastatic liver disease are often treated with a number of liver directed therapies. For small volume disease or isolated lesions, the standard of care is surgical resection. In patients where resection is not a good option due to medical comorbidities or anatomic considerations, more focal therapies such as microwave ablation (MWA), radiofrequency ablation (RFA), or stereotactic body radiation can be offered. In patients with multifocal or more diffuse liver parenchymal disease, modalities that provide greater coverage of tumor burden such as external beam radiation and hepatic intravascular therapies may be preferred.

Transarterial radioembolization (TARE) therapy relies on the dual blood supply of the liver to establish its therapeutic ratio. Tumors are primarily fed by the arterial supply while liver parenchyma receives most of its blood supply from the portal system. For TARE, blockade of arterial supply is not the desired outcome as in embolization or chemoembolization. Instead, the radioactive microspheres become lodged in arterioles in or on the periphery of the tumor, which is then irradiated by Yttrium-90 (⁹⁰Y) inducing DNA damage.^1,2

Indications for TARE

There are several indications for TARE including downsizing intrahepatic tumors, increasing future liver remnant (FLR) size prior to surgery,³ bridging to transplant for hepatocellular carcinoma (HCC),¹ controlling tumor size and inducing hypertrophy of FLR before resection (radiation lobectomy),^1,4,5 palliation or delayed progression for advanced HCC, and primary treatment of isolated liver lesions (radiation segmentectomy [RS]).^6,7

Downsizing Tumors

Evidence for the value of TARE for downsizing tumors prior to surgery is limited in patients scheduled for partial hepatectomy or metastasectomy. One multi-institutional, retrospective study examining TARE preoperatively in 100 patients treated in 16 centers across Asia and Europe for HCC and intrahepatic metastases showed a risk of 24% grade 3 and above complications in patients who received TARE (30% was seen in the matched non-preoperative TARE group).⁸ Liver failure complications were comparable between the 2 groups. Success of downstaging has been examined in a systematic review.³ This review was fairly comprehensive and compiled downstaging by the most common histologies. In HCC, the rate of successful downstaging across 11 studies with 307 patients was variable, ranging from 8% to 100%.³ Similar heterogeneity in response was noted in intrahepatic cholangiocarcinoma and liver metastases.³

Increasing FLR

The literature supporting the use of TARE for inducing hepatic hypertrophy and increasing the FLR is also limited. Prior work on maximizing FLR has been in the context of portal vein embolization prior to resection of liver metastases.^9–11 Preoperative TARE for increasing FLR typically involves a radiation lobectomy, usually of the right hepatic lobe (thereby stimulating increased size of the left lobe with potential tumor control in right lobe). One study of 83 patients who received preoperative right-sided radiation lobectomy noted right lobe atrophy, left lobe hypertrophy, and FLR hypertrophy as early as 1 month post-TARE.⁵ However, only 5 of 83 patients underwent successful right hepatectomy. Smaller series in both HCC and metastatic colorectal cancer (mCRC) have described feasibility, but the patient cohorts usually number 20–40 and there is wide variation in FLR hypertrophy.^5,12–16 There is a systematic review examining the use of TARE for contralateral liver lobe hypertrophy including 7 studies with 312 patients.¹⁷ The authors’ conclusion was that unilobar TARE does result in significant hypertrophy but that the time to achieve this may be very slow (ranging 44 days to 9 months) compared to other methods. The delay in hypertrophy could give rise to tumor progression in the interim which may affect resectability. Another pooled analysis across 9 retrospective studies showed a contralateral lobe hypertrophy rate of 7%–62% but the time to measurement was highly variable as was the evaluation of hypertrophy by imaging.³

