3 Evidence

The appraisal committee (section 8) considered evidence submitted by GlaxoSmithKline, reviews of the submissions by the evidence review group (ERG; section 9) and evidence provided in the Decision Support Unit report (DSU; section 9).

3.1 The company's submission focused on a subgroup of the patients whose disease met the criteria specified in the marketing authorisation. The company explained that, being aware of NHS resources and to identify people who are most likely to benefit from belimumab, the submission focused on a high disease activity subgroup (hereafter referred to as the target population). The target population is adults with active autoantibody-positive systemic lupus erythematosus with evidence for serological disease activity (defined as positive anti-double-stranded DNA and low complement) and a Safety of Estrogen in Lupus National Assessment – Systemic Lupus Erythematosus Disease Activity Index (SELENA-SLEDAI) score of greater than or equal to 10.

3.2 The company submitted clinical data for all of the patients enrolled in the clinical trials and for the target populations in the trials. Data were presented for both populations individually for each trial and combined across trials. The patient characteristics and results described in the clinical effectiveness section of this document focus on the company's target population.

Clinical effectiveness

3.3 The main evidence for the clinical effectiveness of belimumab was from 2 phase III clinical trials. The BLISS‑52 (n=865) and BLISS‑76 (n=819) trials were randomised, double-blind, placebo-controlled, parallel-group studies with follow‑up at 52 weeks and 76 weeks respectively. In these trials, belimumab plus standard care (hereafter referred to as belimumab) was compared with placebo plus standard care (hereafter referred to as standard care). Standard care included: non-steroidal anti-inflammatory drugs, antimalarials, corticosteroids or other immunosuppressants (azathioprine, methotrexate and mycophenolate mofetil) either alone or in combination. Although each of the BLISS trials were 3‑arm trials (belimumab 10 mg/kg, belimumab 1 mg/kg and placebo), only results for the 10 mg/kg belimumab dose were presented in the company's submission because this is the dose covered by the marketing authorisation.

3.4 Adult patients (aged 18 years or older) who met the American College of Rheumatology criteria for systemic lupus erythematosus and had active autoantibody-positive disease and a SELENA-SLEDAI score of 6 or more at screening were eligible for enrolment in the BLISS trials. Patients with severe active lupus nephritis or central nervous system lupus were excluded from the trials. Of the patients in the standard care and belimumab 10 mg/kg arms (n=1,125), 52% (n=585) had disease that met the criteria for the marketing authorisation and 35% (n=396) had disease that met the criteria for the target population.

3.5 The BLISS‑52 trial recruited patients from South America, Asia and eastern Europe, whereas the BLISS‑76 trial recruited patients from the US, Canada, Europe (western and eastern) and Israel. In the BLISS‑52 trial, approximately 42% of the target population were Asian. In the BLISS‑76 trial most of the target population were white (around 65%). Over 90% of the target population included in the trials were women and most (approximately 85%) were aged 45 years or younger. In the target population, over 90% of the patients had at least 1A or 1B British Isles Lupus Assessment Group (BILAG) organ involvement and approximately 70% had at least 1A or 2B BILAG organ involvement. For the target population, mean SELENA-SLEDAI score was approximately 13 in both trials. About 85% of patients in the target population had a physician's global assessment score of between 1 and 2.5. Most of the patients had a range of manifestations of systemic lupus erythematosus, mainly involving mucocutaneous, immunological and/or musculoskeletal damage.

3.6 The company presented results from the BLISS‑52 and BLISS‑76 trials separately and pooled. The primary outcome of both studies was the response rate at week 52 compared with baseline, assessed with the Systemic Lupus Erythematosus Responder Index (SRI). With the SRI criteria, a response was defined as: a reduction of at least 4 points in SELENA-SLEDAI score (regarded as clinically meaningful); no new BILAG A organ domain score; no more than 1 new BILAG B organ domain score; and no worsening in physician's global assessment score (increase of less than 0.3).

