3 Evidence
The appraisal committee (section 5) considered evidence submitted by Eli Lilly and Company and a review of this submission by the evidence review group (ERG). See the committee papers for full details of the evidence.
Clinical effectiveness
3.1 The company submission considered 2 populations; the full population (including people with squamous and non-squamous non-small-cell lung cancer [NSCLC]) and a subgroup of people with non-squamous NSCLC.
3.2 The company's systematic review identified 1 relevant randomised controlled trial: REVEL. This was a phase III, international, multicentre, randomised, placebo-controlled, double-blind trial investigating ramucirumab plus docetaxel (n=628) compared with placebo plus docetaxel (the docetaxel-alone group; n=625) in adults with stage IV NSCLC whose disease had progressed during or after platinum-based therapy for locally advanced or metastatic disease.
3.3 The primary outcome was overall survival; secondary outcomes included progression-free survival, objective response rate, disease control rate and safety and quality of life as captured by the Lung Cancer Symptom Scale (LCSS) and the EQ‑5D health questionnaire.
3.4 In the full population, compared with docetaxel alone, ramucirumab plus docetaxel:
-
improved overall survival by 1.4 months (hazard ratio [HR] 0.86; 95% confidence interval [CI] 0.75 to 0.98; p=0.024) and
-
improved progression-free survival by 1.5 months (HR 0.76; 95% CI 0.68 to 0.86; p<0.0001).
In the subgroup of patients with non-squamous disease, compared with docetaxel alone, ramucirumab plus docetaxel: -
improved overall survival by 1.4 months (HR 0.83; p=0.02) and
-
improved progression-free survival by 0.9 months (HR 0.77; p<0.001).
In the subgroup of patients with squamous disease, compared with docetaxel alone, ramucirumab plus docetaxel: -
improved overall survival by 1.3 months (HR 0.88; p=0.319) and
-
improved progression-free survival by 1.5 months (HR 0.76; p=0.019).
3.5 The company reported that the percentage of patients who had at least 1 adverse event of any grade during treatment was similar between treatment arms: 97.8% in the ramucirumab plus docetaxel group compared with 96.1% in the docetaxel-alone group. Fatigue, neutropenia and febrile neutropenia were the grade 3 or higher adverse events that occurred during treatment in more than 10% of patients.
3.6 The company did a network meta-analysis to estimate the relative treatment effect of ramucirumab plus docetaxel compared with nintedanib plus docetaxel for the subgroup of patients with non‑squamous disease, using data from REVEL and the LUME‑Lung 1 trial. LUME‑Lung 1 compared nintedanib plus docetaxel with docetaxel alone. The company's analyses assumed that the histologies of non‑squamous NSCLC and adenocarcinoma were the same given that nintedanib plus docetaxel is licensed specifically for adenocarcinoma. Hazard ratios for overall survival and progression-free survival were calculated using a Bayesian network meta-analysis and assuming proportional hazards. The results did not show any difference between ramucirumab plus docetaxel and nintedanib plus docetaxel (overall survival HR 1.01; 95% CI 0.82 to 1.25, progression-free survival HR 0.99; 95% CI 0.78 to 1.26).
Cost effectiveness
3.7 The company presented a de novo, partitioned survival economic model based on 3 health states; pre-progression, post-progression and death. Patients remained in the pre-progression state until disease progression or death. In the post-progression state patients had either best supportive care or post-progression treatments.
3.8 Five parametric models were used to consider goodness of fit to the overall survival and progression-free survival data from REVEL. Curves were fitted to both the adjusted (taking into account a number of covariates) and unadjusted Kaplan–Meier data. However the company considered that the adjusted models provided the best fit and used them in its base case.
3.9 For overall survival the company considered that the proportional hazards assumption (that is, the relative risk of an event is fixed irrespective of time) held. Therefore a single parametric curve was fitted to the entire data set with treatment included as a covariate. The company chose a log‑logistic distribution to extrapolate overall survival in its base case.
3.10 For progression-free survival the company noted that the proportional hazards assumption was violated. Therefore the company generated separate parametric curves for ramucirumab plus docetaxel and docetaxel alone and considered that the generalised gamma model provided the best fit for both treatment groups.
3.11 For comparing ramucirumab plus docetaxel with nintedanib plus docetaxel, the company applied its network meta-analysis hazard ratio to the docetaxel-alone curves from REVEL to estimate overall survival for nintedanib plus docetaxel, and used the adjusted log‑logistic model for ramucirumab plus docetaxel.
3.12 The company's deterministic base-case incremental cost-effectiveness ratio (ICER) for ramucirumab plus docetaxel compared with docetaxel alone for the full population was £194,919 per quality-adjusted life year (QALY) gained. The company's deterministic base-case ICER for ramucirumab plus docetaxel compared with nintedanib plus docetaxel for the subgroup of patients with non-squamous NSCLC was £1,106,497 per QALY gained (excluding the nintedanib patient access scheme).
