Idelalisib for treating chronic lymphocytic leukaemia

Systematic review

3.1 The company's systematic literature review identified 6 randomised controlled trials that were relevant to the decision problem because they included patients whose disease was relapsed or refractory. The company noted that none of the trials identified in the literature review compared idelalisib plus rituximab directly with the appropriate comparators identified in the NICE scope.

3.2 One single arm study was identified that investigated the effectiveness of idelalisib with rituximab in untreated chronic lymphocytic leukaemia in patients with a 17p deletion or TP53 mutation.

Previously treated chronic lymphocytic leukaemia (Study 116)

3.3 The company presented the results of Study 116, which was a phase III, double‑blind, randomised controlled trial across 90 centres in the US and Europe (including the UK). The study evaluated idelalisib plus rituximab compared with rituximab plus placebo in people with chronic lymphocytic leukaemia. A total of 220 patients were randomised to either idelalisib (150 mg oral tablets, twice daily) plus rituximab (375 mg/m² at week 0, then 500 mg/m² at weeks 2, 4, 6, 8, 12, 16 and 20) or rituximab (same dose) plus placebo (matching tablet, twice daily, until progression or withdrawal due to tolerability issues). Patients were included if they were aged 18 years or older, had previously had at least 1 treatment line (either an anti‑CD20 or 2 or more cytotoxic regimens) and had a reported Karnofsky performance score of 40 or more. Patients were excluded if their disease had a known histological transformation from chronic lymphocytic leukaemia to a more aggressive lymphoma.

3.4 The primary outcome of Study 116 was progression‑free survival, defined as the interval from randomisation to first documentation of definitive disease progression or death from any cause (whichever was sooner). Definitive disease progression was defined using the criteria from the International Workshop on Chronic Lymphocytic Leukaemia (IWCLL). Secondary outcomes included rates of overall response (complete and partial), lymph node response (defined as a decrease of 50% or more in lymphadenopathy) and overall survival. The primary and secondary end points were examined in pre‑specified subgroups, including: the presence or absence of a 17p deletion, TP53 mutation or both; sex (male or female); and age (less than 65 years or more than 65 years old). Health‑related quality of life was assessed using a change in domain and symptom scores from the Functional Assessment of Cancer Therapy: Leukaemia (FACT‑Leu) instrument, and using the EQ‑5D instrument. These were administered at baseline and at each study visit.

3.5 Patients could cross over from the rituximab plus placebo group to having idelalisib monotherapy in an extension study if their disease progression were confirmed by an independent review committee. The intention‑to‑treat analysis was done according to the treatment to which patients were randomised, and this included patients who had crossed over to the idelalisib monotherapy group. Progression‑free survival was calculated using the Kaplan–Meier method. A Cox proportional hazards model with adjustment for stratification was used to calculate hazard ratios.

Untreated chronic lymphocytic leukaemia with a 17p deletion or TP53 mutation (Study 101‑08)

3.14 Study 101‑08 was a phase II, single‑arm study of idelalisib plus rituximab in patients with untreated chronic lymphocytic leukaemia. A total of 64 patients were enrolled in centres across the US to have idelalisib (150 mg oral tablets, twice daily) and rituximab (375 mg/m² for 8 weeks of continuous treatment). Only a small subset of patients (n=9) had a 17p deletion or TP53 mutation.

3.15 The primary outcome of the study was overall response rate, which was defined as the proportion of patients who achieved a complete or partial response as defined according to IWCLL criteria. Secondary outcomes included adverse events, progression‑free survival and overall survival.

Evidence for other comparators listed in the NICE scope

3.19 In addition to the comparison of idelalisib plus rituximab with rituximab alone, best supportive care and ofatumumab, the company also submitted evidence for the other comparator technologies listed in the NICE scope, namely fludarabine cyclophosphamide plus rituximab, bendamustine with or without rituximab, chlorambucil with or without rituximab, and steroids plus rituximab. The company's systematic review did not identify any evidence directly comparing idelalisib with these comparators, but identified 5 randomised controlled trials and 13 non‑randomised controlled trials in the relapsed or refractory population. One randomised controlled trial was identified in the untreated population.

ERG comments

3.20 The ERG commented on the population of Study 116 and its applicability to people with chronic lymphocytic leukaemia in the UK (of the 220 patients enrolled, 32 were UK patients). It noted that the trial included some patients (43.2%) with a 17p deletion or TP53 mutation; this type of disease does not respond to standard chemotherapy. The ERG noted that the baseline characteristics of patients in the trial represented a high‑risk patient cohort. The ERG also noted that the use of rituximab as a comparator was limited in its relevance to a UK population, because it is neither recommended by NICE nor by the British Committee on Standards in Haematology guidance.

