The ERG report comprised a critical review of the clinical and cost-effectiveness evidence presented in the manufacturer’s submission, assessing the appropriateness of the manufacturer’s interpretation and analysis of the evidence.
3.1 Clinical Evidence Provided By the Manufacturer
The clinical effectiveness evidence in the submission was based predominantly on data from two randomized, controlled trials. RAVE was a multicentre, randomized, placebo-controlled, double-blind, double-dummy, noninferiority trial, which compared RTX (375 mg/m
2 per week for 4 weeks) and CYC (2 mg/kg per day) for induction of remission in 197 patients with new or relapsed severe MPA or GPA [
35]. The Randomized Trial of Rituximab Versus Cyclophosphamide in ANCA-Associated Vasculitis (RITUXVAS) was a phase II, open-label, randomized, controlled, prospective study, which evaluated RTX (375 mg/m
2 per week for 4 weeks) combined with two intravenous doses of CYC (15 mg/kg given with the first and third RTX doses), compared with the CYCLOPS intravenous CYC regimen (15 mg/kg for 3–6 months, 6–10 doses in total) in 44 patients with newly diagnosed severe AAV [
45]. In both studies, patients in the CYC control group received azathioprine as remission maintenance therapy, whereas patients in the RTX group received no remission maintenance therapy. The manufacturer gave precedence to data from the RAVE trial because this reflected the regimen sanctioned in the market authorization, presenting data from RITUXVAS as supporting evidence.
In RAVE, the primary endpoint was remission of disease without use of prednisone at 6 months; glucocorticoids were tapered off, such that all patients who had remission without disease flares had discontinued glucocorticoids by 5 months. Remission was signified by a Birmingham Vasculitis Activity Score for WG (BVAS/WG) [
46] of 0. Secondary endpoints included the rates of disease flares, a BVAS/WG of 0 with less than 10 mg per day of prednisone use, the cumulative glucocorticoid dose, rates of AEs, and HRQoL measured using the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) [
47]. A disease flare was defined as an increase in the BVAS/WG of ≥1 point.
The manufacturer reported the results of the RAVE trial, which are published elsewhere [
35]. The trial demonstrated noninferiority of RTX compared with CYC because the lower bound of the 95 % confidence interval for the primary endpoint was higher than the pre-determined noninferiority margin of −20% (
p < 0.001). Sixty-three patients (63.6 %) in the RTX group achieved complete remission, compared with 52 patients (53.1 %) in the CYC group, an absolute difference of 10.6 % (95 % confidence interval [CI] −3.2 to 24.3 %,
p = 0.132), hence a statistically significant advantage associated with RTX was not observed. Statistically significant changes in the secondary efficacy endpoints measured in RAVE were not observed, other than in exploratory subgroup analyses [
35]. Exploratory subgroup analysis of complete remission in newly diagnosed patients gave an absolute difference of −4.2 % (95 % CI −23.6 to 15.3 %,
p = 0.673) in favour of CYC (60.4 % in the RTX group compared with 64.6% in the CYC group). A similar analysis in patients with recurrent disease gave a statistically significant absolute difference of 24.7 % (95 % CI 5.8 to 43.6 %,
p = 0.013) in favour of RTX (66.7 % in the RTX group compared with 42.0 % in the CYC group).
In RITUXVAS, the primary endpoints were sustained remission (defined as an absence of disease activity [BVAS = 0] for at least 6 months) and the rates of severe AEs at 12 months. Sustained remission occurred in 25 of 33 patients in the RTX group (76 %) and in 9 of 11 patients in the control group (82 %). The absolute difference in sustained remission with RTX as compared with CYC was −6 percentage points (95 % CI −33 to 21, p = 0.68). Among the patients who remained alive at 12 months, 93 % of the patients in the RTX group and 90 % of those in the control group had sustained remission (p = 0.80). There were no significant differences between the treatment arms for any of the secondary efficacy outcomes, such as the median time to remission, the median BVAS and the prednisone doses.
