Long-Term Overall Survival After Selective Internal Radiation Therapy for Locally Advanced Hepatocellular Carcinomas: Updated Analysis of DOSISPHERE-01 Trial

DISCUSSION

After analysis of long-term 65.8-mo follow-up, improvement in median OS was shown to be sustained in the PDA group. The 22.9-mo (95% CI, 11–36.5 mo) median OS reached in the PDA group was observed in a population with severely advanced disease, including PVT involvement for 65% of them and a mean tumor size of 10.6 cm (7). These results compare favorably with results obtained in immunotherapy trials in which the reported median OS was 19.4 mo (95% CI, 11–36.5 mo) with atezolizumab plus bevacizumab (9) and 16.4 mo (95% CI, 14.1–19.5 mo) with durvalumab plus tremelimumab (10).

However, it must be mentioned that a direct comparison between SIRT studies and studies using systemic drugs turns out to be hazardous, especially on account of differences in the patient populations included. Indeed, in SIRT trials, PVT was shown to be more common (∼65% in DOSISPERE-01 (7) vs. only 26%–38% in immunotherapy trials (9,10)); SIRT patients did not exhibit any extrahepatic spread, whereas 53%–63% of immunotherapy-treated patients exhibited distant metastases (9,10); additionally, underlying cirrhosis characteristics and etiology differed (hepatitis B was reported in only 26% of SIRT-treated patients (7) vs. 31%–49% of immunotherapy-treated ones (9,10)).

Results for PVT patients deserve to be further highlighted, given that this patient population is of specific interest. Indeed, PVT patients were classified as advanced patients according to the Barcelona Clinic of Liver Cancer classification (2), similarly to patients with extrahepatic spread, despite portal vein invasion representing only a locoregional spread, which is thus accessible to SIRT (3,11,12), unlike distant metastasis. On the basis of this classification, the recommended treatment of PVT patients is systemic therapy rather than locoregional therapy such as SIRT (2). Although the statistically significant difference in OS for PVT patients was lost in this long-term analysis, this was most likely due to a lack of power, as this study was stopped by anticipation, whereas the trends are still striking, with a median OS of 22 mo (95% CI, 10.3–36.5 mo) in the PDA group versus 9.5 mo (95% CI, 5.3–17.6 mo) in the SDA group (HR, 0.52; 95% CI, 0.26–0.1; P = 0.058). Here, again, the median OS of 22 mo (95% CI, 10.3–36.5 mo) that was reached in the PDA group in PVT patients compares favorably with that obtained in immunotherapy-treated patients, for whom the median OS in the event of macrovascular invasion was 14.2 mo (95% CI, 11–19.4 mo) under atezolizumab plus bevacizumab (9) but was not reported for the durvalumab-plus-tremelimumab combination (10).

Regarding dosimetry parameters, the impact of TD on OS remains significant. Furthermore, this study revealed a significant impact exerted by PLD on OS (not reported in the first report). Indeed, median OS was 22.9 mo (95% CI, 11–48.1 mo) for patients receiving a PLD of 150 Gy or more versus 10.3 mo (95% CI, 5.9–17.6 mo) for those treated with a PLD of less than 150 Gy. Although TD and PLD are not independent prognostic indicators, this point is of particular interest for technical reasons when personalized dosimetry based on TD is difficult to perform. This can be the case given that tumor segmentation can turn out to be challenging in several instances (disease not well delineated, infiltrative disease, or multiple lesions), as when there is a large lesion with multiple feeders. In this situation, ^99mTc-macroaggregated albumin dosimetry would often require, to be accurate, 1 injection of ^99mTc-macroaggregated albumin for each feeder, separated by at least 24 h. In those situations, a PDA based on the PLD is doable, as for radiation segmentectomy (13,14), in this specific patient population exhibiting good liver function and at least 30% of hepatic reserve.

