Rituximab for the first-line treatment of chronic lymphocytic leukaemia

The manufacturer's submission compared rituximab in combination with fludarabine and cyclophosphamide with fludarabine and cyclophosphamide combination therapy. This comparison was based on the CLL-8 trial, a phase 3 randomised controlled trial. The CLL-8 trial was a multicentre, open-label, parallel-group study in people with previously untreated chronic lymphocytic leukaemia. In the CLL-8 trial a total of 817 people were randomised to receive either fludarabine and cyclophosphamide or rituximab in combination with fludarabine and cyclophosphamide; data were reported on 810 people. The median age of trial participants was 61 years and 74% of participants were men. Participants had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1, and 95% of participants had Binet stage B or C disease. People with Binet stage A disease (n=40) were enrolled into the trial until a protocol amendment stopped further enrolment.

Trial participants were randomised to 6 cycles of treatment, with an interim staging after 3 cycles. People with progressive or stable disease at interim staging were offered alternative treatments by their clinicians outside the trial. People in the control group whose disease did not respond to treatment did not cross over to the treatment group, but could be offered rituximab-containing regimens. People whose disease showed a partial or complete response at the interim staging received all 6 cycles of treatment. Each cycle of 28 days consisted of fludarabine and cyclophosphamide chemotherapy (fludarabine [25 mg/m^²] and cyclophosphamide [250 mg/m^²] on days 1, 2 and 3) with or without rituximab (375 mg/m^² on day 0 of cycle 1, 500 mg/m^² on day 1 of cycles 2 to 6). All trial treatments were administered intravenously.

The primary outcome of the trial was progression-free survival, defined as the time between randomisation and the date of the first documented disease progression, relapse or death by any cause. Secondary outcomes were event-free survival, overall survival, disease-free survival, duration of response, time to new chronic lymphocytic leukaemia treatment and response rates. Quality-of-life data were collected in the trial using the Spitzer Quality of Life Index and the European Organisation for Research and Treatment of Cancer-Quality of Life Questionnaire Core 30 (EORTC-QLQC30).

Demographic characteristics and disease characteristics, including Binet stage B symptoms and prognostic markers such as cytogenetic abnormalities, were well balanced between the trial groups. Of all trial participants, 64% had Binet stage B disease, 31% had Binet stage C disease, and 5% had Binet stage A disease. The trial also enrolled 46 people (8%) with p53 deletion, a chromosome abnormality associated with a poorer prognosis.

A pre-planned interim analysis of the trial data after a median follow-up of 20.7 months showed a statistically significant difference in progression-free survival between the treatment groups. At this point the trial was halted and the interim analysis became the main analysis. The reported median progression-free survival was 39.8 months in the rituximab in combination with fludarabine and cyclophosphamide group and 32.2 months in the fludarabine and cyclophosphamide group, with a hazard ratio of 0.56 (95% confidence interval [CI] 0.43 to 0.72, p<0.0001). The trial also reported an overall response rate of 86.1% in the rituximab in combination with fludarabine and cyclophosphamide group and 72.7% in the fludarabine and cyclophosphamide group. At this point median overall survival had not been reached and the trial reported a hazard ratio of 0.64 (95% CI 0.41 to 1.00, p=0.05).

The manufacturer also submitted analyses of data from the CLL-8 trial collected from 3 follow-up points of a longer duration. After a median duration of follow-up of 25.4 months, the reported median progression-free survival was 42.8 months in the rituximab in combination with fludarabine and cyclophosphamide group and 32.5 months in the fludarabine and cyclophosphamide group, with a hazard ratio of 0.60 (95% CI 0.48 to 0.76, p<0.001). At the end of this follow-up period the statistically significant difference in overall survival was not maintained (hazard ratio 0.72, 95% CI 0.48 to 1.09, p=0.13). However, the data remained highly censored, because the majority of people were still alive. A further analysis after a median follow-up of 25.5 months reported a rate of progression-free survival of 76.6% in the rituximab in combination with fludarabine and cyclophosphamide group and 62.3% in the fludarabine and cyclophosphamide group (p<0.001). At this follow-up point, the CLL-8 trial reported an overall response rate of 95% in the rituximab in combination with fludarabine and cyclophosphamide group and 88% in the fludarabine and cyclophosphamide group (p=0.001). The overall survival rate was 91% in the rituximab in combination with fludarabine and cyclophosphamide group and 88% in the fludarabine and cyclophosphamide group (p=0.18). In a further analysis after a median follow-up of 26.4 months the reported mean progression-free survival was 37.1 months in the rituximab in combination with fludarabine and cyclophosphamide group and 30.8 months in the fludarabine and cyclophosphamide group (p<0.001). The hazard ratio for progression-free survival was 0.6 (95% CI 0.47 to 0.75, p<0.0001). Mean overall survival was 47.7 months in the rituximab in combination with fludarabine and cyclophosphamide group and 48.2 months in the fludarabine and cyclophosphamide group (p=0.18).

