Statement of principal findings

Treatment options vary for patients with unresectable HCC according to the stage of the cancer and the underlying liver disease. The AG, therefore, considered three distinct unresectable HCC patient populations, defined with respect to the aim of therapy and eligibility for comparator treatments. These three populations were as follows: (1) patients eligible for transplant, (2) patients ineligible for transplant but eligible for CTT and (3) patients ineligible for CTT. These three populations largely correspond to early-, intermediate- and advanced-stage HCC.

There is a large body of evidence on the clinical effectiveness and safety of SIRT compared with sorafenib or TACE; seven RCTs, seven prospective comparative studies, five retrospective comparative studies and one non-comparative case series were included in the review of clinical effectiveness. However, only two studies were considered to have a low risk of bias, the SARAH¹⁹ and SIRveNIB²¹ RCTs, which both compared SIR-Spheres with sorafenib. These studies enrolled patients with locally advanced HCC not amenable to curative treatment modalities and ineligible for CTT; the evidence for the early- and intermediate-HCC populations was significantly more limited. Both RCTs found no significant difference in OS or PFS between SIR-Spheres and sorafenib, despite a statistically significantly greater tumour response rate in the SIR-Spheres arm of both trials. The SARAH trial¹⁹ reported a significant difference between groups in HRQoL, favouring SIR-Spheres; however, the proportion of patients who completed the questionnaires was low. AEs, particularly grade ≥ 3 events, were more frequent in the sorafenib group in both trials. There are some concerns regarding the generalisability of the results of these two RCTs to the UK HCC population, particularly the SIRveNIB trial,²¹ which was conducted in the Asia-Pacific region, where the aetiology and treatment of HCC differ from those in Europe.

The Sirtex company submission¹⁰² selected a subgroup of patients from the SARAH trial¹⁹ with ≤ 25% tumour burden and preserved liver function, defined as having ALBI 1, for the base-case analysis in its economic analysis. Although results appeared more promising in this subgroup of patients with a better prognosis, the results of this post hoc subgroup analysis should be prospectively validated before being considered relevant for clinical practice.

In studies that directly compared the different SIRTs, patients with PVT appeared to have better survival outcomes with TheraSphere than with SIR-Spheres; however, this result was from a small retrospective comparative study rated as being at a high risk of bias and, therefore, may not be reliable. Other studies comparing TheraSphere with SIR-Spheres that were not restricted to patients with PVT had conflicting results. The only study that compared QuiremSpheres with SIR-Spheres and TheraSphere was provided by Terumo as an addendum to its submission.¹⁰⁴ Clinical outcomes appeared to be similar between treatment groups; however, this was a very small pilot study with several methodological limitations.

Three NMA models were produced to represent the three different populations of unresectable HCC patients described above. Both the NMA in patients eligible for transplant and the NMA in patients eligible for CTT were not conducted owing to the uncertainty of using SIRT for bridging to transplant and downstaging in the UK, and a lack of good-quality evidence in patients eligible for CTT.

The base-case NMA was conducted in adults with unresectable HCC who have Child–Pugh class A liver function and are ineligible for CTT. There were no meaningful differences in OS between SIR-Spheres, sorafenib and lenvatinib in the per-protocol or ITT populations. All treatments appeared to have similar efficacy. There was only one low-quality retrospective study that directly compared TheraSphere with SIR-Spheres in the base-case population.³⁹ Adding this study as a sensitivity analysis had a substantial effect on the NMA results: TheraSphere showed a significant improvement in OS when compared with SIR-Spheres, sorafenib and lenvatinib. However, these results may be biased and unreliable as they rely on only one low-quality retrospective study.

The limitations in the effectiveness evidence had an important role in shaping the economic analysis and restricted the focus of the AG’s economic analysis to the population ineligible for CTT; this was the only population for which there were reliable estimates of the comparative effectiveness of SIRT with comparator technologies. The structure of the AG’s model was broadly similar to the structures of the models developed by BTG and Sirtex for this population and was designed around a decision tree and partitioned survival model. The decision tree was used to model the fact that some patients eligible to receive SIRT will fail the work-up procedure and will not receive SIRT treatment; in a scenario analysis, the decision tree was also used to allow a proportion of patients to go on to receive curative therapies. The partitioned survival model developed was based on three health states: PFS, progressive disease and death.