The manufacturer submitted evidence on the safety of RTX, focussed on 18 months of data collected in RAVE. There were no reported significant differences between the treatment groups in almost all AE outcomes, but there were some notable disparities regarding leucopenia and malignancies. More patients in the CYC group than in the RTX group (32 [33 %] versus 22 [22 %], p = 0.01) had one or more of the predefined selected AEs (including death, malignant conditions, grade 2 or higher leucopenia or thrombocytopenia, grade 3 or higher infections, drug-induced cystitis, venous thromboembolic events, stroke, hospitalizations and infusion reactions that contraindicated further infusions), but more episodes of grade 2 or higher leucopenia in the control group (10 versus 3) accounted for most of this difference. Malignant conditions developed in 7 patients after 6 months: 6 of 124 patients (5 %) who were exposed to RTX at any point during the trial (including after 6 months), as compared with 1 of 73 patients (1 %) without exposure to RTX (p = 0.26).
The manufacturer presented limited safety results from the RITUXVAS trial, but the ERG noted that a total of 31 severe AEs had occurred in 14 of the 33 patients in the RTX group (42 %), and 12 severe AEs had occurred in 4 of the 11 patients in the control group (36 %). The incidence rates of severe AEs did not differ significantly between the groups (p = 0.77). Six of the 33 patients in the RTX group (18 %) and 2 of the 11 patients in the control group (18 %) died (p = 1.00). The causes of death were infections (in 3 patients in the RTX group and in 1 patient in the control group), cardiovascular disease (in 1 patient in the RTX group and in 1 patient in the control group), and complications of end-stage renal failure (in 2 patients in the RTX group).
3.1.1 Critique of Clinical Evidence and Interpretation
The ERG noted that the manufacturer identified the two randomized, controlled trials comparing RTX with CYC as induction therapy for adults with what can be described as generalized, severe AAV. The manufacturer did not conduct a meta-analysis or synthesis, and thus the results were reported as they were in the relevant published studies. The ERG considered that evidence from RAVE suggests that RTX given at a dose of 4 × 375 mg/m2 is noninferior to oral CYC in terms of induction of remission in adults with AAV and de novo disease, and is superior to oral CYC in terms of remission in adults with generalized, severe AAV who have relapsed one time on CYC, on the basis of the exploratory subgroup analysis. Evidence from RITUXVAS suggests that RTX given at a dose of 4 × 375 mg/m2 plus 2–3 intravenous pulses of CYC is noninferior to intravenous pulse CYC in terms of remission in adults with generalized, severe AAV and de novo disease. The ERG stated that the evidence relates only to induction of remission with these specific regimens in adult populations with generalized, severe AAV, and only evidence on the 4 × 375 mg/m2 RTX dose was submitted by the manufacturer, rather than the 2 × 1 mg dose, which the manufacturer recognized as the most widely used off-label dose in the UK for AAV, and which represents a smaller overall dose. No evidence was presented on the efficacy or safety of RTX in adults intolerant of CYC, with contraindications against CYC or with mild AAV; in children; or for use of this regimen as maintenance therapy or for relapse after RTX.
The ERG applied the Cochrane risk-of-bias tool to appraise the RAVE and RITUXVAS trials [
48], as well as the noninferiority trial extension of the CONSORT statement for the RAVE trial [
49]. In general, both trials were considered to be at low risk of bias, although there was a high risk of performance and detection bias in the RITUXVAS trial because of its open-label nature. In addition, the ERG had concerns over the populations, interventions, comparators and outcomes included in the studies, with regard to their usefulness for informing UK recommendations. The RAVE trial considered relatively young adults only (the mean ages were 54 and 51 years in the RTX and CYC treatment arms, respectively) with moderately severe AAV and either de novo disease or following relapse after CYC. The trial did not include adults with severe renal impairment or life-threatening pulmonary haemorrhage, those who had contraindications against CYC or those who were CYC refractory. It is uncertain if RTX alone will demonstrate equal efficacy and safety in other adult populations or children. The RITUXVAS trial considered a much older population (with median ages of 68 and 67 years in the RTX and control treatment arms, respectively) with severe renal impairment, but both arms of the trial included CYC treatment. In the absence of a head-to-head trial, it is uncertain whether RTX combined with CYC is inferior, equivalent or superior to RTX alone.
