Radioembolization of hepatocarcinoma with (90)Y glass microspheres: development of an individualized treatment planning strategy based on dosimetry and radiobiology.

PURPOSE: The aim of this study was to optimize the dosimetric approach and to review the absorbed doses delivered, taking into account radiobiology, in order to identify the optimal methodology for an individualized treatment planning strategy based on (99m)Tc-macroaggregated albumin (MAA) single photon emission computed tomography (SPECT) images.

METHODS: We performed retrospective dosimetry of the standard TheraSphere® treatment on 52 intermediate (n = 17) and advanced (i.e. portal vein thrombosis, n = 35) hepatocarcinoma patients with tumour burden < 50% and without obstruction of the main portal vein trunk. Response was monitored with the densitometric radiological criterion (European Association for the Study of the Liver) and treatment-related liver decompensation was defined ad hoc with a time cut-off of 6 months. Adverse events clearly attributable to disease progression or other causes were not attributed to treatment. Voxel dosimetry was performed with the local deposition method on (99m)Tc-MAA SPECT images. The reconstruction protocol was optimized. Concordance of (99m)Tc-MAA and (90)Y bremsstrahlung microsphere biodistributions was studied in 35 sequential patients. Two segmentation methods were used, based on SPECT alone (home-made code) or on coregistered SPECT/CT images (IMALYTICS™ by Philips). STRATOS™ absorbed dose calculation was validated for (90)Y with a single time point. Radiobiology was used introducing other dosimetric variables besides the mean absorbed dose D: equivalent uniform dose (EUD), biologically effective dose averaged over voxel values (BEDave) and equivalent uniform biologically effective dose (EUBED). Two sets of radiobiological parameters, the first derived from microsphere irradiation and the second from external beam radiotherapy (EBRT), were used. A total of 16 possible methodologies were compared. Tumour control probability (TCP) and normal tissue complication probability (NTCP) were derived. The area under the curve (AUC) of the receiver-operating characteristic (ROC) curve was used as a figure of merit to identify the methodology which gave the best separation in terms of dosimetry between responding and non-responding lesions and liver decompensated vs non-decompensated liver treatment.

RESULTS: MAA and (90)Y biodistributions were not different (71% of cases), different in 23% and uncertain in 6%. Response correlated with absorbed dose (Spearman's r from 0.48 to 0.69). Responding vs non-responding lesion absorbed doses were well separated, regardless of the methodology adopted (p = 0.0001, AUC from 0.75 to 0.87). EUBED gave significantly better separation with respect to mean dose (AUC = 0.87 vs 0.80, z = 2.07). Segmentation on SPECT gave better separation than on SPECT/CT. TCP(50%) was at 250 Gy for small lesion volumes (<10 cc) and higher than 1,000 Gy for large lesions (>10 cc). Apparent radiosensitivity values from TCP were around 0.003/Gy, a factor of 3-5 lower than in EBRT, as found by other authors. The dose-rate effect was negligible: a purely linear model can be applied. Toxicity incidence was significantly larger for Child B7 patients (89 vs 14%, p < 0.0001), who were therefore excluded from dose-toxicity analysis. Child A toxic vs non-toxic treatments were significantly separated in terms of dose averaged on whole non-tumoural parenchyma (including non-irradiated regions) with AUC from 0.73 to 0.94. TD50 was ≈ 100 Gy. No methodology was superior to parenchyma mean dose, which therefore can be used for planning, with a limit of TD15 ≈ 75 Gy.

CONCLUSION: A dosimetric treatment planning criterion for Child A patients without complete obstruction of the portal vein was developed.