Personalized Dosimetry with Intensification Using ⁹⁰Y-Loaded Glass Microsphere Radioembolization Induces Prolonged Overall Survival in Hepatocellular Carcinoma Patients with Portal Vein Thrombosis

DISCUSSION

The radio-induced tumoricidal effect has been found to respond to the determinist rule, with a minimum absorbed TD necessary to induce response (tumoral threshold dose).

To date, 2 independent teams have described this dose–response relationship in the context of hepatocellular carcinoma with a clear tumoral threshold dose identification, based on ^99mTc-MAA scintigraphy, while using ⁹⁰Y-loaded glass microspheres (10–12). By applying a biologically effective dose and a voxel approach, Chiesa et al. were the first to describe a dose–response relationship, with a tumoral threshold dose of 257 Gy (10). Using a simple MIRD approach and mean dose evaluation, we also described a tumoral threshold dose of 205 Gy, with sensitivity and accuracy in the prediction of response of 100% and 91%, respectively (11). TD has been described as the only predictive parameter of response on multivariate analysis (12).

More recently, posttherapeutic dosimetry has also identified a clear dose–response relationship (14) in hepatocellular carcinoma patients treated with ⁹⁰Y-loaded glass microspheres, thus confirming the significant relevance of TD.

^99mTc-MAA SPECT/CT–based dosimetry is, however, the only technique currently available that can be performed before microsphere injection, thus allowing for modification of treatment schedule, such as a personalized dosimetric approach with intensification in cases of low TD. We recently described this approach, comprising personalized dosimetry and intensification at the lobar level, with interesting results (79% response rate) in an unselected cohort of hepatocellular carcinoma patients (12). Using this approach, we found no evident correlation between tumor size and response (12). This finding is highly relevant, as tumor size has been recognized as a critical parameter related to response using radioembolization with no personalized dosimetric approach. Previous studies have reported significant decreases in response rate as tumor size increased (7,15). With the application of standard radioembolization, the complete histologic response rate was 89% for lesions smaller than 3 cm and only 33% for lesions larger than 5 cm (15). In another study, the morphologic response rate was 82% for lesions smaller than 5 cm, 17.6% for those between 5 and 10 cm, and only 2% for those larger than 10 cm (7). This point underlines the fact that intensification should be considered for large lesions. Riaz et al. have also previously described an intensification approach at a segmental level called radiation segmentectomy, producing good results (81% response rate; no clinical toxicity) and thus underlining the usefulness of the intensification concept (16,17). Recent results by means of radiation segmentectomy have demonstrated that the rate of complete histologic response was significantly higher for segments receiving a mean dose of 190 Gy or more (i.e., 66.6% for a segment dose ≥ 190 Gy, compared with only 25% for a segment dose < 190 Gy; P = 0.03) (17). TDs were not provided in that study, thus rendering any direct comparison with our threshold TD of 205 Gy impossible. Nevertheless, the study brings to light new evidence of the dose–response relationship observed with radioembolization. It also demonstrates that intensification at a segmental level for small lesions (median size, 2.6 cm) can be performed in an uncomplicated manner (not requiring evaluation of the TD), producing good clinical results (safety and response rate of 86%), with the objective of delivering a threshold dose of 190 Gy to the segment. Interestingly, the mean ILD of our intensified patients was in the same range, namely 187 Gy.

As PVT patients often present with large lesions, and because of the existence of PVT itself, lobar intensification seemed particularly worth investigating in this selected patient cohort. In this study, intensification was applied in 37% of cases. For intensification, our primary endpoint was to achieve a TD of at least 205 Gy, namely the threshold TD that had already been identified (11,12). We had previously observed (12) that the response rate was higher for patients with a TD of more than 275 Gy, in comparison with patients with a TD of 205–275 Gy. Intensification was therefore also performed for some patients with a predicted TD of more than 205 Gy (as predicted using the standard dosimetric approach), with a little over 50% of intensified patients thus concerned. We might wonder if an interesting endpoint of intensification could be reaching the maximal tolerated dose to the healthy injected liver. One difficulty with this hypothesis is the fact that this maximal tolerated dose is not well defined with radioembolization, especially for cirrhotic patients, and we currently do not know up to which level we can intensify treatment. For these reasons, we conservatively decided to not use, as an endpoint, reaching of any potential maximal tolerated dose.

Nevertheless, using this concept of intensification, we obtained a high response rate (85%) in this cohort of PVT patients with large lesions (mean tumor size, 8.5 ± 3.1 cm). Once again, we observed that TD clearly affected the response rate, as previously described (10–12). Achieving high response rates is a crucial goal that has been associated with extended OS (7,18), especially in PVT patients, with responders exhibiting a 3-y survival rate of 25%, compared with only 4.4% for nonresponders (P = 0.02) (7). In line with previous findings (11,12), we saw no evident correlation between response and size, and we observed a high response rate (91%) for patients with lesions measuring 10 cm or more. This finding demonstrates that, with a controlled TD (using a personalized dosimetric approach and intensification), we were able to restore, at least partially, the prognosis of patients with large lesions. An additional interesting finding was that PFS and OS also strongly correlated with TD, with a median OS of only 4.3 mo (95% CI, 3.7–5 mo) for a TD of less than 205 Gy, compared with 18.2 mo (95% CI, 8.5–28.7 mo) for a TD of 205 Gy or more (P = 0.005).

