The TheraSphere™ Advanced Dosimetry Retrospective Global Study Evaluation in Hepatocellular Carcinoma Treatment (TARGET) study was an international, multi-center, retrospective, single-arm study of patients from 13 centers located across eight countries who were treated using 90Y glass microspheres for HCC. Patients included were treated between January 1st, 2010 and December 31st, 2017. Sites included University Medical Center Utrecht, Netherlands; Northwestern University, Chicago, IL, USA; Eugene Marquis Center, Rennes, France; Indiana University School of Medicine, Indianapolis, IN, USA; MD Anderson Cancer Center, Houston, TX, USA; Stanford University, Stanford, CA, USA; University of Florida College of Medicine, Gainesville, FL, USA; Centre Hospitalier, Universitaire Vaudois, Lausanne, Switzerland; Universitatsklinikum, Essen, Germany; Foundation IRCCS Istituto Nazionale Tumori, Milan, Italy; Istanbul University Medical School, Istanbul, Turkey; Florence Nightingale Hastanesi, Istanbul, Turkey; King Faisal Specialist Hospital and Research Centre, Saudi Arabia. Protocols were approved by each site’s respective Institutional Review Boards (IRBs) and/or Independent Ethics Committees (IECs).
Patients were treated according to instructions for use of the device and local procedure. They were selected in consecutive reverse chronological order at each site to minimize patient selection bias.
The TARGET study consisted of three sub-studies: one evaluating the inter-site variability of pre-treatment dosimetry; one evaluating dosimetry software/methodology accuracy and reproducibility; and the primary study to collect clinical data to generate predictive models for NTAD and TAD association with clinical outcomes.
In addition to the standard inclusion criteria for patients undergoing TARE, the key patient inclusion criteria in this retrospective analysis were liver-dominant disease; up to 10 well-defined HCC tumor(s), with at least one tumor ≥ 3 cm, with or without PVT; Child–Pugh stage A or B7; BCLC stage A, B, or C; male or female; aged 18 years or older; bilirubin ≤ 2 mg/dL; and tumor replacement < 50% of total liver volume by diagnostic imaging. Patients were required to have had diagnostic imaging consisting of multi-phase contrast enhanced CT or contrast enhanced MRI within 3 months prior to 99mTc-MAA SPECT or SPECT/CT imaging; received an infusion of 99mTc-MAA and 90Y glass microspheres from a single location sufficient to cover the tumor(s) based on angiography; had clinical evaluation and laboratory evaluation (at least bilirubin) at baseline. Accumulation of 99mTc-MAA within PVT was not an inclusion criterion.
Exclusion criteria included prior external beam radiation treatment to the liver; prior loco-regional liver-directed therapy (e.g., transarterial chemoembolization [TACE] and/or TARE); prior liver resection or transplantation; any anti-cancer therapy between first TARE treatment and 3-month imaging; hepatic vein invasion; and administration to ≤ 2 segments (i.e., radiation segmentectomy). Radiation segmentectomy patients were excluded as ablative radiation to a minimal liver volume would skew the results.
Patient data were collected through a retrospective review of patient records to identify demographic characteristics, (disease-specific) medical history, Child–Pugh and Barcelona Clinic Liver Cancer (BCLC) status, Eastern Cooperative Oncology Group (ECOG) performance status, treatment-specific variables, and potentially dose-related adverse events (AEs). Demographic data collected included age, gender, race, and ethnicity. Medical history was deemed clinically significant (potentially relevant to identified AEs) by the site investigator. All potentially relevant ongoing medical conditions and baseline symptoms arise from treatment of those conditions to be a part of the patient’s medical history. Additionally, medical histories included clinical laboratory results (biochemistry, coagulation, hematology, and tumor marker) and concomitant medication. Disease-specific medical history included date of HCC diagnosis and presence of liver disease (etiology), hepatic encephalopathy, ascites, and subsequent treatments following diagnosis. Treatment-specific variables included those related to the preparation and imaging of 99mTc-MAA, the administration of 99mTc-MAA and 90Y glass microspheres, and specific treatment parameters. Clinical evaluation results were collected at baseline (+ / − 30 days), 90 days (+ / − 30 days), and 180 days (+ / − 30 days) after treatment. Laboratory testing results were collected at baseline (+ / − 30 days), 42 days (+ / − 17 days), and 90 days (+ / − 30 days) after treatment. Imaging data were collected at baseline (+ / − 30 days), 42 days (+ / − 17 days), 90 days (+ / − 30 days), 180 days (+ / − 30 days) after treatment; any additional imaging data available between 180 and 400 days was also collected. AEs were collected up to 90 days and were defined using the National Cancer Institute’s Common Terminology for Adverse Events (CTCAE) version 4.02.
