Statement of principal findings
Clinical effectiveness results
The main sources of clinical effectiveness evidence were two good-quality RCTs (SELECT51 and DECISION52). Results from these trials show that treatment with either lenvatinib or sorafenib statistically significantly improves median PFS and ORR when compared with placebo. Median OS results demonstrate that there is no statistically significant difference in effect when treatment with lenvatinib or sorafenib are compared with placebo. Treatment crossover confounds the OS results from both trials and, to adjust for this effect, OS data were modified using RPSFTM. The results from the adjusted analyses show that, when compared with placebo, treatment with lenvatinib statistically significantly improves OS but there is still no statistically significant improvement in OS from treatment with sorafenib. However, the AG considers that the assumption of PH for unadjusted OS, adjusted OS and PFS is violated in SELECT and is violated for adjusted OS and PFS in DECISION; therefore, these results should be interpreted with caution. Nonetheless, clinical advice to the AG is that the improvements in PFS and the benefits from active treatment do appear to be clinically meaningful.
The AG considers that the improvements in OS and PFS for patients treated with lenvatinib and sorafenib when compared with placebo are likely to reflect improvements in OS and PFS when compared with BSC, notwithstanding the possible differences in the BSC received by the patients in the two trials.
The AG highlights that differences exist between the median OS and PFS results from the observational studies,59,77,78,81,88,101,103,126,135,137 and those from SELECT and DECISION: OS for patients treated with lenvatinib and sorafenib in SELECT and DECISION was longer than the OS reported in the observational studies. In contrast, in DECISION, PFS for patients treated with sorafenib was shorter than for in any of the prospective observational studies and the two meta-analyses.127,138 Median PFS for patients treated with lenvatinib in SELECT was longer than suggested by the prospective, observational results from Study 20177 and shorter than in Study 208.135
Results from indirect comparisons and MAICs7,8,57,97 show that treatment with lenvatinib leads to better PFS (but not OS) than treatment with sorafenib. The AG did not conduct an indirect comparison as preliminary analyses suggested that using data from SELECT and DECISION in the same network would generate unreliable results. The AG’s preliminary analyses showed that the PFS risk profiles (as demonstrated by a comparison of K–M data) of the SELECT and DECISION populations receiving placebo were not comparable. In addition, results from the AG’s analyses showed that, within SELECT and DECISION, the PH assumption did not hold for the majority of survival outcomes. For data to be included in a network, the assumption of PH should hold both across trials and within trials. The AG’s analyses have demonstrated that this assumption is often violated. As a consequence of this violation, the AG has been unable to compare lenvatinib with sorafenib. The AG considers that the relative clinical effectiveness of these two drugs cannot currently be reliably determined.
As expected, both treatment with lenvatinib and with sorafenib resulted in more AEs than treatment with placebo. Both all-grade and grade ≥ 3 diarrhoea were common for patients treated with lenvatinib and those treated with sorafenib. However, the most common AE experienced by patients treated with lenvatinib was hypertension and the most common AE experienced by patients treated with sorafenib was hand–foot syndrome. Dose reductions were frequent for patients treated with lenvatinib (67.8%) and for patients treated with sorafenib (64.3%). The results of published indirect comparisons97 suggest that when treatment with sorafenib is compared with lenvatinib, the incidence of alopecia is higher but the incidence of hypertension is reduced, and those treated with sorafenib experience fewer grade ≥ 3 AEs, SAEs and withdrawals owing to AEs.
The impact of treatment with lenvatinib on HRQoL was not assessed in SELECT and is, therefore, unknown; this is a limitation of the trial given the difference in the safety profiles for some of the AEs associated with lenvatinib and sorafenib. Sorafenib is reported7,120 to have a ‘mild’ negative impact on patients’ HRQoL, possibly attributable to the high rates of AEs experienced by patients in DECISION.
Cost-effectiveness evidence
The two submitting companies and the AG agree that there are no published cost-effectiveness studies relevant to the decision problem set out in the final scope issued by NICE.53 The AG considered that none of the cost-effectiveness studies identified via the AG’s literature review were carried out from a NHS England perspective and that, when treatment with lenvatinib and sorafenib were compared, the results were based on the results of flawed indirect comparisons. In addition, the prices of the drugs reported in the studies were generally not consistent with the discounted prices that will likely be charged in the NHS in England. As a result of the absence of relevant published evidence, the AG developed a de novo cost-effectiveness model for the specific purpose of this appraisal and carried out several cost-effectiveness comparisons.
As the AG did not consider that it was appropriate to carry out an indirect comparison, the AG compared the cost-effectiveness of treatment with lenvatinib with the cost-effectiveness of BSC (using data from SELECT) and the cost-effectiveness of treatment with sorafenib with the cost-effectiveness of BSC (using data from DECISION). The AG also compared the cost-effectiveness of each of the SELECT and DECISION intervention drugs with BSC data from the other trial as a sensitivity analysis.