Bridge to Transplant for HCC

HCC is unique in that liver transplant is a curative modality, particularly in patients with early stage disease. TARE has been described as a potential therapy as a bridge to transplant. Bridging to transplant is a variation of tumor downstaging (as are many of the other local treatment approaches) in which patients who fall outside the criteria for transplant based on tumor number and size can have a response to TARE, and be rendered eligible for transplantation. One study from Germany described 40 patients who underwent TARE prior to transplant and noted that 87.5% of explant specimens had evidence of complete (17/40) or partial treatment effect/necrosis (18/40) post-Y-90; partial necrosis was defined as the target lesion >50% necrotic with clusters of viable tissue.¹⁸ Another study from Italy showed successful downstaging in 78.9% of cases in 22 patients who received TARE prior to transplant.¹⁹ It should be noted that these studies were retrospective and that patients were not selected up front for downstaging. A pooled analysis of HCC patients who received up front TARE noted a transplant rate ranging from 10% to 100% but patient characteristics and stage were highly varied.³

Locally Advanced HCC

The role of TARE in multifocal HCC has been the subject of 2 randomized trials comparing sorafenib and TARE, the SARAH²⁰ and SIRveNIB²¹ trials. The SIRveNIB trial included 360 patients with 182 randomly assigned to TARE and 178 to sorafenib (standard of care chemotherapy pill for multifocal HCC) across 11 Asian countries.²² There was no difference in median overall survival (8.8 months in the TARE group vs 10 months in the sorafenib group). TARE had fewer attributable grade 3 and above adverse events.

The SARAH trial included 467 patients with 237 randomly assigned to TARE and 222 to sorafenib over 25 centers in France. Median survival of each group was identical to the SIRveNIB trial. Similarly, TARE had a reduced incidence of adverse events and also increased radiologic tumor response. The authors also collected quality of life data and noted improved quality of life with TARE.²⁰ Overall, the reaction within the HCC community was that these were negative clinical trials which provided no improvement in overall survival. There is now an ongoing VESPRO analysis to pool the individual patient data from the SARAH and SIRveNIB trials in a combined meta-analysis.²³ Northwestern University has published on their experience of over 14 years with 948 patients and noted that survival was improved in responders to TARE; however, this was retrospective, and radiologic response by itself in the Phase III setting has not always translated into an overall survival benefit.^24,25

Radiation Segmentectomy

Riaz et al published on their experience in 84 patients treated with RS for unresectable HCC noting low toxicity (9% grade 3 and 4 biochemical toxicity) and median time to progression of 13.6 months.⁶ The authors published an update with 70 patients and limited their selection of patients to solitary HCC <5 cm, unresectable, and not amenable to RFA, with RS dose >190 Gy (Medical Internal Radiation Dose [MIRD] treated volume) in patients with good liver function (Child Pugh class A), and no vascular invasion or extrahepatic metastases. The resulting 1-, 3-, 5 year, and median overall survival were 98, 66, 57%, and 80%, respectively.⁷ Comparison of RS in 235 patients with transarterial chemoembolization (TACE) + MWA in 417 patients in a propensity score-matched analysis of patients with solitary HCC < 3 cm showed similar local control (83% complete response rate in both groups) but lower complication rates for RS.²⁶ In comparison, stereotactic body radiation local control is reported as 87%−99% at 1 year and is noninvasive.^27–30 To become a standard treatment for early stage HCC, a clinical trial comparing RS to RFA or MWA should be performed, as RFA or MWA shows excellent local control in similarly sized tumors <2–3 cm.^27,31

Portal Vein Tumor Thrombus

There is some controversy as to whether portal vein tumor thrombus (PVTT) in HCC poses a contraindication for TARE. This has been a relative contraindication as there is a partial embolic effect of TARE in the arterial system and compromised portal circulation in the liver could result in greater hepatic toxicity and ischemia, although a smaller number of glass spheres could be used by treating with as high a specific activity as possible (~2000 Bq/sphere) to try and limit additional ischemic effects. Many interventional radiologists will treat patients with TARE even in the presence of PVTT. A recently published retrospective review of 120 patients from Italy suggested that TARE was effective in advanced HCC with segmental PVTT so long as bilirubin was not elevated (<1.2 mg/dL) and tumor burden did not exceed 50% of liver volume.^32,33 Voxel-based dosimetry from the 99mTc SPECT was used to limit parenchymal (ie, nontumoral liver [NT]) absorbed dose to 40 Gy initially, but then after generating a dose-toxicity curve, a limit of about 70 Gy was set followed by a change in the specific activity from 1000 Bq/sphere to 360 Bq/sphere.^33,34 The decrease in specific activity at time of administration may be accomplished by ordering a higher activity vial and increasing the time between vendor’s calibration date and administration. The authors did not calculate dose to the PVTT, and such personal dosimetry may offer additional gains. Similar small series from China, Korea and Italy also suggest a possible survival benefit of TARE.^35,36 The evidence for improvement in overall survival in PVTT for TARE is similar to that reported for EBRT.³⁷