The manufacturer presented a number of subgroup analyses. For the people with the p53 mutation the hazard ratio for progression-free survival was 0.6 (95% CI 0.31 to 1.19). The hazard ratio for progression-free survival for people with Binet stage A disease was 0.13 (95% CI 0.03 to 0.61, p=0.01), Binet stage B disease was 0.46 (95% CI 0.32 to 0.63, p<0.0001) and Binet stage C disease was 0.88 (95% CI 0.58 to 1.33, p=0.54). The CLL-8 trial was not powered to detect differences in treatment effect for any of these subgroups.

In the CLL-8 trial, 77% of people in the rituximab in combination with fludarabine and cyclophosphamide group experienced a grade 3 or 4 adverse event compared with 62% in the fludarabine and cyclophosphamide group. In the rituximab in combination with fludarabine and cyclophosphamide group 46% of people experienced a serious adverse event; this figure was 41% of people in the fludarabine and cyclophosphamide group. The main adverse events were haematological toxicities, with neutropenia, leucopenia, febrile neutropenia and pancytopenia having a higher incidence (at least 2% difference) in the rituximab in combination with fludarabine and cyclophosphamide group, and thrombocytopenia, anaemia and pyrexia having a higher incidence (at least 2% difference) in the fludarabine and cyclophosphamide group. There were no differences in the rate of other adverse events between the trial groups.

The manufacturer provided data from 4 uncontrolled phase 2 trials on the efficacy and tolerability of combining rituximab with different chemotherapy regimens. The combination chemotherapies included fludarabine, pentostatin, cyclophosphamide and mitoxantrone for the first-line treatment of chronic lymphocytic leukaemia. One of the studies (n=300) compared a group of people treated with rituximab in combination with fludarabine and cyclophosphamide with a group of people who had been treated with fludarabine-based regimens in the past and provided data with a median follow-up of 6 years. For the group receiving rituximab the rate of overall survival after 6 years was 77% with a 95% overall response rate. Median time to progression was 80 months. In comparison with the historical control group, rituximab in combination with fludarabine and cyclophosphamide was associated with statistically significant overall survival and was the strongest independent predictor of survival (hazard ratio 0.48, p<0.001).

In their submission, the manufacturer also compared rituximab in combination with fludarabine and cyclophosphamide with chlorambucil using a mixed treatment comparison. A mixed treatment comparison was conducted because there were no head-to-head studies comparing rituximab with comparators other than fludarabine and cyclophosphamide. As well as chlorambucil, this analysis also included alemtuzumab, fludarabine alone and bendamustine. In addition to CLL-8, a further 7 trials were identified and used to create a network of evidence to make indirect comparisons of rituximab in combination with fludarabine and cyclophosphamide with the other comparators. The studies were combined using a fixed effect model because there was no apparent gain in goodness of fit when a random effects model was used. The mixed treatment comparison showed that chlorambucil had the shortest progression-free survival and therefore this was used as the reference treatment. The mean hazard ratios for other treatments compared with chlorambucil were 0.24 for rituximab in combination with fludarabine and cyclophosphamide, 0.43 for fludarabine and cyclophosphamide, 0.59 for alemtuzumab and 0.86 for fludarabine alone. The mean hazard ratio for progression-free survival was 0.56 for rituximab in combination with fludarabine and cyclophosphamide compared with fludarabine and cyclophosphamide, 0.24 for rituximab in combination with fludarabine and cyclophosphamide compared with chlorambucil, 0.42 for rituximab in combination with fludarabine and cyclophosphamide compared with alemtuzumab and 0.28 for rituximab in combination with fludarabine and cyclophosphamide compared with fludarabine alone.

