Assessment of radiation sensitivity of unresectable intrahepatic cholangiocarcinoma in a series of patients submitted to radioembolization with yttrium-90 resin microspheres

Radioembolization is a valuable therapeutic option in patients with unresectable intrahepatic cholangiocarcinoma. The essential implementation of the absorbed dose calculation methods should take into account also the specific tumor radiosensitivity, expressed by the α parameter. Purpose of this study was to retrospectively calculate it in a series of patients with unresectable intrahepatic cholangiocarcinoma submitted to radioembolization. Twenty-one therapeutic procedures in 15 patients were analysed. Tumor absorbed doses were calculated processing the post-therapeutic 90Y-PET/CT images and the pre-treatment contrast-enhanced CT scans. Tumor absorbed dose and pre- and post-treatment tumor volumes were used to calculate α and α3D parameters (dividing targeted liver in n voxels of the same volume with specific voxel absorbed dose). A tumor volume reduction was observed after treatment. The median of tumor average absorbed dose was 93 Gy (95% CI 81–119) and its correlation with the residual tumor mass was statistically significant. The median of α and α3D parameters was 0.005 Gy−1 (95% CI 0.004–0.008) and 0.007 Gy−1 (95% CI 0.005–0.015), respectively. Multivariate analysis showed tumor volume and tumor absorbed dose as significant predictors of the time to tumor progression. The knowledge of radiobiological parameters gives the possibility to decide the administered activity in order to improve the outcome of the treatment.

www.nature.com/scientificreports/ Based on the well-known Linear-Quadratic (LQ) radiobiological model 13,14 , the knowledge of α could help to improve SIRT planning personalized activity in order to reach better clinical outcomes.
Purpose of this retrospective study was to calculate the radiobiological α parameter in a series of patients with unresectable intrahepatic cholangiocarcinoma (ICC) submitted to SIRT with 90 Y resin microspheres.

Materials and methods
Radiobiology: theoretical aspects. The well-known LQ model represents the main equation of radiobiology: with N 0 the number of tumor cells before the treatment; N the number of tumor cells remaining after treatment; D the average tumor absorbed dose; α and β the so-called intrinsic radiosensitivity; G the Lea-Catcheside factor, taking into account the capability of the cells to self-repair themselves.
A reasonable simplification for radioembolization (low dose-rate) is to neglect quadratic effects (i.e., Gβ = 0): Assuming a linear relationship between the number of tumor cells and the macroscopical mass of the lesion (M) the simplified LQ model (Eq. 2) becomes: with M f the minimum value of post-treatment tumor mass and M 0 the pre-treatment tumor mass.
In Eq. (3), α is the only unknown parameter: in fact, M f and M 0 can be measured and D can be easily calculated. Note that the knowledge of α allows to predict the best response M f to the therapy.
From Eq. (3), α can be calculated as: In SIRT, the distribution of the activity in the targeted liver cannot be considered homogeneous. Dividing the targeted liver in n voxels of the same volume (m 0 ), Eq. (3) becomes: with d i the i-th voxel absorbed dose. Again, it is possible to evaluate α 3D by numerically solving Eq. (5).
Study participants and study design. This retrospective study enrolled twenty-six patients with histologically-proven unresectable ICC, who referred to our center for 90 Y-SIRT from July 2013 to June 2018. Eight patients were excluded because of unavailability of post-therapy 90 Y-PET/CT (n = 4) or post-treatment radiological follow-up (n = 4). Moreover, 3 patients with disease progression at the first radiological follow-up were excluded because of a very low average tumor absorbed dose (< 35 Gy).
The clinical data of the 15 patients included in the study are summarized in Table 1. Patients were indicated to SIRT after multidisciplinary tumor board discussion. In 7 patients (47%), SIRT was associated to systemic chemotherapy as consolidation therapy.