Metastatic Disease

The use of TARE in mCRC to the liver is most highly featured in 3 randomized trials comparing chemotherapy and chemotherapy + TARE, FOXFIRE, FOXFIRE-global, and SIRFLOX trials. SIRFLOX reported first and FIREFOX was combined into a single analysis with the SIRFLOX data.^38,39 In the latter pooled analysis, 549 patients were randomized to FOLFOX chemotherapy alone (standard of care treatment for mCRC) and FOLFOX plus first-line TARE using radiation lobectomy.³⁹ The addition of TARE failed to improve overall survival or progression-free survival compared to FOLFOX alone although there was a reduction of intrahepatic progression in the TARE + FOLFOX arm.³⁹ As expected, toxicity was higher in the arm with TARE given dual therapies.³⁹ The conclusion from that study was not to use TARE as first-line therapy for colorectal metastases. There is also the ongoing EPOCH trial which evaluates TARE as a second-line treatment after chemotherapy for mCRC to the liver.⁴⁰

There is retrospective literature supporting the use of TARE in metastatic neuroendocrine carcinoma to the liver. One of the largest series reported on 244 patients across multiple institutions from the United States and Europe.⁴¹ The outcomes reported were radiologic response at 3 and 6 months as well as biochemical toxicity. Overall, response rates were relatively low with 8% complete response, 35% partial response, and 48% stable disease. In contrast, a systematic analysis of 11 studies showed median disease control of 86% at 3 months after TARE. Per the authors, the report of toxicity was limited by the low quality of the studies.⁴²

Patient Selection

There are relatively standardized guidelines, based on the manufacturer’s insert⁴³ and prior retrospective studies, for the delivery of TARE. It is generally recommended that patients fulfill the following eligibility criteria prior to TARE: Eastern Cooperative Oncology Group (ECOG) 0–2, normal liver function (aspartate aminotransferase (AST)/alanine aminotransferase (ALT) <5× upper limit of normal, bilirubin <2 mg/mL), normal creatinine (as patients will need contrast material for hepatic arteriography and catheterization), Child Turcotte Pugh score A-B7 (for HCC patients), minimal comorbidities, noninfiltrative tumor type, and <70% bulk disease (tumor volume <70% of the targeted liver volume) or tumor nodules that are not too numerous to count.

Toxicity of TARE

The most common side effect of TARE is postradioembolization syndrome. It consists of fatigue, nausea/vomiting, abdominal pain, and loss of appetite/weight loss, and its incidence ranges from 20% to 70% peaking at the first 2 weeks post-TARE administration.⁴⁴ A less common side effect is radioembolization/radiation-induced liver disease (REILD). There has been a comprehensive review published evaluating 19 studies.⁴⁵ The risk of REILD ranged from 0% to 11% for patients with HCC and 0% to 20% for patients with metastases. Randomized trials for both HCC and mCRC showed much a lower incidence of REILD with only a 0%−1% incidence of “radiation hepatitis” reported.^{20,21,38,39,46} Gastroduodenal ulcer/bleeding,⁴⁴ biliary toxicity,⁶ and radiation pneumonitis are all relatively unusual with TARE.^44,47,48

Microsphere Devices

TARE uses ⁹⁰Y, which is a high-energy beta emitter. The beta particles have a maximum energy of 2.28 MeV, average energy of 0.93 MeV, half-life of 64.1 hours, maximum range in tissue of 11 mm, and average range in tissue of 2.5 mm. A new device available in Europe is labelled with Holmium-166, which has a slightly less penetrating beta (max 1.85 MeV) but emits a gamma (81 keV) suitable for SPECT/CT and can also be imaged by magnetic resonance imaging (MRI).⁴⁹ Details regarding regulations of radioactive materials have been discussed elsewhere.^50–52