The manufacturer's submission presented an economic analysis comparing rituximab in combination with fludarabine and cyclophosphamide with fludarabine and cyclophosphamide, and rituximab in combination with fludarabine and cyclophosphamide with chlorambucil. The manufacturer developed a 3-state Markov model with a cycle length of 1 month and a 15-year time horizon (to represent a lifetime horizon). The health states in the model were 'progression-free survival', 'progressed', or 'death'. People entered the model in the progression-free survival health state. The probability of transition from the progression-free survival to the progressed health state was taken from the groups in CLL-8. For the transition from the progression-free survival to the death health state, trial data were used and supplemented with Office of National Statistics data to inform the background mortality rate. Transition from the progressed to the progression-free survival health state was not possible. For the transition from the progressed to the death health state, data for people from both groups of the trial were aggregated and a single probability from the trial applied as there was a non-significant difference in survival following progression between the groups in the trial.

In the model, the drug costs were calculated assuming a body surface area of 1.93 m^², which reflects the average body surface area of the people in the CLL-8 trial. The CLL-8 trial used fludarabine and cyclophosphamide administered intravenously, but it is more common to use oral chemotherapy in the UK. In the model it was assumed that the efficacy of fludarabine and cyclophosphamide is the same regardless of the route of administration if the dosage is adjusted to ensure equivalent bioavailability. The costs of fludarabine and cyclophosphamide treatment in the model were adjusted to allow for the difference in the route of administration. The drug costs for rituximab were £1,397 for the first cycle of treatment and £1,746 for subsequent cycles. For 6 cycles of treatment the total drug cost of rituximab was £10,128. The total drug costs of fludarabine, cyclophosphamide and chlorambucil were calculated as £2,790, £22 and £286, respectively. In the base case, all people received 6 cycles of therapy unless disease progression occurred before the end of the 6 cycles.

The model included costs for supportive care that varied between the health states. This included costs for blood transfusions and bone marrow transplant in the progression-free survival health state taken from the CLL-8 trial and costs for second-line therapies for the progressed health state. In the model rituximab had a cost for intravenous administration of £430 per cycle of treatment and the cost for an appointment to prescribe oral fludarabine and cyclophosphamide chemotherapy was £280. It was assumed that oral chemotherapy could be prescribed in the same appointment as rituximab so no additional cost of prescribing oral chemotherapy was included for the rituximab treatment group. Costs were also added for the pharmacist's time to prepare the infusion.

The utility values used in the manufacturer's submission were taken from a health technology assessment report (Hancock et al. 2002) that assessed the cost effectiveness of fludarabine as a first-line treatment for chronic lymphocytic leukaemia. A utility of 0.8 was attached to the progression-free survival health state and 0.6 to the progressed health state. The estimates of utility were not preference based, and were estimated by the authors of the health technology assessment report from condition-specific health-related quality-of-life data. No disutility for adverse events was included in the model. The manufacturer provided an interim analysis of 11 people from an observational study of utility in people with chronic lymphocytic leukaemia. The value for progression-free survival was consistent with that used in the manufacturer's submission. No conclusions could be drawn about the utility value appropriate for the progressed health state, as data for only 2 people were available.

The manufacturer provided a base-case estimate of incremental cost effectiveness of rituximab in combination with fludarabine and cyclophosphamide in comparison with fludarabine and cyclophosphamide. The incremental quality-adjusted life year (QALY) gain was 0.88 at an incremental cost of £11,617, giving an incremental cost-effectiveness ratio (ICER) of £13,189 per QALY gained. The probabilistic sensitivity analysis presented suggested that rituximab in combination with fludarabine and cyclophosphamide had a 91.9% probability of being cost effective at £20,000 and 98.6% probability of being cost effective at £30,000 when compared with fludarabine and cyclophosphamide. The manufacturer also provided an estimate of the incremental cost effectiveness of rituximab in combination with fludarabine and cyclophosphamide in comparison with chlorambucil. The incremental QALY gain was 1.91 at an incremental cost of £12,250, giving an ICER of £6,422 per QALY gained. Probabilistic sensitivity analysis suggested that the probability of rituximab in combination with fludarabine and cyclophosphamide being cost effective in comparison with chlorambucil was 100% at both £20,000 and £30,000 per QALY gained.

A sensitivity analysis was presented in the manufacturer's submission using different parametric models for the progression-free survival extrapolation. Additional sensitivity analyses were completed as follows:

• including costs for adverse events

• including costs for febrile neutropenia episodes (as in CLL-8)

• increasing and decreasing supportive care costs for the health states by 50%

• assuming utility values for the health states such that the difference in the values between the health states was 0.4 and 0.1.
  