Five out of the 15 patients selected for α calculation had bilobar disease treated in two different sessions within 4 weeks (3 patients presented mass-forming lesions, whereas 2 patients were characterized by multifocal lesions); in one patient treatment was repeated on the same lesion one year after the first treatment; thus, overall 21 SIRT procedures were performed in the selected population. Since it was not possible to obtain two separated values of M f in patients with bilobar mass-forming lesions, the data of two separate SIRT procedures performed in these patients were summed up to calculate the α values from Eq. (4), resulting in 18 α values. For the measurement of the α 3D values from Eq. (5), the data of SIRT procedures performed in mass-forming lesions were excluded, resulting in 15 α 3D values. SIRT procedure. All patients had a preliminary angiography, followed by the intra-arterial injection of In the absence of clinically relevant lung shunt fraction (LSF < 20%) and abdominal extra-hepatic shunts, patients underwent SIRT within 21 days, injecting 90 Y resin microspheres at the level of the planned injection site. In case of bilobar tumors (5/15), patients were treated with two subsequent procedures, at an interval of approximately 4 weeks. www.nature.com/scientificreports/ The administered activity was determined by using the BSA method. The volumes of total liver, targeted liver and tumor were obtained on pre-treatment contrast-enhanced CT by an implemented software using a semiautomatic contouring on a dedicated workstation (Advantage Window 4.7, GE Healthcare). Considering the liver/tumor tissue density as 1 g/mL, the liver/tumor mass was calculated from liver/tumor volume.
Post-treatment data and outcomes. Within 15 h from the SIRT procedure, patients performed posttherapeutic 90 Y-PET/CT scan of upper abdomen with a hybrid scanner Discovery710 (GE Healthcare Milwaukee, Wisconsin, USA) equipped with a 64-slices Optima CT660. PET acquisitions were obtained including the whole liver lasting 20 min/bed (usually 2 beds for each acquisition). Vue Point FX algorithm including Time-of-Flight information was used for reconstruction (2 iterations, 16 subsets). The 90 Y-PET/CT was useful to confirm the hepatic distribution of microspheres and to calculate the average absorbed dose and the 3D distribution of the absorbed dose within each tumor. 90 Y-PET/CT and contrast-enhanced CT scans were processed by using a dosimetry software (Simplicit90Y, Mirada Medical, Oxford, UK). Volumes of interest (VOIs) were semi-automatically defined on the contrastenhanced CT over the total liver, targeted liver and tumor, and transferred on the corresponding PET slices after the co-registration between the contrast-enhanced CT and the low-dose CT of the PET/CT scan. After entering the LSF values and the actual administered activity, the software turns out the dose-volume histograms (DVH) and the values of average absorbed dose of the targeted liver and tumor (Fig. 1).
The minimum dose absorbed by the 70% of tumor volume (D70) and the percentage of tumor volume that absorbed at least 100 Gy (V100) were extrapolated from the DVHs, as good predictors of therapeutic efficacy 15,16 .
During the follow-up, patients underwent contrast-enhanced CT scan 4-6 weeks after treatment and then every 3 months, to evaluate radiological tumor response according to the Response Criteria in Solid Tumors (RECIST) 1.1 criteria. The minimal mass of the treated tumor (M f ), corresponding to the maximum reduction of M 0 after SIRT, was evaluated for each treatment on the contrast-enhanced CT scans, using the same method employed to obtain M 0 . www.nature.com/scientificreports/ Statistical analysis. Quantitative values were expressed in terms of median in association of 95% confidence interval (95% CI). The non-parametric Wilcoxon comparison test was used to calculate differences between repeated measurements, before and after treatment. Correlations between absorbed dose measurements and tumor volume changes were performed. Time to progression (TTP) was calculated as the time between first SIRT and radiological tumor progression, using the Kaplan-Meier method. Tumor-related continuous variables were analyzed for an association with TTP by univariate and multivariate Cox regression models and the log-rank test. Variables with p < 0.1 in the univariate analysis were included in the multivariate analysis model. JMP8 software (Statistical Discovery ™) was used for statistical analysis.
Ethics approval. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This study was approved and informed consent was waived by the Ethics Committee for the Vast North West Area (CEAVNO 19/11/2020)-Tuscany, Italy. This article does not contain any studies with animals performed by any of the authors.
Median TTP was 7.3 months (95% CI 3.2-18.0 months). Table 3 shows the results of the univariate and multivariate Cox's proportional hazards model analysis for some tumor-related continuous variables. Tumor burden, D70 and V100 had no effects on TTP. Multivariate analysis showed a statistically significant negative effect of tumor volume on TTP (p < 0.03) as well as a positive effect of tumor absorbed dose on TTP (p = 0.05).