The ERG noted that the manufacturer’s submission focused upon oral CYC as a comparator. It is uncertain whether RTX without CYC would demonstrate equal efficacy and safety if it were compared with intravenous CYC. The ERG received clinical expert advice that intravenous CYC is used more often in expert clinical practice, and this might have a better safety profile [
50,
51]. In addition, the ERG noted that no evidence was submitted on the comparability of RTX with other potentially relevant comparators specified in the NICE scope, such as mycophenolate mofetil and methotrexate. The ERG noted that some evidence on these interventions—such as the NORAM trial [
36,
52] and the MYCYC trial [
37]—is available, which could have informed an indirect comparison; however, the manufacturer’s submission did not contain such an analysis.
The ERG considered the lack of longer-term data on efficacy and safety to be a significant concern. RAVE and RITUXVAS did not monitor outcomes beyond 18 months, and primary endpoints were measured at 6 and 12 months, respectively. The duration of remission was not reported as an outcome in either trial, yet it has important implications from both a clinical perspective and an economic perspective. Regarding safety, RAVE provided evidence that RTX with concurrent glucocorticoid therapy has a safety profile similar to that of oral CYC with concurrent glucocorticoid therapy, and the RITUXVAS trial suggested that RTX plus intravenous pulse CYC with concurrent glucocorticoid therapy has a safety profile similar to that of intravenous pulse CYC with concurrent glucocorticoid therapy. However, the safety data were collected over relatively short time periods (maximum 18 months), and much of the observed toxicity was related to steroids rather than to RTX or CYC. Thus it is not possible to draw conclusions specifically comparing the safety of RTX and CYC. In addition, part of the hypothesis for the RITUXVAS trial was the potential for improved safety through treatment with RTX rather than CYC. This was not demonstrated in RAVE or RITUXVAS. One reason suggested for the comparable rather than superior safety profile of RTX in the RITUXVAS and RAVE trials relates to the short durations of these trials [
35]. A further reason might be that the high cumulative dose of RTX in both trials [
53‐
55] caused AEs. The ERG concluded that longer-duration trials with more comparable groups are needed—particularly to assess malignancies and fertility outcomes—and patients who have previously received CYC need to be included in such trials.
3.2 Cost-Effectiveness Evidence
The manufacturer conducted a systematic review on the cost effectiveness of one or more interventions for patients with AAV. No suitable studies were found, and the manufacturer therefore submitted a de novo model-based economic evaluation, which assessed the cost effectiveness of different treatment strategies for patients with AAV. Separate analyses were undertaken for patients with newly diagnosed AAV, people with recurrent AAV, all patients (including both newly diagnosed and relapsed patients) and patients intolerant of CYC. The model was constructed using a cohort Markov approach with a 6-month cycle length. Patients entered the model at an age of 52.8 years, based upon the mean age at baseline in the RAVE trial. A lifetime time horizon was modelled. Health states were included for remission, non-remission, uncontrolled disease and death (Fig.
1). The structure of the model is outlined in the ERG report [
2].
Model parameter values for response rates were primarily taken from the RAVE trial. However, the model incorporated various treatment sequences, including two courses of CYC and RTX for selected patients, and for some sequences, data were not available from RAVE, and so assumptions were made. For each subgroup that was modelled, the relevant response rate was taken from RAVE where possible—RAVE provided data on response rates for CYC and RTX in the treatment-naïve subgroup, the recurrent-disease subgroup and all patients. For the recurrent-disease subgroup, no data on remission rates associated with a second course of CYC could be derived from the RAVE trial, and no data were available on the remission rates associated with a second course of RTX. The manufacturer assumed reduced response rates for these parameters.
