Obinutuzumab in combination with chlorambucil for untreated chronic lymphocytic leukaemia

Clinical effectiveness

3.1 The company identified 1 relevant randomised controlled trial to include in its submission. The CLL11 trial was a multicentre, open‑label, 3‑arm trial that compared obinutuzumab plus chlorambucil, rituximab plus chlorambucil and chlorambucil alone in patients with untreated chronic lymphocytic leukaemia for whom full‑dose fludarabine‑based therapy was not appropriate.

3.2 Patients in CLL11 had untreated chronic lymphocytic leukaemia needing treatment (that is, those with Binet stage C or symptomatic disease). Patients who were eligible for the trial had either a total cumulative illness rating scale score greater than 6 or a creatinine clearance of less than 70 ml/minute, or both; no evidence of bone marrow dysfunction other than that caused by chronic lymphocytic leukaemia (determined by an absolute neutrophil count of 1.5×10⁹/litre or greater and platelet count of 75×10⁹/litre or greater); and a life expectancy greater than 6 months. The cumulative illness rating scale calculates the number and severity of chronic illnesses in patients with comorbidities.

3.3 There were 2 stages of recruitment to CLL11. In stage 1, 589 patients were randomised in a 2:2:1 ratio to have obinutuzumab plus chlorambucil, rituximab plus chlorambucil, or chlorambucil alone. In stage 2, an additional 192 patients were randomised to either the obinutuzumab plus chlorambucil group or the rituximab plus chlorambucil group. The stage 1 analysis compared obinutuzumab plus chlorambucil with chlorambucil alone, and rituximab plus chlorambucil with chlorambucil alone. The stage 2 analysis compared obinutuzumab plus chlorambucil with rituximab plus chlorambucil.

3.4 Patients in each of the 3 treatment groups had a dose of chlorambucil on days 1 and 15 of cycles 1–6 equivalent to 0.5 mg/kg body weight (up to a maximum of the dose associated with a BMI of 35 kg/m²). Patients in the obinutuzumab plus chlorambucil treatment group also had 1000 mg of obinutuzumab on days 1, 8 and 15 of cycle 1 and on day 1 of cycles 2–6. Patients in the rituximab plus chlorambucil group also had 375 mg/m² rituximab on day 1 of cycle 1 and 500 mg/m² rituximab on day 1 of cycles 2–6. Each treatment cycle lasted 28 days.

3.5 The primary outcome of CLL11 was progression‑free survival as assessed by the investigator. This was defined as the time from randomisation to the first occurrence of progression, relapse or death from any cause. The analysis of the primary end point used an intention‑to‑treat population. Median progression‑free survival and 95% confidence intervals were estimated using Kaplan–Meier survival methodology. Based on a data cut‑off of March 2014, there were statistically significant improvements in median investigator‑assessed progression‑free survival:

3.6 The secondary outcomes in CLL11 were progression‑free survival as assessed by an independent review committee; overall survival; event‑free survival (time before disease progression or relapse, death, or start of a new anti‑leukaemic therapy); disease‑free survival; duration of response; time to re‑treatment or new anti‑leukaemic therapy; end of treatment response (response occurring more than 56 days after the end of treatment); best overall response; best overall response within 1 year of start of study treatment; molecular remission; rate of negative testing for minimal residual disease; adverse events and patient‑reported outcomes. The results for the secondary outcomes at the end of stage 1 showed that obinutuzumab plus chlorambucil and rituximab plus chlorambucil were statistically significantly better than chlorambucil alone for most outcomes. At the end of stage 2, obinutuzumab plus chlorambucil was statistically significantly better than rituximab plus chlorambucil for most of the secondary outcomes. Overall survival was statistically significantly greater for obinutuzumab plus chlorambucil than for rituximab plus chlorambucil and chlorambucil alone at the end of stage 1. There was no statistically significant difference in overall survival between obinutuzumab plus chlorambucil and rituximab plus chlorambucil at the end of stage 2. However, the company stated that the overall survival data were immature. Deaths and disease‑free survival rate were not statistically significantly different between the groups at the end of stage 2.