The analysis of OS rates from 2 to 5 y likewise discloses additional information of interest. For patients with poor features, namely those randomly assigned to SDA, receiving a TD of less than 205 Gy or a PLD of less than 150 Gy, or not downstaged to resection, OS rates were dramatically decreased, from 13% to 22% at 2 y and to less than 10% at 4 y. For patients with good features, besides those who were resected, OS rates were between 44% and 50% at 2 y and between 33% and 37% at 3 y. Only resected patients displayed an OS rate of more than 50% at 5 y.

The huge prognostic impact of secondary surgery on long-term OS in this population of patients with large lesions and often PVT, even with PDA, is highlighted by the loss of difference in median OS censored at the time of surgery between arms, as by the multivariate analysis. This result, the major impact of secondary resection to achieve prolonged long-term OS even with personalized dosimetry with large lesions and PVT, was not necessarily intuitive as it is not the case for small lesions. Indeed, in the LEGACY study (15), with 94% of lesions smaller than 5 cm (and 62% < 3 cm), OS was not driven by secondary surgery. Indeed, for patients who received SIRT as a unique treatment, OS was similar to that of patients with resection or transplantation (3-y OS rate of 86.6% without secondary surgery vs. 92.8% for patients with resection or transplantation) (15). The difference in the impact of surgery between these 2 kinds of populations can be explained by the fact that complete pathologic response is more frequently observed for small lesions, that is, in 67% of the patients of the LEGACY study who underwent resection or transplantation (14) versus only 10% for large lesions with often PVT in the DOSISPHERE-01 study (7). Furthermore, patients with large lesions and often PVT have a much higher risk of recurrence (median PFS was only 6 mo in the DOSISPHERE-01 study vs. not reached at 24 mo in the Legacy study (15)), then SIRT allowed to accurately evaluate the biological test of time, allowing surgery to be performed only on patients with a low risk of recurrence.

Two key messages arise from these observations. First, everything possible has to be done with SIRT to downstage patients to surgery, even including PVT patients, as it is the only way to achieve acceptable prolonged median OS rates, including an OS of more than 50% at 5 y. Such prolongation of median OS has recently been described after post-SIRT surgery performed for initially unresectable patients (16). In that study, including 18 patients who were accurately downstaged and then resected, 78% of whom presented with PVT, median OS was 61.8 mo (95% CI, 31.4 mo to not reached) (16). Thus, SIRT is most likely to optimize preparation and accurate selection of good PVT candidates who are eligible for surgery on account of its strong debulking effect (including portal vein complete response and revascularization) (7,11) and its ability to ensure contralateral liver hypertrophy, which is usually attained within 3–6 mo (17,18). In addition, biological tests performed in a timely manner permit exclusion of patients with early relapse after SIRT.

The second key message arising from our analysis is that even for patients with good features, excepting the resected ones, the OS rates were seen to decrease quite rapidly between 2 and 4 y, which is another strong argument to evaluate SIRT delivered in combination with immunotherapy in this patient population. Indeed, several arguments are in favor of combining SIRT with immunotherapy, including the strong debulking effect of SIRT in the treated area, even in the presence of large lesions (7). The systemic action of immunotherapy will be complementary to the local action of SIRT. Furthermore, a potential synergy between both approaches is awaited as SIRT is known to induce an immune response (19,20). Additional arguments for these combinations could be situations in which immunotherapy efficacy is likely diminished, such as in patients without viral hepatitis, in whom the HR was found to be 1.05 with atezolizumab plus bevacizumab versus sorafenib (9) but 0.58 for patients with viral hepatitis B. This may also apply to large lesions. Indeed, the response rate (modified RECIST) for the atezolizumab-plus-bevacizumab arm was significantly (P = 0.0097) lower in the presence of lesions larger than 5 cm than in the presence of lesions smaller than 5 cm, being 26.1% versus 40.9%, respectively (data calculation based on results presented in Fig. 1 (21)).

In addition to the rather small number of patients included, a major limitation of this study is that SIRT using personalized dosimetry was not randomized to this population’s standard treatment, consisting of atezolizumab plus bevacizumab or double immunotherapy. Therefore, drawing definite conclusions about the role of SIRT in this specific population is not possible. Thus, further randomized studies are warranted to better define this role.