The relapse rates used in the manufacturer’s model were derived from the RAVE trial. Exponential models were fitted to flare data from patients who had experienced remission at 6 months, in order to estimate the time-to-event for relapse. The same relapse rate was applied after different courses of treatment within each subgroup analysis, but different relapse rates were used in each subgroup.
Age- and sex-specific mortality risks were based upon UK life tables [
56], adjusted according to the age and gender distributions in the RAVE trial and a published standardized mortality ratio (SMR) comparing a general population with an AAV population [
57]. The AAV SMR was applied directly to the non-remission health state, and the SMR was arbitrarily altered by ±10 % to reflect mortality in the remission and uncontrolled-disease states.
Utility scores were based upon UK evidence for the general population [
58], adjusted according to the RAVE trial demographics and weighted for remission (weight = 0.98) and non-remission (weight = 0.88) health states, according to SF-36 data (subsequently transformed into EQ-5D utility scores) collected from the RAVE trial. An additional utility decrement was applied to patients in the uncontrolled-disease health state (weight = 0.79), assuming that the difference between the non-remission and uncontrolled-disease health states is the same as the difference between the remission and non-remission health states. Further decrements were made according to the probability of experiencing AEs in each health state. The AE rates were based upon data from RAVE and included anaemia, leucopenia, deep vein thrombosis, dyspnoea, diarrhoea and pneumonia. The costs of managing these AEs were based upon relevant costs taken from the NHS Trusts reference costs schedules for 2009–2010 [
59].
Drug acquisition costs were taken from British National Formulary No. 64 [
60] and were based upon licensed doses, apart from glucocorticoid costs, which were based upon the average dose used in the RAVE trial. Administration costs for RTX and intravenous CYC were assumed to be equal and were based upon a previous economic evaluation of infliximab [
61]. Monitoring costs were included for oral CYC and for patients receiving azathioprine maintenance therapy. Regular outpatient appointments with relevant consultant specialists were also assumed, the frequency of which differed across health states—notably, once patients were in the uncontrolled-disease health state, it was assumed that they attended one specialist palliative care outpatient appointment every 1.5 weeks for all remaining years of life. The costs of these appointments were valued using NHS reference costs.
The results from the manufacturer’s base-case analysis are shown in Table
1. These results represent those provided by the manufacturer after it responded to clarifications requested by the ERG, as some amendments were made to the economic model. They indicate that the RTX treatment strategy results in an incremental cost-effectiveness ratio (ICER) of £8,544 per QALY gained for all patients, but in the treatment-naïve and recurrent-disease subgroups, the ICERs were £55,175 per QALY gained and £43,003 per QALY gained, respectively. In the CYC-intolerant subgroup analysis, RTX dominated the comparator, which was deemed to represent best supportive care. The manufacturer undertook probabilistic sensitivity analysis but did not report the probabilistic ICERs. For the all-patients analysis, the manufacturer reported that the probabilities of the RTX treatment sequence being cost effective, compared with the CYC treatment sequence, were 61.7 % and 64.6 % for cost-effectiveness thresholds of £20,000 and £30,000 per QALY gained, respectively. The manufacturer presented several deterministic sensitivity analyses and found that the results of the all-patients analysis were relatively insensitive to variations in the tested model input parameters, with the exception of (i) the CYC remission rate; (ii) the uncontrolled-disease utility; (iii) the frequency of consultant visits in each health state; and (iv) the reference cost applied to the consultant appointments in the uncontrolled-disease health state. In addition, several structural sensitivity analyses were presented, testing different numbers of CYC and RTX courses for different patient groups. As would be expected, the results were highly sensitive to these structural assumptions.