3.21 The ERG noted that the results of Study 116 should be interpreted with caution because the trial was stopped early for benefit, and this type of stopping can lead to an overestimation of treatment effect.

Economic model structure

3.22 The company submitted a de novo economic model for the relapsed or refractory chronic lymphocytic leukaemia population only. The company did not submit an economic model for the untreated group because of the limited evidence available from Study 101‑08. The base‑case analysis modelled the following:

3.23 The company included an additional exploratory analysis of idelalisib plus rituximab compared with the other comparators listed in the NICE scope:

3.24 The company used a Markov model with time‑dependent transition probabilities. It used weekly cycle lengths (with half‑cycle corrections) with a time horizon of 25 years. The mean age of patients entering the model was 71 years. A discount rate of 3.5% was applied to costs and health benefits and the analysis was done from an NHS and personal social services perspective.

3.25 The base‑case analysis used data from Study 116. The model consisted of 5 health states, namely pre‑progression on treatment, pre‑progression off treatment, post progression, terminal care and death. The on treatment state was determined by the area under parametric time‑on‑treatment curves. The pre‑progression off treatment and post‑progression states were informed by the parametric survival curve analysis of progression‑free and overall survival data.

Clinical parameters and assumptions

3.26 To inform the clinical parameters for idelalisib plus rituximab compared with ofatumumab, the company assumed equal efficacy for rituximab and ofatumumab. The company explained that this assumption reflected the results of the ORCHARRD study, a network meta‑analysis in patients with diffuse large B‑cell lymphoma which found no difference in efficacy between ofatumumab and rituximab. The company also assumed equal efficacy for rituximab and best supportive care because there were insufficient data available to model this comparator.

3.27 To inform the model transition probabilities, the company extrapolated the overall survival data beyond the trial cut‑off. The company used the crossover‑adjusted overall survival data from Study 116 (patients could crossover from the rituximab plus placebo group to instead have idelalisib plus rituximab). The company used the Akaike Information Criterion statistic (an indication of the statistical fit between the observed Kaplan–Meier data and the parametric model estimates) to assess the most appropriate functional form. The results of the analyses showed that the exponential model provided the most appropriate fit, followed by the Weibull. The company explained that, on inspection of the model, 5% of patients were still alive after 20 years which it deemed inappropriate. The company therefore chose the Weibull model, as the next best fit, to extrapolate the overall survival data.

3.28 The company explained that progression‑free survival data did not need to be adjusted for crossover before extrapolation, because disease progression in the trial was the main reason for patients crossing over to the idelalisib plus rituximab group. The Akaike Information Criterion statistic suggested that the Weibull model was the most appropriate curve to use to extrapolate beyond the trial data.

3.29 The company used time‑on‑treatment data from Study 116 to estimate the drug resource use for idelalisib in the pre‑progression on treatment state. In the trial, idelalisib was indicated to be taken until disease progression or unacceptable toxicity, so the company noted that time on treatment followed a similar course to progression‑free survival. To extrapolate the data, it determined that a Cox proportional hazards model should be used to calculate a hazard ratio for time on treatment compared with progression‑free survival. This produced a hazard ratio of 1.31. It then applied this to the progression‑free survival curve for idelalisib plus rituximab to derive a time‑on‑treatment curve. For consistency, the company used the same hazard ratio to estimate a time‑on‑treatment curve for the rituximab plus placebo group of Study 116.

3.30 The company used the overall response rates reported in Study 116 to inform resource use assumptions in the model. The overall response rates were 84% in the idelalisib plus rituximab group and 15% in the rituximab plus placebo group. For the comparison with best supportive care, patients were assumed to have no overall response. For the comparison with ofatumumab, the overall response rate for rituximab plus placebo was applied.

Table 1 Results of the studies used to inform the company's additional comparator analysis

Utility values and adverse events

3.32 The company used EQ‑5D data collected alongside Study 116 to inform the utility values for patients in the pre‑progression on treatment state. It used a generalised estimation equation regression to determine whether there was a difference in quality of life between the study groups. The company also assumed that utility values in the terminal care state were equal to those in the post‑progression state. However, no EQ‑5D trial data were collected for patients in the post‑progression or post‑treatment discontinuation states, so the company conducted a systematic literature review to identify studies reporting utility values for different chronic lymphocytic leukaemia health states. The company identified a range of studies, and chose to use Dretzke et al. (2010) for the post‑progression and pre‑progression off treatment states.

3.33 The company derived adverse event frequencies directly from Study 116. Those considered in the model were grade 3 or 4 events which occurred in at least 3% of patients in either treatment groups of Study 116.