3.7 The median observation times at the March 2014 data cut‑off were:

3.8 No direct evidence comparing obinutuzumab and bendamustine was identified by the company. To compare these 2 treatments, the company created 2 network meta‑analyses of randomised controlled trials (a large network and a small network). The large network included studies regardless of whether full‑dose fludarabine therapy was suitable for the enrolled patients (n=17, including CLL11). The results of the large network meta‑analysis showed that the mean progression‑free survival hazard ratio statistically significantly favoured obinutuzumab plus chlorambucil compared with bendamustine (HR 0.399, 95% CI 0.218 to 0.672, fixed‑effects model, adjusted for age; HR 0.546, 95% CI 0.367 to 0.783, fixed‑effects model, not adjusted for age). The small network only included studies of patients for whom fludarabine‑based therapy was unsuitable (n=3), in line with obinutuzumab's licensed indication. The small network meta‑analysis did not compare obinutuzumab with bendamustine because no studies of bendamustine monotherapy were included in the network. The network meta‑analyses did not allow the calculation of the hazard ratio for obinutuzumab plus chlorambucil compared with bendamustine plus rituximab. This was because the results of MaBLe (an ongoing trial comparing rituximab plus bendamustine with rituximab plus chlorambucil) had not been published and so were not included in the large network meta‑analysis. Because the hazard ratio for obinutuzumab plus chlorambucil compared with bendamustine plus rituximab was needed to inform the cost‑effectiveness model, the company estimated a hazard ratio of 0.68 based on data from CLL11 at the March 2014 data cut‑off and the power calculation assumptions from the ongoing MaBLe trial, and used this value in its base case.

3.9 The company carried out safety analyses on data from all patients who had at least 1 dose of study medication in CLL11. The most frequent adverse events were infusion‑related reactions, neutropenia and nausea. The most frequent adverse events of grade 3 or higher were infusion‑related reactions, neutropenia and anaemia. The most frequent serious adverse events were infection, neoplasm and infusion‑related reactions. The incidence of adverse events, serious adverse events and adverse events leading to discontinuation of study treatment was higher in the obinutuzumab plus chlorambucil arm than in either of the other groups. The differences were mainly a result of infusion‑related reactions. However, the company did not report whether the differences were statistically significant. Obinutuzumab treatment was associated with increases in common chlorambucil‑related adverse events (neutropenia, thrombocytopenia, anaemia). These events were mainly mild‑to‑moderate in severity, easily managed, and rarely led to discontinuation of all treatment.

3.10 Infusion‑related reactions occurred in 166 patients (69%) in the obinutuzumab plus chlorambucil group and 88 patients (39%) in the rituximab plus chlorambucil group in stage 2 of CLL11. Most infusion‑related reactions were low grade in intensity and were clinically manageable. No deaths were associated with infusion‑related reactions. There were 21 deaths (6.5%) due to adverse events in the rituximab plus chlorambucil group, 15 deaths (4.5%) due to adverse events in the obinutuzumab plus chlorambucil group and 11 deaths (9%) due to adverse events in the chlorambucil alone group. The company did not report whether the differences between the groups in the number of deaths due to adverse events were statistically significant.

Cost effectiveness

3.16 Roche identified 1 published cost‑effectiveness model of patients with chronic lymphocytic leukaemia (Walzer et al. 2013) that was relevant to this appraisal. Roche had developed this model and updated the published version to include in this appraisal.

3.17 The company's model evaluated the cost effectiveness of obinutuzumab plus chlorambucil compared with rituximab plus chlorambucil, bendamustine plus rituximab, bendamustine alone and chlorambucil alone. The model consisted of 3 health states, namely 'progression‑free survival', 'progressed', and 'death', with the progression‑free survival health state further divided into 'on therapy' and 'not on therapy'. The model had weekly cycles and a half‑cycle correction was applied, except to the drug, administration and pharmacy costs. The model used a lifetime time horizon (maximum 20 years) and a discount rate of 3.5% per year for costs and quality‑adjusted life years (QALYs).