3.13 The company carried out a number of scenario analyses for both populations (see the company submission for more details). For the full population, the ICERs for ramucirumab plus docetaxel compared with docetaxel alone were between £189,068 and £230,272 per QALY gained. For the subgroup of patients with non-squamous disease, the company's scenario analyses ranged from ramucirumab plus docetaxel being dominated (that is, more expensive and less effective) by nintedanib plus docetaxel to £1,246,442 per QALY gained (excluding the nintedanib patient access scheme).
Evidence review group key issues
3.14 The ERG considered that REVEL was good quality and accurately presented the risks and benefits of ramucirumab plus docetaxel compared with docetaxel alone.
3.15 The ERG was concerned that the company used the population from REVEL with non‑squamous disease rather than the population with adenocarcinoma when comparing ramucirumab plus docetaxel with nintedanib plus docetaxel. However, when the ERG compared the overall survival curves for the non‑squamous and adenocarcinoma groups from REVEL they appeared to have some similarities. The ERG also found similarities in the progression-free survival data and therefore considered that the inconsistency in the populations compared would have little effect on the cost-effectiveness results.
3.16 The ERG noted that although the company's log‑logistic model provided a good fit for the ramucirumab plus docetaxel group, the fit for the docetaxel-alone group was poor. From approximately 10 months onwards the docetaxel-alone log‑logistic curve underestimated the observed overall survival in the Kaplan–Meier plot. The ERG considered that if the curve was fitted to any comparator (docetaxel alone or nintedanib plus docetaxel) of ramucirumab plus docetaxel it would underestimate the efficacy of the comparator. Therefore, separate curves should be fitted to the groups.
3.17 The ERG was concerned about using the network meta-analysis hazard ratios to model overall survival and progression-free survival because:
-
they imposed proportional hazards between compared treatments
-
they forced a log‑logistic curve shape onto the comparator, which was unlikely to reflect the observed data
-
they attached the generated curve on the time axis (of the survival curve plot) according to the position of the REVEL docetaxel survival curve and
-
the log‑logistic model could be an inaccurate estimate of the intervention and comparators.
The ERG considered that the resulting survival curves may not represent the situation fully. It therefore explored using a linear trend model to estimate overall survival in scenario analyses (see the ERG report for more details).
3.18 The ERG noted that the company assumed that quality of life was the same in each group while on treatment (that is, the company pooled the EQ‑5D values from the trial) but made small allowances for different side effects. The ERG also had some concerns about the way the company had calculated the cost of ramucirumab based on the average number of weeks of treatment rather than the average number of ramucirumab doses. However the ERG did not consider that this significantly affected the ICER.
3.19 The ERG made some adjustments to the company's base-case, resulting in an ICER of £175,000 per QALY gained for ramucirumab plus docetaxel compared with docetaxel alone for the full population. The ERG's adjustments to the company's base-case produced an ICER for ramucirumab plus docetaxel compared with nintedanib plus docetaxel of £1,600,000 per QALY gained for the subgroup of patients with non‑squamous NSCLC (excluding the nintedanib patient access scheme).
3.20 The ERG carried out a number of scenario analyses for both populations. For the full population, the ICERs for ramucirumab plus docetaxel compared with docetaxel alone were between £167,000 and £247,000 per QALY gained, with an ICER of £177,000 when the linear trend model was used to estimate overall survival. For the subgroup of patients with non‑squamous disease, the ERG's scenario analyses ranged from ramucirumab plus docetaxel being dominated by nintedanib plus docetaxel to £1,900,000 per QALY gained (excluding the nintedanib patient access scheme). When the ERG included the nintedanib patient access scheme (confidential simple discount), this increased the ICERs further. The ERG also carried out a scenario analysis using a linear trend model for overall survival in the subgroup of patients with squamous disease. This resulted in an ICER of £167,000 per QALY gained for ramucirumab plus docetaxel compared with docetaxel alone.
3.21 The ERG applied the linear trend model to the REVEL results for patients receiving docetaxel alone. Overall survival was:
-
14.4 months (full population)
-
15.32 months (subgroup with non‑squamous disease) and
-
11.19 months (subgroup with squamous disease).
Comparing ramucirumab plus docetaxel with docetaxel alone, the linear trend model showed a mean extension in overall survival of: -
2.2 months (full population)
-
3.9 months (subgroup with non‑squamous disease) and
-
1.1 months for the population with squamous disease.
Comparing ramucirumab plus docetaxel with nintedanib plus docetaxel for the subgroup of patients with non‑squamous disease, the mean extension in overall survival was 0.16 months, with less gain if only the adenocarcinoma population was considered.