Two radioactive microsphere devices are approved by the United States Food and Drug Administration for TARE. Theraspheres (BTG International Ltd., UK) are glass microspheres (20–30 micron diameter) with a mass density of approximately 3.3 g/cc with ⁹⁰Y as an integral constituent, and contain 2500 Bq/sphere at calibration.^43,53 Sir-Spheres (SIRTeX Medical, Australia) are resin spheres (90% between 30 and 35 micron diameter)⁵⁴ and contain ⁹⁰Y bound to the resin with approximately 37.5–75 Bq/sphere at calibration. Due to differences in activity per sphere between glass and resin, a larger number of resin spheres are required to deliver the same amount of activity. Glass spheres are administered with saline, while resin uses dextrose 5% plus sterile water.⁵⁵

Segmentation, Body Surface Area (BSA) Model, and Dosimetry Models

Segmentation to Separate Tumor From Nontumoral Liver

Radiation oncologists are well versed in segmentation and understand that to perform rational personalized treatment planning, the tumors (T) should be separated from NT. In TARE, where body-surface area models are still prevalent, explicit segmentation is often not required. Glass sphere dosimetry is more patient specific by using the perfused volume, but it still does not require the full segmentation typical in radiation oncology. Segmentation is important for determining the optimal prescribed activity, but it can be difficult in practice with multiple registrations of diagnostic scans, multiple tumors, cirrhotic livers, lack of perfusion on the MAA, and diffuse tumors on anatomic scans.^56,57 Differences between anatomic and functional segmentation methods can lead to large differences due to the inclusion/exclusion of hypovascular regions.³⁴ Microspheres are not metabolized, they become physically lodged in the distal arterioles of tumors and liver parenchyma. Consequently, tumors should not be delineated solely on uptake in ^99mTcMAA or post-therapy ⁹⁰Y imaging – there needs to be some correlative anatomical or functional imaging for the underlying disease.

BSA and Empirical Models for Resin

The empirical fixed-activity method for resin spheres is not an absorbed dose calculation, does not take into account a patient’s liver volume, and is not recommended for determining activity.⁵⁸ The Body Surface Area (BSA) method is slightly personalized assuming a correlation between BSA and liver volume.⁵⁷ It represented an improvement over the empirical activity model and was reported safer based on a large retrospective study of 680 treatments, where 21 of 28 cases of radiation induced liver disease (RILD) occurred from a single center using the empirical model.⁵⁹ The BSA method is the most common method for determining activity to administer for resin spheres. However, it does not use absorbed dose to determine the activity and has rules for decreasing activity based on the lung shunt fraction (LSF) and amount of liver treated; the lack of personalization in the BSA technique has been studied by several authors.^56,60–62

Organ-Level MIRD Dosimetry Model

The organ-level MIRD committee absorbed dose calculations represent the current state of microsphere absorbed dose calculations in clinical practice.⁶³ MIRD assumes activity is uniformly distributed throughout the source region. Thus uptake at the macroscopic and microscopic scale is assumed to be uniform, and heterogeneity in activity (and absorbed dose) throughout the region is not modeled. However, in vivo microsphere distributions are nonuniform and spheres are known to cluster in the microvasculature.^64–66 Assuming microspheres are trapped permanently, the equation for absorbed dose to a volume of interest depends on a constant specific to ⁹⁰Y, the activity of microspheres in the volume, and the mass of volume:

$${\mathit{\text{Absorbed Dose}}}_{\mathit{\text{VOI}}}\left(Gy\right)=\frac{50\left(\frac{J}{GBq}\right)\times {\mathit{\text{Activity}}}_{\mathit{\text{VOI}}}\left(GBq\right)}{{\mathit{\text{Mass}}}_{\mathit{\text{VOI}}}\left(kg\right)}$$

Given the assumptions inherent to organ-level MIRD dosimetry, this equation can be used to calculate absorbed dose to lungs, tumors, NT, or the whole liver; the activity within the volume of interest must be known as well as the mass.