Primary and secondary endpoints
The primary endpoint was to determine the relationship between total perfused NTAD and occurrence of ≥ grade 3 hyperbilirubinemia without disease progression in 15% or more of patients treated. This endpoint was selected based on consensus between study investigators during a study planning meeting. Secondary endpoints included assessment of treatment-related (S)AEs within 90 days of treatment with 90Y glass microspheres and their relationship with NTAD; determining the relationship between predicted TAD and best objective response (defined as complete response [CR] or partial response [PR] by mRECIST and RECIST 1.1); OS; the relationship between TAD and OS; and the relationship between tumor marker (alpha fetoprotein [AFP]) response and TAD. AEs and SAEs related or potentially related to TARE were collected up to 90 days after treatment. The maximum grade AE was recorded during this window. Best response was evaluated based on available imaging between 25 and 400 days post treatment, e.g., evaluable target lesions available at Day 90 (n = 130) and Day 180 (n = 74).
Imaging and dosimetry
Imaging was based upon institutional practice but required, at a minimum, angiography documenting the catheter position for 99mTc-MAA and 90Y glass microsphere administration, diagnostic contrast enhanced imaging (CT or MRI), and 99mTc-MAA SPECT or SPECT/CT imaging. Type of diagnostic and response imaging used at each site (e.g., CT, MRI) was recorded.
Dosimetry analyses were performed by local clinical teams as determined using standard instructions for dosimetry calculations. Lung shunt fraction (LSF) was calculated on planar 99mTc-MAA images per routine clinical practice (maximum 30 Gy for single administration; cumulative 50 Gy for multiple administrations).
Multi-compartment dosimetry was calculated retrospectively using Simplicit90Y™ software (Mirada Medical LTD., Oxford, UK). Dosimetry analysis included three main steps: registration, segmentation, and dose calculation. For the 99mTc-MAA images, a co-registered SPECT/CT was utilized over a SPECT, when available (178 SPECT/CT and 31 SPECT).
Registration was performed using data from diagnostic contrast enhanced CT/MRI with 99mTc-MAA SPECT/CT or SPECT. The registration was adjusted as needed using dosimetry software tools, then the quality of the registration was evaluated. The volume of interest (VOI) segmentation process involved the following: (i) segmenting lung volume on planar 99mTc-MAA scintigraphy; and (ii) segmenting whole liver, whole liver normal tissue, perfused liver, perfused normal liver (i.e., perfused liver minus sum of all tumors ≥ 2 cm), total perfused tumors (i.e., sum of all tumors ≥ 3 cm) and target lesion (single largest lesion) on multiphase contrast enhanced CT or MRI. Tumors ≤ 2 cm were conservatively added to normal perfused liver because of increased dosimetry errors at low volumes caused by registration errors, limited SPECT resolution, and partial volume effects. For accurate dosimetry, the total perfused tumor volume included all tumors ≥ 3 cm. A board-certified radiologist, not involved in assessment of tumor response, inspected all segmented volumes to confirm that the volume corresponded to the correct description and evaluated the quality of the registration. Dose calculation was performed using the VOI and counts from the registered 99mTc-MAA SPECT/CT according to the MIRD schema using a patient relative calibration factor and the local deposition method. The mean total perfused tumor absorbed dose was calculated based on a tumor size weighted average. NTAD and TAD were calculated primarily using 99mTc-MAA imaging; 90Y PET was only available for a minority of patients.
When the imaging described above was not available, alternative modalities were used, including SPECT being used instead of SPECT/CT, and, in one patient, cone beam CT (CBCT) was used instead of CT for tumor volume assessment.
The total projected sample size of between 200 and 300 patients was determined using simulated logistic regression curves of the relationship between the occurrence of ≥ grade 3 hyperbilirubinemia and NTAD, where the width of the 95% confidence intervals for the simulated curves suggested that between 200 and 300 patients would provide a reliable estimate of the adverse event of interest.
The main analysis population included all patients enrolled in the study who satisfied the eligibility criteria. Logistic regression was used to determine the relationship between the occurrence of ≥ grade 3 hyperbilirubinemia and NTAD, the relationship between OR and TAD, and between AFP response and TAD. Kaplan–Meier methodology was used to analyze OS. Cox regression was used to evaluate the relationship between TAD and OS.
Logistic regression was also used to assess the relationship between OR and TAD after taking account of pre-defined covariates of interest, according to the following steps:
Step 1: Each covariate was included, one at a time, together with TAD, in a series of univariable models. If the covariate had 2-sided p value < 0.1, it was included in the model selection procedure described in Step 2.
Step 2: A backward elimination procedure was used to determine the final multivariable model, starting with all the significant covariates identified in Step 1. A 2-sided significance level of 10% was used for a covariate to remain in the model. The process was repeated until all the remaining covariates had 2-sided p values < 0.1.
Similarly, Cox regression was used to assess the relationship between OS and TAD after taking account of pre-defined covariates of interest, according to the steps described above.