The results of the AG’s base-case analysis (probabilistic analysis), which assumed equal efficacy across all three SIRTs, suggested that TheraSphere is cost saving relative to both SIR-Spheres and QuiremSpheres. However, the incremental costs between TheraSphere and SIR-Spheres are < £300 and result from the additional cost of angiography required as part of the SIR-Spheres administration procedure. Pairwise NMB, assuming a £30,000 WTP threshold, for SIR-Spheres compared with TheraSphere was, therefore, close to zero (–£182). QuiremSpheres is associated with an incremental cost of £6955 relative to TheraSphere (exclusive of PAS). Pairwise NMB between QuiremSpheres and TheraSphere in the AG’s base case was –£6599, exclusive of PAS. In the analysis including the confidential PAS for QuiremScout, QuiremSpheres remained more costly than both TheraSphere and SIR-Spheres and, as such, the pairwise NMB remained negative (see Appendix 17 for full results).

In a fully incremental analysis, exclusive of the PAS discounts available for QuiremScout, sorafenib, lenvatinib and regorafenib, lenvatinib was the most cost-effective therapy and dominated TheraSphere (the lowest-costing SIRT treatment). Predicted NMB for lenvatinib compared with TheraSphere was –£2154. In a pairwise comparison of sorafenib with TheraSphere, the ICER for sorafenib was £31,974 per QALY, with an estimated NMB of –£150 (implying that TheraSphere is cost-effective compared with sorafenib at a WTP threshold of £30,000). In a fully incremental analysis inclusive of all confidential PAS discounts, lenvatinib remained the most cost-effective therapy across all scenarios, and dominated all three SIRTs, generating greater health benefits at lower costs. In pairwise comparisons of sorafenib with each SIRT, sorafenib also dominated all three SIRTs. Lenvatinib remained the most cost-effective option across 15 of the 17 AG scenarios when PAS discounts were included.

The results of the scenario analyses presented at list price showed that SIRTs were more likely to be cost-effective in the low tumour burden and ALBI 1 subgroup of patients, and when downstaging was permitted. The results of analyses conducted including PAS discounts for QuiremScout, sorafenib, lenvatinib and regorafenib, however, showed that the results of the AG’s economic analysis were robust to a range of alternative parameter values and assumptions, with a negative incremental NMB predicted for all SIRTs at a £30,000 WTP threshold (see Appendix 17 for details).

The AG’s economic analysis suggests that, although current life expectancy in patients ineligible for CTT is likely to be < 24 months, the predicted life extension generated by SIRT is likely to be < 3 months.

Strengths and limitations of the assessment

The key strengths of this assessment are as follows:

• The reviews of clinical effectiveness and cost-effectiveness were based on comprehensive searches of the literature, which were supplemented by data identified in recent systematic reviews of CTT treatments.
• The review of clinical effectiveness evidence included a detailed mapping and quality assessment of all comparative evidence on SIRT treatments across a range of alternative positions in the treatment pathway.
• The AG’s economic evaluation includes a fully incremental analysis of the three SIRTs (SIR-Spheres, TheraSphere and QuiremSpheres) and relevant systemic therapies (sorafenib and lenvatinib) in patients with CTT-ineligible HCC.
• The AG appropriately accounts for the fact that some patients eligible for SIRT treatment will fail the work-up procedure and will not go on to receive SIRT. Importantly, it recognises that patients who fail work-up are different from patients who successfully receive SIRT and tend to have inferior progression and survival outcomes.
• The AG’s economic analysis includes an exploratory analysis of two potentially plausible prospective subgroups: (1) low tumour burden/ALBI 1 and (2) no MVI.
• The AG’s economic analysis includes an exploration of the impact of downstaging in CTT-ineligible patients. The AG economic analysis also avoids double-counting the outcomes of patients who are downstaged to curative therapies.

The main weaknesses of the assessment are largely a consequence of weaknesses and gaps in the clinical evidence base:

• There is very limited evidence on the comparative effectiveness of SIRT with CTT in patients with either early- or intermediate-stage HCC. The AG did not consider the identified clinical evidence sufficient to produce an economic analysis and, therefore, the presented independent economic assessment covers only part of the NICE scope. The BTG company submission included an economic analysis of downstaging in CTT-eligible patients, whereas Sirtex presented a cost-minimisation model. The limits of the clinical evidence supporting these analyses and uncertainties regarding the equivalence of SIRT and CTT in this population mean that these analyses may be of limited relevance for decision-making.
• The AG did not have access to IPD from the SIRveNIB trial;²¹ instead, PFS and OS outcomes were replicated using a published algorithm. Although the precision of this replication is likely to be good, this process may have introduced a small loss of accuracy relative to the use of IPD directly. Furthermore, the lack of IPD meant that the SIRveNIB trial could not be included in scenario analyses exploring the low tumour burden/ALBI 1 and no-MVI subgroups.
• Lack of IPD for the REFLECT trial,⁸¹ comparing lenvatinib with sorafenib, meant that there were limited options for including lenvatinib in the economic analysis and the modelled HRs were based on a subgroup that did not fully align with the population eligible for SIRT. Furthermore, the AG’s base case makes the assumption of proportional hazards between lenvatinib and sorafenib despite some evidence presented in previous technology appraisals that this assumption may not hold.
• There was limited evidence on the relative effectiveness of TheraSphere compared with other SIRTs or systemic therapy, with the limited studies identified all rated as being at a high risk of bias.
• There is no evidence on the comparative effectiveness of QuiremSpheres, with the exception of one small, methodologically weak pilot study provided as a late addendum by Terumo.
• There is limited evidence on the long-term outcomes of patients who receive therapy with curative intent. The AG’s analysis and the Sirtex model present data from a historical US cohort study; these data are now several years old and potentially reflect a broader population of patients with HCC.

Uncertainties

The main uncertainties associated with the appraisal are as follows:

• The comparative effectiveness of SIRT in patients eligible for transplant or eligible for CTTs, such as DEB-TACE, TACE and TAE, is highly uncertain, with identified evidence limited to a small number of mainly observational studies.
• The comparative effectiveness of alternative SIRT (SIR-Spheres, TheraSphere and QuiremSpheres) in all HCC populations is largely unknown. The limited evidence available suggests that TheraSphere may be superior to SIR-Spheres for advanced HCC with PVI. The identified evidence is, however, of very low quality and, therefore, it is unknown whether or not the observed effects are the result of confounding bias. There is also no evidence on the comparative effectiveness of QuiremSpheres with any therapy, other than a very small pilot study with several methodological limitations that was provided as an addendum. This is significant, as QuiremSpheres uses a different work-up procedure and different radioactive isotope and therefore it is plausible that QuiremSpheres may have differential effectiveness when compared with SIR-Spheres and TheraSphere.
• The Sirtex submission¹⁰² puts forward a subgroup of patients with a low tumour burden and preserved liver function as a potential subgroup of patients who may benefit from treatment with SIR-Spheres. This subgroup was, however, not prespecified and the randomisation procedure did not stratify for these characteristics. The subgroup analysis is also based on very few patients. The extent of any benefits in this subgroup are, therefore, subject to considerable uncertainty and a confirmatory study would be required to be confident that the observed benefits are not spurious.
• The role of downstaging in a CTT-ineligible population is currently unclear. In the SARAH trial,¹⁹ a small proportion of patients were successfully downstaged to curative therapies. Advice received by the AG from clinical experts, however, suggests that downstaging in this population is likely to be very rare, and it is unclear whether or not the SARAH trial is representative of UK practice in this regard.
• In the SARAH trial,¹⁹ patients with bilobar HCC had each lobe treated in separate SIRT administrations to avoid the risk of REILD. The Sirtex submission, however, suggests that, in UK practice, patients with bilobar HCC would have both lobes treated simultaneously. The impact of sequential versus simultaneous treatment is largely unknown and it is not fully clear what practice would be adopted in the UK; advice received from the AG’s clinical advisors, however, suggests that sequential treatment would be more likely to be used in the UK.
• There is currently only limited evidence on the comparative effectiveness of combination therapy (SIRT combined with a systemic therapy). The searches of trial registration databases completed as part of the clinical effectiveness review, however, identified that a large RCT, STOP-HCC,⁷⁴ is set to report shortly. This RCT compares TheraSphere plus sorafenib with sorafenib alone and will provide new evidence on this comparison.
• In the NHS, systemic therapies are recommended only for those with Child–Pugh class A liver function; thus, the current standard of care for those with Child–Pugh class B liver function is BSC. There is a potential place for SIRT in a Child–Pugh class B7 population, who were represented in in the SARAH¹⁹ and SIRveNIB²¹ trials. However, there is currently no direct evidence on the comparative effectiveness of SIRT with BSC in this population, and currently no means of comparing them indirectly.