Resource use and costs

3.34 The company used the time‑on‑treatment curves from Study 116 to estimate the length of time patients would have idelalisib plus rituximab and rituximab alone in the pre‑progression on treatment state. The company used the same dosing regimen from Study 116 for idelalisib plus rituximab and rituximab alone. For the other comparators, treatment costs were applied to the proportion of patients remaining in the progression‑free health states in the model, up to maximum treatment durations. The proportion of modelled patients who progressed before maximum treatment durations were not assumed to complete the full course of treatment.

3.35 Based on expert advice, the company assumed that intravenous immunoglobulin therapy would be used in 45% of non‑responders and 0% of responders in the pre‑progression health states. The company estimated that the cost of intravenous immunoglobulin therapy, incorporating the acquisition cost and administration of 5 weekly infusions, was £13,706. Active treatment was assumed to be every 3.5 weeks, based on British Committee for Standards in Haematology guidelines.

Company's base‑case results and sensitivity analyses

3.36 When using the simple discount agreement price for idelalisib, treatment with idelalisib plus rituximab was associated with higher costs and greater quality‑adjusted life year (QALY) gains compared with rituximab alone. The deterministic incremental cost‑effectiveness ratio (ICER) for idelalisib plus rituximab compared with rituximab alone was £13,634 per QALY gained (incremental costs £26,128; incremental QALYs 1.92). Compared with best supportive care, the ICER for idelalisib plus rituximab was £20,461 per QALY gained (incremental costs £39,211; incremental QALYs 1.92). Compared with ofatumumab, the ICER was £1527 per QALY gained (incremental costs £2926; incremental QALYs 1.92).

3.37 The company conducted a range of deterministic sensitivity analyses on the base‑case parameters. The results showed that the survival curve parameter estimates had the greatest influence on the results.

3.38 The company conducted a probabilistic sensitivity analysis for the base‑case parameters, presenting scatter plots and cost‑effectiveness acceptability curves for the 3 base‑case comparisons. The results showed that idelalisib plus rituximab was cost effective with a 90% probability compared with rituximab alone, an 80% probability compared with best supportive care, and a 100% probability compared with ofatumumab (at a maximum acceptable ICER of £30,000 per QALY gained). The mean probabilistic ICER for the comparison of idelalisib plus rituximab with rituximab alone was £13,680 per QALY gained. For idelalisib plus rituximab compared with ofatumumab, the mean probabilistic ICER was £1692 per QALY gained; for the comparison with best supportive care, it was £20,021 per QALY gained. The mean probabilistic ICER for the comparison of idelalisib plus rituximab with rituximab alone was £13,680 per QALY gained. For the comparison with ofatumumab, the mean probabilistic ICER was £1692 per QALY gained; for the comparison with best supportive care, it was £20,021 per QALY gained.

Company's exploratory scenarios

3.39 When using the simple discount agreement price for idelalisib, treatment with idelalisib plus rituximab was associated with both higher costs and greater QALY gains compared with:

3.40 The company presented a subgroup analysis for patients in Study 116 with a 17p deletion or TP53 mutation. The results showed an increase in the ICERs for all 3 base‑case comparators: £20,200 per QALY gained for the comparison with rituximab alone, £27,543 per QALY gained for the comparison with best supportive care and £7066 per QALY gained for the comparison with ofatumumab.

ERG comments

3.41 The ERG noted the company's assumption that treatment effects continue beyond the trial. The ERG acknowledged that the treatment benefits of idelalisib may continue beyond the time horizon of the trial, but would be unlikely to continue for the rest of a patient's life. It noted that any reduction in treatment benefit following a discontinuation could result in a smaller treatment benefit for idelalisib with rituximab compared with the other regimens.

3.42 The ERG highlighted a possible issue with the company's methodology to adjust for crossover in the trial, noting that the company's analysis relies on the assumption that idelalisib alone has equal efficacy to idelalisib plus rituximab. It stated that a lower efficacy for idelalisib alone would result in a lower treatment benefit for idelalisib plus rituximab compared with rituximab and the magnitude of this change would only be substantial if idelalisib alone is considered inferior to idelalisib plus rituximab.

3.43 The ERG commented on the company's use of the constant shape parameter for the survival curves for the other treatments. It noted that the company could have fitted survival models directly using the digitised Kaplan–Meier plots that were generated. The ERG noted that this would improve the extrapolation for the fludarabine, cyclophosphamide and rituximab data, where the model (assuming a constant shape) was a poor fit for the data. In addition, the ERG questioned why the company had chosen to use hazard ratios from Badoux et al. to adjust for baseline differences in its extrapolations instead of the other studies identified in the literature review.