Following the RECIST 1.1 criteria, radiological PR corresponds to the reduction ≥ 30% of the diameter of the lesion. Considering the approximation of spherical volumes, this corresponds to a final tumor volume M f ≤ 0.34 M 0 .    www.nature.com/scientificreports/ Using the above-mentioned α 3D and α values, the corresponding Equivalent Uniform Dose (EUD PR ) needed to obtain a radiological PR was: Similarly, the average dose (D PR ) needed to obtain a radiological PR resulted to be:
In the present study, the radiosensitivity α of unresectable ICC treated with SIRT was retrospectively analysed, obtaining α values that are substantially in line with previous studies 30 Table 4. Comparison between patients submitted to SIRT alone and SIRT plus chemotherapy in terms of α and α 3D values. No statistically significant differences were found between two groups. NS not significant. www.nature.com/scientificreports/ α value of 0.010 ± 0.001 Gy −131 in a series of patients with different liver tumors (including HCC and ICC) treated with external beam radiotherapy.
The α values resulted to be lower than those of α 3D . Using Eq. (3), the adoption of a lower α translates into a higher average tumor absorbed dose, resulting into a higher probability of obtaining a significant reduction of the final tumor mass (M f ). Thus, the use of the α value would be preferable to the use of α 3D and of the 3D distribution of the tumor absorbed dose (that is even more difficult to assess).
In the setting of HCC, a significant difference between the absorbed dose needed to reach Tumor Control Probability (50%) for glass and for resin microspheres has been reported 11,32 . It probably depends on the intrinsic different microscopic distribution of resin and glass microspheres 10 due to the different specific activity (activity per microsphere). The differences between these two compounds applies also to ICC. In a series of 64 ICC patients treated with glass microspheres, Bourien et al. 33 obtained a threshold value of 260 Gy of the tumor absorbed dose to obtain a significant difference in overall survival (28.2 vs 11.4 months), while Levillain et al. 34 found a significant improvement of overall survival with doses above 86 Gy using resin microspheres (14.9 vs 5.5 months in a population of 58 patients). Thus, the estimated values of tumor radiosensitivity obtained in the present study should be considered specific for resin microspheres.
The relationship between the tumor absorbed dose and the reduction of tumor mass validates the use of the LQ model because it correlates the two variables in a logarithmic way. In the multivariate analysis, TTP was significantly associated to the tumor absorbed dose, supporting the relationship between dose and tumor reduction.
The knowledge of the α value enables personalized dosimetry. In fact, using the Eq. (7), the calculated tumour average absorbed dose (D PR ≥ 180 Gy) needed to obtain a radiological response resulted to be higher than that derived from the standard BSA method (100 Gy in our series). A recent Phase 2 Clinical Trial 35 enrolled 41 patients with locally advanced ICC treated with chemotherapy combined to SIRT with glass microspheres as first-line treatment obtaining an objective response of 39% at 3 months and a median OS of 22 months. Treatment personalization with the aim to provide at least 205 Gy to the tumor (317 Gy as median dose delivered to the tumor) played one of the main roles in the promising outcomes of this trial.
The main limitation of this retrospective study is represented by the limited number of patients enrolled through a long period of time, with heterogeneous indications and different therapies performed before and after SIRT. As a result, the calculated mean values of α and α 3D have a wide standard deviation, probably due to the tumor heterogeneity. No significant differences were observed in the α and α 3D values comparing patients treated with SIRT to patients treated with SIRT and chemotherapy; however, the role of concomitant systemic treatments requires further investigation. Moreover, the result of Eq. (7) (D PR ≥ 180 Gy) could be influenced by the tumor approximation of spherical volumes and the use of RECIST criteria that take into account the longest diameter of the lesion to evaluate the response.
Despite the obvious limitations, the study does not aim to draw conclusions on SIRT in patients with unresectable ICC, but it should make a methodological contribution to a greater comprehension of the intrinsic radiosensitivity of ICC in a selected series patients submitted to SIRT with 90 Y resin microspheres. The knowledge of these radiobiological parameters would enable further advances in the field of personalized dosimetry for SIRT, by calculating the tumor absorbed dose given the tumor specific radiosensitivity and the desired final mass of the treated lesion. Further studies are warranted to investigate how this approach could affect clinical outcomes, in terms of safety, tumor response and survival.