Table 1
Headline cost-effectiveness results presented by the manufacturer
All patients
|
CYC | 8.03 | £95,819 | – | – | – |
RTX | 8.19 | £97,210 | 0.1628 | £1,391 | £8,543.69 |
Treatment-naïve subgroup
|
CYC | 8.45 | £81,327 | – | – | – |
RTX | 8.53 | £86,021 | 0.0851 | £4,694 | £55,174.92 |
Recurrent-disease subgroup
|
CYC | 7.89 | £100,699 | – | – | – |
RTX | 7.98 | £104,550 | 0.0896 | £3,851 | £43,003.05 |
CYC–intolerant subgroup
|
BSC | 7.49 | £102,721 | – | – | – |
RTX | 8.02 | £97,836 | 0.5386 | −£4,885 | RTX dominates |
3.2.1 Critique of the Cost-Effectiveness Evidence and Interpretation
The ERG had serious concerns with the manufacturer’s economic model. These included technical errors, such as minor mistakes in the estimation of mortality rates, some inaccuracies in the estimation of standard errors for relapse rates and cost parameters, and a failure to characterize uncertainty around all uncertain parameters. In addition, some parameter value estimates were inappropriate, and some structural assumptions were implausible [
2]. The ERG took the view that the manufacturer’s economic model could not be considered reliable or robust.
The most important issues highlighted by the ERG surrounded the treatment sequences modelled by the manufacturer, and the definition of the uncontrolled-disease health state. Clinical advice received by the ERG suggested that a lifetime cumulative dose of 20–30 g of CYC should not be exceeded. Typically, this represents 1–2 courses of oral CYC or 2–3 courses of intravenous CYC. Hence, the treatment sequence that was modelled should reflect the treatment history of the patient, and thus it should differ for the different subgroups that were modelled. For treatment-naïve patients, the ERG suggested that two courses of CYC were plausible, whereas for patients with recurrent disease, only one course of CYC was likely to be plausible. In addition, the manufacturer considered treatment sequences only in which RTX was given first-line; the ERG stated that there was no reason to assume that this should be the case—it was relevant to consider its cost effectiveness when it was given both before and after CYC. Finally, the ERG disagreed with the manufacturer’s assumption that patients who did not respond to an initial course of RTX would immediately receive a second course of RTX—the ERG believed it was more appropriate to include only one course of RTX within the modelled treatment sequences.
The ERG’s clinical advisors suggested that the uncontrolled-disease health state modelled by the manufacturer was unrealistic, and that a more common health state is one in which the most effective induction treatments have been used but some other treatment or combination of treatments is utilized in order to afford patients a reasonable level of disease control. This may be described as low-grade ‘grumbling’ disease [
62]. The ERG noted that the level of HRQoL associated with such a health state was unlikely to be substantially worse than that experienced in the non-remission health state. Also, although treatment would be received in this state, the manufacturer’s assumption that specialist palliative care is received at hospital outpatient appointments once every 1.5 weeks for all remaining years of life appeared to be a substantial overestimate. The ERG’s clinical advisors suggested it would be more appropriate to assume that patients in this health state would continue to receive maintenance treatment, and that outpatient appointments would occur each month initially, followed by less frequent visits over time. In the base-case version of the manufacturer’s model, patients in the CYC group spent 70.7 % of their mean life expectancy in the uncontrolled-disease health state, compared with 63.2 % in the RTX group. The cost savings and utility benefits associated with RTX due to this were likely to be substantially overestimated.
3.3 Additional Work Undertaken By the Evidence Review Group
The ERG amended the identified technical errors in the manufacturer’s model, amended several parameter value estimates to better reflect reality (particularly for the uncontrolled-disease health state) and amended the modelled treatment sequences to allow for fully incremental analyses of introducing RTX into the treatment sequence either before or after CYC. The cost-effectiveness results for all patients and for the treatment-naïve, recurrent-disease and CYC-intolerant subgroups are shown in Table
2.