  The assumption of a similar rate of adverse events for chlorambucil and fludarabine and cyclophosphamide was tested by assuming no bone-marrow transplants, fewer transfusions and less febrile neutropenia for the chlorambucil arm. One-way sensitivity analyses suggested that the results were not sensitive to a variety of parameter assumptions including utility values, monthly supportive care costs and drug administration costs. The results were sensitive to the function used to extrapolate progression-free survival (exponential, Gompertz), and the highest ICER reported (using a Gompertz function) was £22,661 per QALY gained.

The manufacturer's submission also included a scenario analysis to explore the impact on the ICER of using intravenous administration of fludarabine and cyclophosphamide chemotherapy instead of oral administration. This analysis demonstrated that the ICER was not sensitive to assumptions about the mechanism of administration. A further scenario analysis modelled the cost effectiveness of rituximab in combination with chemotherapy agents other than fludarabine and cyclophosphamide. The results of this analysis suggested that the QALY gain from combining rituximab with chemotherapy would need to decrease to about 40% of that in the base case, all else remaining the same, for the ICER for rituximab to increase to over £30,000 per QALY gained.

At the request of the ERG, the manufacturer performed a further 1-way sensitivity analysis to explore the impact on the ICER of using alternative assumptions about different mortality rates between the progression-free survival and progressed health states. The manufacturer increased the mortality rate in the rituximab in combination with fludarabine and cyclophosphamide group by 315% so that the life years gained in both treatment groups were the same (0.24 QALYs). In this scenario, the incremental QALY gain was 0.24 at an incremental cost of £7,226, giving an ICER of £30,336 per QALY gained.

The ERG considered that all the relevant studies had been identified. The ERG noted that the manufacturer's submission was based on only 1 completed clinical trial, and that this was unpublished. However, it considered this study to be of good quality. In addition, it noted that this study used intravenous administration of fludarabine and cyclophosphamide rather than oral administration which is normally used in UK clinical practice. The ERG considered the study population was appropriate. It noted that the subgroup of people with chronic lymphocytic leukaemia and the p53 deletion was only considered in relation to progression-free survival and not assessed in the cost-utility model. The ERG considered that the main comparators used in the cost-effectiveness analysis (fludarabine and cyclophosphamide, and chlorambucil) were appropriate. It noted that the mixed treatment comparison provided estimates of clinical effectiveness comparing rituximab in combination with fludarabine and cyclophosphamide with additional comparators, including alemtuzumab, fludarabine monotherapy and bendamustine. The ERG considered that the mixed treatment comparison completed by the manufacturer was appropriate.

The ERG considered the fact that in the manufacturer's economic model people in the progressed health state could not move back into the progression-free survival health state was unrealistic due to the natural history of chronic lymphocytic leukaemia. People with chronic lymphocytic leukaemia may receive further treatment at progression, which may then result in further periods of progression-free survival. The relapsing nature of chronic lymphocytic leukaemia means that subsequent relapses are less likely to respond to further treatment. This implies that subsequent relapses are likely to be associated with higher disease-related mortality. Therefore, the ERG considered that the manufacturer's assumption of a constant hazard of death after progression may not be appropriate. The ERG highlighted that the overall effect of the aggregated progressed health state and constant hazard of death from this health state was to imply a correlation between progression-free survival and overall survival which it did not consider had been empirically demonstrated in the manufacturer's submission. The ERG further considered that the sensitivity analyses presented by the manufacturer did not fully investigate the uncertainty associated with the extent to which gains in progression-free survival led to gains in overall survival.

The ERG performed an exploratory analysis of the comparison of rituximab in combination with fludarabine and cyclophosphamide with fludarabine and cyclophosphamide. First, it conducted a component analysis to examine the relative contributions to utility gain from the gain in progression-free survival and the gain in overall survival. This analysis showed that progression-free survival contributed to 0.24 QALYs and overall survival to 0.64 QALYs (of a total gain of 0.88 QALYs). The ERG concluded that this demonstrated that in the model the majority of the benefit is derived from overall survival, making it sensitive to changes in assumptions about overall survival benefits from rituximab. The ERG noted that within the model, because a single transition probability is attached to all people in the progressed health state, the benefit in overall survival is derived almost entirely from the different rate of transfer from the progression-free health state to the progressed health state.