Lung Shunt Fraction and MIRD Lung Dosimetry

LSF following package insert instructions is determined from gamma camera planar images acquired after injection of the ^99mTc MAA at the end of the interventional radiology mapping procedure.^43,54 LSF is traditionally calculated using the geometric mean of ^99mTc MAA planar imaging. The lungs and total liver are contoured on each image, the geometric means are calculated, and the fraction of geometric mean counts in the lungs relative to the lungs + liver is reported. Planned lung dose is calculated assuming a LSF of the injected activity will go to lungs and uses the above organ-level MIRD equation. Excessive shunting to the lungs is a contraindication for TARE due to concerns of pneumonitis and LSF >10% will lead to decreased activity for BSA prescriptions. Lung dose limits on the glass package insert are 30 Gy for a single administration and 50 Gy for cumulative absorbed doses. For resin, the package insert lists a 20% LSF or 30 Gy to the lung. The limits were not derived from a dose-escalation study, but rather observational reports with resin spheres using nominal 1 kg mass for lung.^48,67

MIRD Perfused Liver Volume Dosimetry for Glass

For glass microspheres, the activity to deliver is based on the organ-level MIRD equation using the perfused liver volume (V_perf), while satisfying the lung dose constraint. The V_perf is defined using 3D imaging such as CT, MRI, cone beam CT (CBCT), or ^99mTc MAA SPECT/CT. This volume will include T and typically also a portion of the NT. In practice, V_perf is a lobe or segment, thus there is additional NT not explicitly included. Recommended doses when using glass spheres range from 80 to 150 Gy.⁴³ The activity will distribute between V_perf and the lungs (via LSF) with a small residual remaining in the expended vial and tubing. A nominal residual fraction can be assumed when planning, while the measured residual fraction can be applied after administration to report the absorbed dose. In addition to MIRD limitations, the glass dosimetry method does not fully separate NT from T, which complicates dose-response and toxicity studies.

Partition Model (PM) Dosimetry

The PM was developed to separately estimate absorbed dose to T, NT, and lungs.^67,68 It is an extension of the organ-level MIRD calculation. The PM requires an estimate of the tumor-to-nontumoral liver uptake ratio (TNR) from ^99mTc MAA SPECT/CT. TNR represents relative vascularity between T and NT compartments and is used to estimate how the activity distributes between them. The PM requires segmentation of T and NT volumes and is best used for solitary, clearly demarcated tumors.^56,57,69 Diffuse disease makes it difficult to contour while multiple tumors reduce the utility of a single TNR. Being MIRD-based, standard limitations still apply within the individual compartments. The PM is the only model listed on package inserts that explicitly separates both T and NT when determining activity to administer.

Voxel-Level Dosimetry

The vast majority of response data to date, and all randomized clinical data have not included the underlying spatial distributions and corresponding heterogeneity of absorbed doses throughout the liver. Voxel-level dosimetry of nonuniform activity distributions is described in MIRD Pamphlet 17.⁷⁰ Several different absorbed dose calculation algorithms have been used for ⁹⁰Y at the voxel-level including Monte Carlo, grid-based Boltzmann solvers, convolution superposition, other kernel-based methods, and local deposition.^70–79 The liver is relatively homogeneous and all calculations are equivalent to within 5% in soft tissue, but there are differences in lung where the beta range is longer.^73,80 The accuracy of the input activity distributions from SPECT/CT or PET/CT will directly affect the absorbed dose estimates. Further details can be found in an excellent review on the physics of radioembolization by Bastiaannet et al.⁸¹ The Figure demonstrates spatial differences between organ-level MIRD, PM, and voxel-level dosimetry.

Figure.