In the AG’s base-case analysis, using list prices only, the comparison of treatment with lenvatinib versus BSC yields an ICER per QALY gained of £65,872, and the comparison of treatment with sorafenib versus BSC yields an ICER per QALY gained of £85,644. The base-case deterministic and probabilistic results were similar for both comparisons. The AG’s deterministic sensitivity analysis involved varying 18 parameters; the results showed that none of the variations lowered the AG’s base-case ICERs to < £50,000 per QALY gained.
When the AG compared the cost-effectiveness of treatment with lenvatinib with the cost-effectiveness of BSC (using placebo data from SELECT) and the cost-effectiveness of treatment with sorafenib with the cost-effectiveness of BSC (placebo data from DECISION), the ICERs per QALY gained were approximately doubled (£130,592) and halved (£41,716), respectively. These results confirm that the choice of BSC comparator is hugely influential in this appraisal.
Strengths and limitations of the assessment
Strengths
A key strength of this review is that it has brought together all of the available relevant evidence (RCTs, observational studies, systematic reviews, indirect comparisons and cost-effectiveness studies) for assessing the clinical and cost-effectiveness of treatment with lenvatinib versus sorafenib in patients with RR-DTC.
The wide array of clinical results available demonstrate that treatment with lenvatinib is more effective when compared with placebo/BSC for all patients and that prior VEGFR-targeted therapy (or even a treatment delay) does not influence the potential for a patient to benefit from treatment.
Another strength of the research is the AG’s detailed investigation of the PFS (and OS) risk profiles of the patients in the two main trials. The AG’s analytical critique shows that the assumptions of PH underpinning the indirect comparison calculations are violated and explains why data from these two trials should not be compared in an indirect comparison. The AG’s critique challenges the validity of published indirect comparison results7,8,57,97 as well as those from published economic evaluations7,8,38,160,162 that have used indirect comparison results in their analyses.
The results from the AG’s economic analyses demonstrate that the choice of BSC comparator has a big influence on the size of the estimated ICERs per QALY gained.
Limitations
The main limitation of this review is that the AG was unable to compare the clinical effectiveness and cost-effectiveness of lenvatinib with those of sorafenib. The AG did not consider that it was appropriate to conduct an indirect comparison because of key differences in the intervention and placebo arms of SELECT and DECISION (both within and across the trials) and because the results of AG analyses demonstrated that the risk profiles of the patients in the placebo arms were different. Therefore, the AG concluded that it was not possible to determine the comparative clinical effectiveness and cost-effectiveness of lenvatinib versus sorafenib; this is problematic as lenvatinib and sorafenib are two relatively new treatments that appear to work well compared with placebo/BSC for patients with RR-DTC who have limited treatment options.
Uncertainties
Although it is recommended4,23–25 that only patients who are symptomatic and/or have rapidly progressing disease are treated with lenvatinib or sorafenib, it is unclear how many patients in SELECT and DECISION met these criteria. As there are no universally accepted objective criteria for describing patients who are symptomatic and/or rapidly progressing, it is difficult to retrospectively identify these groups of patients with any confidence.
Therefore, it is unclear whether or not the efficacy findings from SELECT and DECISION differ in patients who are symptomatic and/or rapidly progressing compared with those who are not. It is also unknown whether or not the frequency and type of AEs differ between these groups of patients and/or whether patient HRQoL is also influenced by symptom status.
There is considerable uncertainty around the HRQoL of patients with RR-DTC in general. Although it appears that treatment with sorafenib may have a ‘mild’ negative impact on HRQoL, the HRQoL data collected during DECISION were limited. As HRQoL data were not collected as part of SELECT, the impact of treatment with lenvatinib on HRQoL, whether positive or negative, is unknown. To what extent a patient’s HRQoL is affected by their symptom status (symptomatic vs. asymptomatic) is also unknown.
Although, for patients with RR-DTC, RCT evidence has shown clinically meaningful improvements in PFS for those treated with lenvatinib and sorafenib compared with placebo, the question remains whether treatment with lenvatinib or sorafenib can deliver a true OS benefit to patients. The adjusted RPSFTM OS estimates suggest that this may be the case for patients treated with lenvatinib but not for patients treated with sorafenib.
Other relevant factors
The AG considers that it is important to reiterate that the cost–utility analyses presented in this MTA report are based on list prices only. As lenvatinib has a confidential PAS price and sorafenib has a confidential Commercial Unit Access price, the cost-effectiveness comparisons presented in this AG report cannot be used as the basis for decision-making. The AG provided cost-effectiveness results generated using the discounted prices for lenvatinib and sorafenib in a confidential appendix presented to NICE.