The ERG repeated the analysis completed by the manufacturer (see section 3.18) that removed the differences in overall survival between the 2 groups in the model. This was done by decreasing the probability of death in the progressed health state for the fludarabine and cyclophosphamide group. A decrease in the probability of death in the fludarabine and cyclophosphamide group to 57% of the base-case level removed the difference in overall survival between the groups and resulted in a QALY gain of 0.24 at an incremental cost of £7,228 and an ICER of £30,304 per QALY gained. When assuming no difference in overall survival between the 2 treatment groups in the analysis, the results suggested that the probability of rituximab in combination with fludarabine and cyclophosphamide being cost effective compared with fludarabine and cyclophosphamide at £20,000 per QALY gained was 29% and at £30,000 per QALY gained was 49%. The ERG identified that if it is assumed that there is no difference in overall survival between the rituximab in combination with fludarabine and cyclophosphamide and fludarabine and cyclophosphamide groups, the model outputs become sensitive to the assumed utility differences between the progression-free and the progressed health states. If the difference in utility between the health states is reduced by 0.1 (that is from 0.2 to 0.1), the ICER increases to £60,302 per QALY gained.

The ERG completed another exploratory analysis that assumed that the actual overall survival benefit from treatment with rituximab was somewhere between the manufacturer's base case and the assumption of no overall survival benefit. The ERG incorporated this assumption into the probabilistic sensitivity analysis by adding an additional variable, in which the decrease in probability of death in the fludarabine and cyclophosphamide group was sampled as a uniform distribution between 1 and 0.574. The results suggested that rituximab in combination with fludarabine and cyclophosphamide had a 72% probability of being cost effective compared with fludarabine and cyclophosphamide at £20,000 per QALY gained and 88% probability of being cost effective at £30,000 per QALY gained.

In response to consultation on the appraisal consultation document, the manufacturer submitted additional evidence in support of the combination of rituximab with chlorambucil compared with chlorambucil alone. The manufacturer presented data from 4 randomised controlled trials in follicular lymphoma, another low-grade B-cell cancer with a relapsing and remitting course. In these studies the addition of rituximab to a range of chemotherapy regimens showed a benefit in progression-free survival and response rates. A further phase 2 trial of rituximab in combination with chlorambucil in 29 people with a range of low-grade lymphoproliferative disorders showed an overall response rate of 89%, with a complete response rate of 63%.

The manufacturer also provided additional economic analysis comparing rituximab in combination with chlorambucil with chlorambucil alone. This used the same model as the original submission, with a number of amendments. The age of the cohort was increased to 70 years to reflect the fact that people treated with chlorambucil in routine clinical practice are generally older than those treated with fludarabine and cyclophosphamide. The baseline risk for the chlorambucil group was taken from the mixed treatment comparison included in the original submission. It was assumed that the hazard ratio for the addition of rituximab to fludarabine and cyclophosphamide observed in the CLL-8 trial (0.595) could be applied to this baseline risk to estimate the effect of adding rituximab to chlorambucil. Drug and administrative costs for rituximab and chlorambucil were included and all other model inputs and assumptions remained the same. Assuming the same hazard ratio meant the relative effect of adding rituximab to chlorambucil was the same as adding rituximab to fludarabine and cyclophosphamide. However, the absolute treatment effect would be smaller because single agent chlorambucil is less effective than fludarabine and cyclophosphamide. The analysis suggested the incremental QALY gain was 0.51 at an incremental cost of £11,570, giving an ICER of £22,490 per QALY gained. A probabilistic sensitivity analysis using the 95% CIs around the assumed hazard ratios showed an 18.6% probability of the ICER being below £20,000 per QALY gained and a 99.7% probability of the ICER being below £30,000 per QALY gained.

The manufacturer also noted that a single arm phase 2 trial is underway of rituximab in combination with chlorambucil in people with untreated chronic lymphocytic leukaemia who are not fit for fludarabine-based treatment. The primary objective of the study is a safety analysis of the combination of rituximab and chlorambucil. Secondary objectives include: response rate, progression-free survival, overall survival, disease-free survival and duration of response. An interim analysis of the data is to be presented in December 2009 and a full analysis is expected towards the end of 2010.

Full details of all the evidence are in the manufacturer's submission and the ERG report.