An example showing 3, Y-90 dosimetry models applied to the right lobe of a patient’s liver with tumor. (A) The organ-level MIRD using glass spheres assumes the activity is uniformly distributed throughout the perfused lobe and calculates a uniform dose of 94 Gy (orange colorwash). (B) The partition model (PM) with a TNR of 2.4 gives a nontumoral liver (NT) lobe dose of 74 Gy (yellow-orange colorwash) and a tumor (T) dose of 178 Gy (red colorwash). (C) Voxel-level dosimetry (Monte Carlo) gives a spatially varying dose at the macroscopic level with an average dose to the shown tumor and NT of 210 and 71 Gy, respectively. Other tumors in T compartment are not on this slice.

Role of Imaging for Planning and Treatment

Interventional Radiology Imaging

The interventional radiologist/oncologist reviews contrast-enhanced CT or MRI to help identify variant anatomy and aberrant arterial supply before the mapping procedure. Two-dimensional angiography, digital subtraction angiography,⁸² and more recently, CT or CBCT hepatic angiography are used by interventional radiology to document visceral anatomy, identify variant anatomy, interrogate perfusional flow in targeted regions, and occlude extrahepatic vessels if needed.^58,83,84 Additional technical details and aspects have been reviewed by several authors.^16,82,83

^99mTc MAA – Planar Imaging and SPECT/CT

In a mapping study performed prior to the Y-90 administration, the ^99mTcMAA is administered after the catheter is in the desired position. The purpose of the MAA is to (1) calculate the LSF as described previously, (2) identify extrahepatic deposition, and (3) perform treatment planning for institutions that use the ^99mTcMAA SPECT/CT for treatment volume segmentation or to generate a TNR for the partition model. As most institutions use MIRD for glass or BSA for resin, the majority of users will only be interested in LSF and extrahepatic deposition.

Planar imaging for LSF is the current standard of practice, but it does have known limitations and institutions are exploring SPECT/CT-based LSF.^57,85,86 MAA is an imperfect surrogate (shape, size, number, and breakdown) for microspheres and relying on it is the source of much debate.^87–92 Both PM and voxel-level dosimetry planning rely on the ^99mTcMAA SPECT/CT, with TNR for PM being estimated from activity concentration in T and NT compartments. There is not a standard method of segmenting T and NT or calculating TNR on the SPECT to determine activity with voxel-level dosimetry or PM.^69,81 Both SPECT thresholding and manual segmentation on CT are used.

Post-Therapy Imaging

⁹⁰Y bremsstrahlung SPECT/CT is difficult to quantify due to the lack of a photopeak, collimator detector scatter, and septal penetration from the high energy bremsstrahlung photons. To make it quantitative would require additional compensations not readily available on most commercial systems.^93–95 However, the image is still useful for qualitative comparison between delivered and planned distributions, and in checking for extrahepatic uptake.

⁹⁰Y PET/CT is another option for post-therapy imaging. Lhommel et al⁹⁶ showed that imaging ⁹⁰Y microspheres with PET/CT was feasible, even with the low positron yield of 32 ppm per decay.⁹⁷ Studies have shown that time-of-flight information helps with quantifying the noisy ⁹⁰Y PET images.^98,99 Thus, ⁹⁰Y post-therapy imaging is capable of providing the delivered activity distributions with a spatial resolution of a few mm. These can then be converted to absorbed doses, preferably using the voxel-level methods listed above. Additional details on post-therapy PET imaging can be found in a review by Tafti and Padia.¹⁰⁰

Future Directions

Investigators are currently using post-therapy ⁹⁰Y PET/CT to calculate absorbed doses to T and NT.^101–105 For lesions, strong absorbed dose-outcome relationships have been demonstrated, but there is considerable variability in the reported threshold absorbed doses potentially due to differences in methodology including imaging.^34,81,103 Furthermore, there is a difference between absorbed dose thresholds for resin and glass microspheres, and radiobiological dosimetry has been proposed to reconcile the differences.¹⁰² For NT, the dose-toxicity relationship has been more elusive and the potential need for microdosimetry to account for nonuniform dose deposition at the microscale level using models of the liver lobules, arterial tree, and microsphere clustering is being investigated.^106,107 Once robust thresholds for response and toxicity are found and validated in trials, TARE will transition away from BSA-style planning to the more personalized PM and voxel-based dosimetry.