Clinical Outcomes After Personalized Dosimetry for ⁹⁰Y Radioembolization of Advanced Hepatocellular Carcinoma: Defining the Role of a Device in a Pharma-Centric Landscape

In the United States, ⁹⁰Y microspheres have been approved for more than 20 y for treatment of hepatic malignancies. Glass microspheres (TheraSphere; Boston Scientific) received a humanitarian device exemption for treatment of hepatocellular carcinoma (HCC) from the Food and Drug Administration (FDA) Center for Devices and Radiologic Health in 1999 on the basis of 3 single-arm studies each with 7–24 patients (1). SIR-Spheres (Sirtex) received FDA premarket approval for treatment of metastatic colon cancer in combination with intraarterial chemotherapy in 2002 after a single randomized trial in which 35 patients received ⁹⁰Y microspheres through a surgically implanted hepatic artery infusion pump (1).

The process for FDA approval of devices is markedly different from that for pharmaceuticals. New anticancer drugs undergo highly regulated phase 3 randomized trials involving hundreds or thousands of patients to obtain approval by the Center for Drug Evaluation and Research. Most drugs enter clinical practice with level 1 evidence that qualifies them for inclusion into guidelines and for reimbursement. In contrast, 95%–98% of medical devices come to market through the FDA 510k approval process without requiring a definitive clinical trial, making adoption and reimbursement challenging (2).

The vulnerability of devices to the standards of approval used by the Center for Devices and Radiologic Health is particularly felt in oncology, where radiologists work as part of multidisciplinary disease management teams and are held to the same levels of evidence as those used by other oncologic specialties. Because the studies used for FDA approval of ⁹⁰Y microspheres were underpowered for efficacy assessment, acceptance in cancer guidelines has been limited. To address this obstacle, multiple large, randomized, controlled trials of ⁹⁰Y microspheres for HCC and metastatic colorectal cancer were completed. The SIRFLOX, FOXFIRE, and FOXFIRE Global trials collectively randomized more than 1,000 patients with liver-dominant metastatic colorectal cancer to first-line chemotherapy with or without SIR-Spheres and found no improvement in progression-free survival or overall survival (OS) (3). Similarly, the SARAH, SIRveNIB, and SORAMIC trials randomized patients with HCC to TheraSphere or sorafenib and found no significant difference in progression-free survival or OS (4). A subsequent systematic review and metaanalysis of 1,439 patients in 8 studies concluded that transarterial radioembolization (TARE) does not improve OS in patients with unresectable HCC (5). In contrast to these disappointing outcomes in advanced HCC, TARE has been utilized in early-stage HCC and shown high and durable complete response rates using a radiation segmentectomy approach in which an ablative dose is given into the arterial supply of the tumor-bearing segment. In 2021, the Center for Devices and Radiologic Health granted TheraSphere a premarket approval for HCC based on a multicenter retrospective report of objective response rate and duration in solitary tumors (6).

Given the high reported response rates to TARE, the lack of clinical efficacy in advanced HCC was a surprise to the radiologic community and stimulated investigation into how the treatment could be improved. The original dosimetry guidelines for both microspheres were simplistic. TheraSphere used a MIRD model in which a fixed dose of 120 Gy was delivered to the target volume of liver regardless of the tumor burden or tumor-to-liver uptake ratio. SIR-Spheres extrapolated the target volume from the patient’s body surface area and hence was vulnerable to both over- and underdosing depending on the actual size of the tumor-bearing liver relative to the patient’s body habitus.

Recent approaches to personalized dosimetry use ^99mTc-macroaggregated albumin SPECT and cone-beam CT images to segment actual tumor and liver volumes and calculate the liver and tumor activity using partition models or voxel-based dosimetry. These approaches ensure an adequate or boosted dose to the tumor while keeping the normal-liver dose below an acceptable threshold. The DOSISPHERE-01 phase 2 trial randomized patients with unresectable HCC to personalized versus standard dosimetry (7). It was stopped at the first interim analysis after reaching its primary endpoint of response rate by European Association for the Study of the Liver criteria, which were 71% for personalized dosimetry versus 36% with standard dosimetry.

Objective response rate is an appealing early surrogate for treatment efficacy but in fact tells us what happened only to the patients’ images, not to the patients. It is an insidious heuristic to associate the two when the correlation is quite poor. Among 146 randomized chemotherapy trials for colorectal cancer, the correlation between objective response rate and OS was only 0.1 (8). Novel classes of drugs such as tyrosine kinase inhibitors and immune checkpoint inhibitors have low objective response rates yet generate prolonged disease stabilization that translates to meaningful survival benefit. Hence, it is essential to report clinically meaningful survival outcomes for all trials of new therapeutic approaches.

In The Journal of Nuclear Medicine, Garin et al. update the DOSISPHERE-01 trial to report the long-term survival outcomes (9). The treated cohorts had advanced disease with a median tumor size of more than 10 cm: 65% had portal vein tumor invasion, and 85% were Barcelona clinic liver cancer class C. At a median follow-up exceeding 5 y, median OS was better with personalized dosimetry for the group as a whole (25 vs. 11 mo), and even the subset with macroscopic portal vein tumor invasion did well (22 vs. 9 mo). Importantly, the rate of adverse events did not differ between the 2 arms.

However, multivariate analysis revealed that the survival benefit was limited to those patients who subsequently underwent resection. When such patients were censored in the survival analysis, median OS dropped to 11.7 mo in the personalized dosimetry arm versus 10.7 mo with standard dosimetry (P = 0.23). Hence, personalized dosimetry was beneficial among patients downstaged to resection even with portal vein invasion but not among never-resectable patients such as those with bilobar disease.

The ability to downstage is a critical outcome. Eleven of 31 (32%) patients in the personalized dosimetry arm were downstaged to resection, resulting in improved survival in this subset. Recent guidelines from the European Society of Organ Transplantation recommend that all HCC patients without extrahepatic disease or macrovascular invasion be considered for downstaging (10). More data are required, but advanced HCC patients with unilobar disease and adequate hepatic reserve should be afforded the opportunity to be downstaged to either resection or transplantation. Ongoing and future studies will investigate the role of combining TARE with immunotherapy in this patient population.

There are a few take-home messages from this paper. First, whereas objective response rate is a routinely reported outcome measure in the early assessment of new therapies, it is an unreliable surrogate, and a clinically meaningful outcome measure should be primary. Second, the DOSISPHERE trial supports use of personalized dosimetry in the neoadjuvant setting when resection or transplantation is the ultimate therapeutic goal. Third, evidence for a survival benefit from TARE outside the neoadjuvant setting remains elusive. The role for TARE in the advanced HCC population has yet to be determined, and studies combining TARE and immunotherapy will help to further define future treatment algorithms.

• © 2024 by the Society of Nuclear Medicine and Molecular Imaging.

• Received for publication December 8, 2023.
• Revision received December 19, 2023.