Table 2
Cost-effectiveness results of the Evidence Review Group’s preferred analyses
All patients
|
CYC → CYC → BSC | £18,926.57 | 8.5810 | – | – | – |
CYC → CYC → RTX → BSC | £22,820.93 | 8.9035 | 0.32 | £3,894.36 | £12,075.42 |
CYC → RTX → CYC → BSC | £23,176.00 | 8.9086 | 0.0051 | £355.07 | £69,709.63 |
RTX → CYC → CYC → BSC | £23,755.25 | 8.9131 | 0.0045 | £579.25 | £127,456.12 |
Treatment-naïve subgroup
|
CYC → CYC → BSC | £18,645.81 | 8.6491 | – | – | – |
CYC → CYC → RTX → BSC | £22,429.08 | 8.9435 | 0.29 | £3,783.27 | £12,850.76 |
CYC → RTX → CYC → BSC | £22,793.54 | 8.9480 | 0.0045 | £364.46 | £81,603.50 |
RTX → CYC → CYC → BSC | £23,636.83 | 8.9507 | 0.0027 | £843.29 | £317,037.96 |
Recurrent-disease subgroup (eligible for additional CYC treatment)
|
CYC → BSC | £17,593.48 | 8.2548 | – | – | – |
CYC → RTX → BSC | £22,295.52 | 8.6773 | 0.4225 | £4,702.04 | £11,129.22 |
RTX → CYC → BSC | £22,620.65 | 8.6836 | 0.0063 | £325.14 | £51,841.87 |
Recurrent-disease subgroup (ineligible for additional CYC treatment)
|
BSC | £15,747.48 | 7.9379 | – | – | – |
RTX → BSC | £21,132.39 | 8.4412 | 0.5033 | £5,384.90 | £10,699.45 |
CYC-intolerant subgroup
|
BSC | £15,747.48 | 7.9379 | – | – | – |
RTX → BSC | £21,184.13 | 8.4200 | 0.48 | £5,436.64 | £11,277.29 |
The additional work undertaken by the ERG indicated that including RTX in the treatment sequence increases health benefits, compared with the current standard treatment sequence (that is, a treatment sequence that does not include RTX). In the analyses for all patients, the treatment-naïve subgroup and the recurrent-disease subgroup (for patients who are eligible for further CYC treatment), the ICER associated with adding RTX after CYC treatment had been exhausted was in the range of £11,129 to £12,851 per QALY gained. However, in each of these analyses, the ICERs associated with administering RTX earlier in the treatment sequence were greater than £50,000 per QALY gained—sometimes substantially so. It is particularly important to note the substantial reductions in total lifetime costs resulting from the ERG’s amended model, compared with the manufacturer’s original model (the manufacturer estimated lifetime costs in the region of £100,000 per patient, compared with the ERG’s estimates of around £20,000). This is almost entirely due to the apparent substantial overestimation of costs associated with the uncontrolled-disease health state in the manufacturer’s original model.
In the recurrent-disease subgroup (for patients who are ineligible for further CYC treatment) and in the CYC-intolerant subgroup, the ICER associated with treating patients with RTX rather than best supportive care was in the range of £10,699 to £11,277 per QALY gained. In these scenarios (and in all other scenarios), best supportive care represents continued treatment to maintain patients in a state of low-grade ‘grumbling’ disease. The ERG noted that while these analyses were useful, they were limited and may represent underestimates of the true ICER because relevant comparators such as mycophenolate mofetil were not included in the model.
3.4 Conclusions of the Evidence Review Group Report
On the basis of the clinical evidence provided in the manufacturer’s submission, RTX given at a dose of 4 × 375 mg/m2 has an effectiveness and safety profile similar to that of oral CYC in terms of induction of remission in adults with AAV and de novo disease. For patients who have relapsed once on CYC, RTX appears to be more effective than oral CYC in terms of inducing remission in adults with generalized, severe AAV. However, the submitted evidence was limited because it included only one trial that incorporated the licensed regimen, and it did not include evidence on long-term safety and effectiveness.
The ERG could not offer robust estimates of the likely cost effectiveness of RTX based upon the original version of the manufacturer’s economic model. Several amendments had to be made. On the basis of the ERG’s amended version of the manufacturer’s economic model, RTX appeared to represent a cost-effective addition to the treatment sequence at a cost-effectiveness threshold of £20,000 per QALY gained, provided it was received only by patients who had exhausted their use of CYC.