Aim: Our goal was to limit liver toxicity and to obtain good efficacy by developing a dosimetric treatment planning strategy. While several dosimetric evaluations are reported in literature, the main problem of the safety of the treatment is rarely addressed. Our work is the first proposal of a treatment planning method for glass spheres, including both liver toxicity and efficacy issues.
Methods: Fifty-two patients (series 1) had been treated for intermediated/advanced hepatocellular carcinoma (HCC) with glass spheres, according to the Therasphere® prescription of 120 Gy averaged on the injected lobe. They were retrospectively evaluated with voxel dosimetry, adopting the local deposition hypothesis. Regions of interest on tumor and non tumor parenchyma were drawn to determine the parenchyma absorbed dose, averaged also on non irradiated voxels, excluding tumor voxels. The relationship between the mean non tumoral parenchyma absorbed dose D and observed liver decompensation was analyzed.
Results: Basal Child-Pugh strongly affected the toxicity incidence, which was 22% for A5, 57% for A6, 89% for B7 patients. Restricting the analysis to our numerically richest class (basal Child-Pugh A5 patients), D median values were significantly different between toxic (median 90 Gy) and non toxic treatments (median 58 Gy) at a Mann-Withney test, (P=0.033). Using D as a marker for toxicity, the separation of the two populations in terms of area under ROC curve was 0.75, with 95% C.I. of [0.55-0.95]. The experimental Normal Tissue Complication Probability (NTCP) curve as a function of D resulted in the following values: 0%, 14%, 40%, 67% for D interval of [0-35] Gy, [35-70] Gy, [70-105] Gy, [105-140] Gy.
Discussion: A limit of about 70 Gy for the mean absorbed dose to parenchyma was assumed for A5 patients, corresponding to a 14% risk of liver decompensation. This result is applicable only to our administration conditions: glass spheres after a decay interval of 3.75 days. Different safety limit (40 Gy) are published for resin spheres, characterized by higher number of particle per GBq (more uniform irradiation, bigger biological effect for the same absorbed dose).
Conclusion: As result of this study we suggest a constraint of about 70 Gy mean absorbed dose to liver non tumoral parenchyma, corresponding to about 15% probability of radioinduced liver decompensation while still aiming at achieving an absorbed of several hundreds of Gy to lesions.