The other parameter that strongly correlated with OS in our study was PVT targeting. The 4 patients who underwent no ^99mTc-MAA PVT targeting exhibited severe acute liver toxicity leading to death. One explanation is that radioembolization induces transient portal hypertension (19), which can be poorly tolerated by PVT patients who present with poor liver function and no accurate treatment of PVT because of the absence of targeting. The median OS was only 3 mo (95% CI, 3–3.7 mo) for patients with poor PVT targeting, compared with 20.2 mo (95% CI, 12–25.1 mo) for those with good PVT targeting (P < 0.0001). Interestingly, we found that PVT location (main vs. branch/segmental) was not statistically related to OS. This finding is also of significant interest since patients with main PVT are often considered poor candidates for radioembolization or the procedure may be considered contraindicated in these cases. In reality, the key parameter for PVT is ^99mTc-MAA targeting and not PVT location. Thus, irrespective of PVT location, radioembolization should be considered and PVT targeting evaluated. Only poor PVT targeting should be considered a contraindication.

In this study, using intensification did not increase liver toxicity. The rate of severe and permanent hepatic toxicity was no higher in intensified patients (6%) than in those treated with the standard approach (12%). The only parameter found to strongly correlate with liver toxicity was the presence of main or branch PVT without ^99mTc-MAA targeting. No patient exhibited a combination of small hepatic reserve (<30%) with a HILD of more than 120 Gy, which represented a previously identified liver toxicity factor (12). It is for this reason that 2 patients did not receive intensification, given that the process would have induced this at-risk situation.

In this group of patients treated with a personalized dosimetric approach and intensification when necessary, the global median OS was 18 mo, regardless of TD or PVT targeting. Such a prolonged OS in nonselected PVT patients has never previously been described, whether using radioembolization with a standard dosimetric approach or sorafenib. Previous studies have, in fact, reported a median OS of between 6.4 and 13 mo for PVT patients using glass or resin microspheres (6–10), which contrasts with values of only 6.5–8.1 mo with sorafenib (4,9).

Child–Pugh A and lobar PVT patients appeared good candidates for radioembolization in our study, with a median OS of 23.3 mo (95% CI, 8–23.7 mo). A relatively long median OS ranging between 15.7 and 17 mo has also been described in this type of patient subgroup in other studies (5,7,20), thus emphasizing the particular interest of radioembolization in this context.

The median OS was 20.9 mo for good radioembolization candidates identified before therapy, namely patients with both a TD higher than 205 Gy and good PVT targeting.

Lastly, OS was not attained for patients who underwent surgery, though it did exceed 24.5 mo and was significantly longer in these patients than in those who did not undergo surgery (15 mo, P = 0.0493), underlining the potential interest in surgery for PVT patients exhibiting a good response to radioembolization.

The results of our study were observed with glass microspheres on day 3, after calibration, applying a specific activity of approximately 1,250 Bq/sphere. A recent simulation (21) has demonstrated that the specific activity of microspheres is hugely influential from a radiologic point of view, with lower radiobiologic effects observed with high specific activity because of the more heterogeneous distribution of radiation. Lewandowski et al. (22) proposed using glass microspheres on week 2, typically 8 d after calibration, with a lower specific activity of approximately 393 Bq/sphere. This technique has been shown to provide good clinical results, with a high response rate (57%) and a low hepatic toxicity profile (2% of grade 3/4 bilirubin toxicity). This approach also provides a good opportunity to optimize glass microsphere radioembolization. The use of glass microspheres with an extended shelf-life, combined with a personalized dosimetric approach, could be a promising method for further improving therapeutic effectiveness. New tumoral threshold dose values and maximal HILD values will, however, have to be defined for the use of glass microspheres with a specific activity of 393 Bq/sphere.

To our knowledge, only a single retrospective nonrandomized study has to date sought to compare sorafenib and ⁹⁰Y-loaded resin microsphere radioembolization (9). No statistical difference in OS was identified between the therapeutic options (8.6 mo for sorafenib vs. 6.4 mo for resin microsphere radioembolization, P = 0.879). Nevertheless, this study’s findings should be interpreted with caution, given the major bias against radioembolization that was present (9). The body-surface-area method was used for activity calculation, despite the fact that the partition method is now the technique preferred by several experts (23,24). Moreover, some patients were treated with radioembolization despite presenting with inappropriately high lung shunting, leading to death, and no personalized dosimetric approach used.

The main drawback of this study was its retrospective and uncontrolled nature. Despite the valuable data it has provided regarding the potential use of radioembolization in PVT patients, as in other trials, further randomized studies are still warranted.