3.18 All people in the model started in the progression‑free survival health state. At the end of each weekly cycle, people in the progression‑free survival health state either remained there, moved to the progressed health state, or died. People in the progressed health state either remained in the progressed health state or died. Once they moved to a different state in the model, people could not return to the previous health state. The company used data from the CLL5 trial (a randomised controlled trial comparing fludarabine with chlorambucil in untreated chronic lymphocytic leukaemia) to model overall survival distribution, because the overall survival data from CLL11 were immature. The overall survival distribution in the model was validated using the Kaplan–Meier overall survival data that were available from CLL11.

3.19 The company calculated the number of people in the progression‑free survival health state using data from CLL11 for obinutuzumab plus chlorambucil, rituximab plus chlorambucil, and chlorambucil alone. For bendamustine, the company took data from the large network meta‑analysis and it used its own estimated hazard ratio for rituximab plus bendamustine (see section 3.8). It modelled the transition from the progression‑free survival health state using data from CLL11 and background mortality. The proportion of people in the progressed health state in each cycle was the difference between the proportion of people who were alive and the proportion of people who were progression free.

3.20 The company applied costs from the perspective of the NHS and personal social services for drug acquisition, drug administration, health state and adverse events. It made an assumption of no vial sharing for all intravenous drugs (obinutuzumab, rituximab and bendamustine); therefore all calculations of price include drug wastage. The drug costs per treatment course were £26,496 (£3312 per 1000‑mg vial) for obinutuzumab, £9953.91 (£174.63 per 100‑mg vial and £873.15 per 500‑mg vial) for rituximab, and £5809.92 (£69.45 per 25‑mg vial and £275.81 per 100‑mg vial) for bendamustine. The cost per treatment course of chlorambucil was £369.45 (£40.51 per pack of 25 tablets of 2 mg each). European Society of Medical Oncology guidelines informed resource use in the progression‑free and progressed health states, and this was validated with clinical experts at an advisory board. The progression‑free survival health state assumed 1 outpatient appointment lasting 60 minutes every 3 months and the post‑progression state assumed 1 outpatient appointment every month.

3.21 Adverse events were included in the model if they were grade 3, 4 or 5 and occurred in 2% or more people in CLL11 (obinutuzumab plus chlorambucil, rituximab plus chlorambucil or chlorambucil alone), Knauf et al. (2009; bendamustine alone), or the MaBLe trial (rituximab plus chlorambucil and rituximab plus bendamustine). Because there was a lack of complete data for bendamustine plus rituximab from the MaBLe study, the company assumed that the frequency and cost of adverse events were the same as for rituximab plus chlorambucil in stage 2 of CLL11. The company applied the total cost of all adverse events in each treatment group as a one‑off event in the first cycle of each Markov state.

3.22 Quality‑of‑life data were collected in CLL11 using the European Organisation for Research and Treatment of Cancer Quality of Life – Core 30 (EORTC‑QLQ‑C30) questionnaire. However, the company did not map these data to the EuroQol (EQ‑5D) questionnaire and it did not use the quality‑of‑life data from CLL11 in its health economic model. To determine relevant utility values, the company did a utility elicitation study with a sample of 100 members of the UK general public. The study used health state descriptions to explore societal preferences for quality of life associated with chronic lymphocytic leukaemia. The company used utility values from the study in the model; these were: 0.71 for progression‑free survival on oral treatment (chlorambucil); 0.67 for progression‑free survival on intravenous treatment (rituximab and bendamustine); 0.55 for progression‑free survival on initial therapy with increased hospital visits (obinutuzumab); 0.82 for progression‑free survival after initial treatment was completed (all treatment arms); and 0.60 for progressed disease (all treatment arms). The company assumed health‑related quality of life to be constant over time within each health state in the model.

3.23 The company's base case and incremental results (without the patient access scheme) are presented in tables 1 and 2.