3.44 The ERG noted that in the company's base case, costs for idelalisib, rituximab and ofatumumab were accrued until treatment discontinuation, but for other comparators patients were assumed to complete the full maximum dosing indicated for that product. The ERG stated that a more realistic approach would be to use time‑on‑treatment data from Study 116 to estimate the proportion of the maximum number of doses actually administered for rituximab monotherapy, in the same way as was done for idelalisib, and apply the costs to these estimated time‑on‑treatment data.

3.45 The ERG noted that because no attempt was made to account for differences between the study populations and UK patients, the results of the company's analyses may be uncertain. In addition, it noted that the plausibility of the results may be affected by assuming that ofatumumab and best supportive care have equal efficacy to rituximab alone. Finally, the ERG stated that even though the methodology used to conduct the analyses for the other comparators in the NICE scope was less reliable than that used in the formal evidence synthesis, many of the comparators are used in the UK and therefore results from these analyses (with appropriately conservative assumptions) are important to understanding the cost‑effectiveness of idelalisib compared with other available treatments.

3.46 The ERG raised concerns with the assumptions relating to the frequency of resource use parameters, specifically about the number of patients whose disease did not respond to treatment being given intravenous immunoglobulin therapy. The ERG noted that this is important because the biggest difference in clinical outcomes between idelalisib plus rituximab and rituximab alone was the overall response rate. This meant that the clinical assumptions made in the model resulted in considerably higher disease management costs for patients having rituximab alone than those having idelalisib plus rituximab.

3.47 The ERG highlighted that utility values for patients who had discontinued treatment with idelalisib plus rituximab (taken from Dretzke et al.) in the pre‑progression states remained higher than those for patients having rituximab alone. It noted that this difference was more than could be explained by the adverse event disutilities calculated from Study 116, and meant that patients discontinuing idelalisib maintained a higher quality of life than those still having rituximab (an assumption not justified in the company submission)

ERG exploratory analyses

3.48 The ERG conducted an exploratory analysis changing the following parameters:

3.49 Applying the simple discount agreement price for idelalisib, treatment with idelalisib plus rituximab was associated with higher costs and greater QALY gains compared with:

3.50 The ERG did an additional analysis exploring the frequency of intravenous immunoglobulin therapy for patients whose disease responds to treatment compared with patients whose disease does not respond. In the company's model, 45% of patients whose disease does not respond had 1.24 cycles of intravenous immunoglobulin therapy. Patients whose disease does respond had none at all. The ERG explored increasing the number of cycles to responders and decreasing the number of cycles for non‑responders. The results showed that the ICER was sensitive to the changes; when responders had 0.5 cycles of intravenous immunoglobulin and non‑responders had 1.0 cycle, the ICER for idelalisib with rituximab compared with rituximab alone increased from £16,947 per QALY gained to £52,369 per QALY gained.

ERG comments

3.57 The ERG reviewed the new evidence submitted by the company. It validated the company's revised base‑case analysis (see 3 .51) against the results it produced (sections 3.48–3.49). Although the results were not identical (£16,977 per QALY gained compared with £16,947 per QALY gained for idelalisib plus rituximab compared with rituximab alone), they were similar and the ERG had no concerns about the validity of the new results submitted.

3.58 The ERG reviewed the results of the proposed base case. It noted the changes to both the assumptions for intravenous immunoglobulin dosing (see section 3.52) and the changes to the utility values (see section 3.53), and considered them appropriate. However, the ERG highlighted that apart from the sensitivity analysis exploring the proportion of patients having intravenous immunoglobulin therapy, the company had continued to assume that 45% of non‑responders and 0% of responders have the therapy in all the analyses.

3.59 The ERG reviewed the results of the probabilistic sensitivity analysis for the comparisons in the proposed base‑case analysis. The results indicated that the proposed base case was highly sensitive to the cost at which treatments were considered to be cost effective. Although idelalisib plus rituximab was cost effective compared with rituximab alone in 97% of simulations at £50,000 per QALY, it was only cost effective in 62% and 12% of simulations at £30,000 and £20,000 per QALY respectively. The equivalent results for idelalisib plus rituximab compared with best supportive care were 88% at £50,000 per QALY, 22% at £30,000 per QALY and 0% at £20,000 per QALY.

3.60 The ERG conducted further analyses on the results of the company's proposed base case to generate ICERs for idelalisib plus rituximab compared with the other comparators listed in the NICE scope. The results of these analyses showed that the ICERs for idelalisib plus rituximab (using the simple discount agreement price for idelalisib) were:

3.61 The ERG did an exploratory analysis to examine the effect on the ICERs (for the comparisons outside of Study 116) when changing the proportions of patients who had intravenous immunoglobulin therapy. The results of these analyses showed that the ICERs for idelalisib plus rituximab (using the simple discount agreement price for idelalisib) when 20% of non‑responders and 10% of responders had intravenous immunoglobulin therapy were:

3.62 Full details of all the evidence are available.