Company's response to consultation

3.36 In response to consultation, the company requested that the clinical and cost effectiveness of obinutuzumab plus chlorambucil for people who cannot have fludarabine‑based therapy be considered in 2 different subgroups: those who can have bendamustine‑based treatment and those who cannot. The company highlighted that these groups are distinct populations because not all people covered by the scope can have bendamustine‑based therapy because of comorbidities. The company submitted new cost‑effectiveness estimates for the 2 subgroups, which included the following revisions to its economic model:

3.37 The company presented revised cost‑effectiveness results for the population who could not have bendamustine‑based treatment. The details of the patient access scheme are confidential and therefore the ICERs cannot be presented here because, having previously released the estimates without the patient access scheme, the estimates with the patient access scheme could reveal the confidential discount agreed between the company and the Department of Health. However, the estimates including the patient access scheme were fully taken into account during the appraisal. The revised base‑case ICERs for obinutuzumab plus chlorambucil compared with both chlorambucil alone and with rituximab plus chlorambucil were between £20,000 and £30,000 per QALY gained. In its sensitivity analysis, the company applied an alternative utility value for progression‑free survival while off treatment of 0.76 (instead of 0.71). This had the effect of decreasing the ICERs. In the population who could not have bendamustine‑based treatment, the company stated that the probability of obinutuzumab being considered cost effective was 74.4% at £30,000 per QALY gained. If the model used an alternative progression‑free survival utility value of 0.76, this probability increased to 87.9% at £30,000 per QALY gained.

3.38 The company presented revised cost‑effectiveness results for the population who could have bendamustine‑based treatment. In amending its model to include all the changes mentioned in section 3.36, the ICERs for obinutuzumab plus chlorambucil compared with both bendamustine alone and with rituximab plus bendamustine were all above £30,000 per QALY gained. In this population, the probability of obinutuzumab plus chlorambucil being considered cost effective was less than 10% at £30,000 per QALY gained. In its sensitivity analysis, the company applied an alternative utility value for progression‑free survival while off treatment of 0.76 (instead of 0.71). Again, this had the effect of decreasing the ICERs, but they remained above £30,000 per QALY gained.

Evidence Review Group's critique of the company's response to consultation

3.39 The ERG agreed with the company that it is appropriate to consider 2 subgroups in this appraisal, that is, people who can have bendamustine‑based treatment and those who cannot.

3.40 The ERG highlighted that the company's updated costs of treating neutropenia were based on those estimated from NICE's ongoing technology appraisal of erlotinib and gefitinib for treating non-small-cell lung cancer that has progressed after prior chemotherapy (review of NICE technology appraisal guidance 162 and 175). The ERG commented that the evidence to support this change was inadequate and more justification was needed. However, the ERG stated that this issue was of minor importance because it had little effect on the company's cost‑effectiveness estimates (see section 3.42).

3.41 The ERG commented that the alternative utility value for progression‑free survival off treatment (0.76 rather than the ERG's preferred value of 0.71) was from the randomised controlled trial COMPLEMENT‑1. It commented that the trial was relevant because the patients were representative of the patient populations being considered in this appraisal. The ERG highlighted that it still preferred the utility value of 0.71 for progression‑free survival off treatment. This was because the data from COMPLEMENT‑1 were collected during the first half of the time in progression‑free survival, whereas this appraisal was interested in the mean utility value 6 months from the start of progression‑free survival (end of treatment) until the end of progression‑free survival. Therefore, the utility data from COMPLEMENT‑1 cited by the company only partly represented the time period of interest. The ERG stated that the utility value of 0.76, estimated from Ara and Brazier (2010) using the EQ‑5D, was from a large study of the general population in England. The ERG highlighted that it was logical that the utility value appropriate for patients with multiple comorbidities with chronic lymphocytic leukaemia could be less than 0.76.

3.42 The ERG confirmed that the amendments to the company's base case had been done correctly. The ERG did scenario analyses implementing all the company's revised assumptions, but excluding the change in costs of treating neutropenia, and using a utility value for progression‑free survival off treatment of either 0.71 or 0.76. The ICERs are confidential and cannot be reported here, but the effect of changing the costs of treating neutropenia was marginal. Although using the alternative source of the utility value for progression‑free survival off treatment had a more moderate effect, overall, the ICERs for obinutuzumab plus chlorambucil compared with both chlorambucil alone and with rituximab plus chlorambucil remained between £20,000 and £30,000 per QALY gained. For obinutuzumab plus chlorambucil compared with both bendamustine alone and with rituximab and bendamustine, the corresponding ICERs remained above £30,000 per QALY gained.

3.43 Full details of all the evidence are available.