3.7 For the primary outcome of SRI response at 52 weeks, statistically significant differences were observed between belimumab and standard care in both trials. In the BLISS‑52 trial, for the target population, 67% of patients on belimumab had disease that responded compared with 41% of patients on standard care (odds ratio [OR] 3.0; 95% confidence interval [CI] 1.7 to 5.2). In the BLISS‑76 study, for the target population the response was 57% for belimumab compared with 34% for standard care (OR 2.5; 95% CI 1.3 to 4.6). In the pooled analysis for the target population, 63% of the patients on belimumab had disease that responded, compared with 38% of those on standard care (OR 2.7; 95% CI 1.8 to 4.1). In the BLISS‑76 trial, the target population showed a statistically significant difference in response rate between belimumab and standard care at 76 weeks (p=0.02).

3.8 For the individual components of the SRI, which were secondary outcomes in the trials, a greater proportion of patients on belimumab in both BLISS trials had a reduction of at least 4 points in SELENA-SLEDAI score compared with standard care. In the pooled analysis for the target population, 65% of patients on belimumab had a reduction of at least 4 points in SELENA-SLEDAI score compared with 41% on standard care (OR 2.6; 95% CI 1.7 to 3.9), which was statistically significant. For the outcomes of no new BILAG 1A or 2B organ domain involvement and no worsening in physician's global assessment, results from BLISS‑52 for the target population showed a statistically significant improvement with belimumab compared with standard care, whereas results from BLISS‑76 did not. However, there was a statistically significant improvement for both these outcomes in the pooled analysis for the target population.

3.9 For other secondary outcomes, in the pooled analysis of the target population 16% of patients on belimumab compared with 7% of patients on placebo (OR 2.43; 95% CI 1.05 to 5.65) had an average prednisone dose reduction of greater than or equal to 25% from baseline, to less than or equal to 7.5 mg per day, during weeks 40 to 52. There were no differences in the Systemic Lupus International Collaborating Clinics (SLICC) index of organ damage in BLISS‑52, BLISS‑76 or the pooled analysis.

3.10 Quality-of-life measures, the SF‑36 and EQ‑5D, were also collected as secondary outcomes. At week 24 in the pooled analysis of the target population, there was a statistically significant mean change from baseline EQ‑5D index for belimumab compared with standard care, but this was not maintained at week 52. The pooled analysis of the target population showed no statistically significant difference in mean SF‑36 physical component summary scores between belimumab and standard care at weeks 24 or 52. In the pooled analysis of the target population for functional assessment of chronic illness therapy (FACIT)-fatigue scores, the difference in scores was statistically significant at weeks 8 and 12 but not thereafter. In the individual trials, there was a statistically significant difference in FACIT-fatigue scores for the total population in favour of belimumab in the BLISS‑52 trial at week 52 but not in the BLISS‑76 trial.

3.11 In the pooled total trial population, the percentage of people defined as being of African American or African family origin (n=100) meeting the primary end point was higher in the standard care group (44%) than in the belimumab group (36%). This compared with an overall response rate of 39% in the standard care group and 51% in the belimumab group in the pooled total trial population. For patients of all other family origins, the belimumab group had higher response rates than the standard care group.

3.12 Adverse event data were taken from the total population included in the BLISS trials (that is, not just the target population) and from a phase II extension study (LBSL99). Over 90% of patients in each arm experienced 1 or more adverse events. The most frequent (occurring in more than 10% of patients) events were headache, upper respiratory tract infection, arthralgia, nausea, urinary tract infection, diarrhoea and fatigue. Of these events, only diarrhoea and nausea occurred slightly more frequently in the belimumab groups than in the standard care groups. Serious adverse events were experienced by 17% in the 10 mg/kg belimumab group, compared with 16% in the standard-care group. Across the double-blind treatment periods, 14 people died, including 3 (0.4%) in the standard care group, 5 (0.7%) in the 1 mg/kg group and 6 (0.9%) in the 10 mg/kg belimumab group. Four deaths were infection-related: 1 in the standard care group, 1 in the 1 mg/kg belimumab group and 2 in the 10 mg/kg belimumab group. Infection may have contributed to the deaths of 2 further patients (1 in the 1 mg/kg belimumab group and 1 in the 10 mg/kg belimumab group). There were 2 suicides, both in patients receiving belimumab (1 in the 1 mg/kg group and 1 in the 10 mg/kg group), and 1 cancer-related death in a patient receiving 1 mg/kg belimumab. In the long‑term open-label extension phase of the phase II study, the incidence of adverse events and severe adverse events remained stable or declined over time through 5 years of exposure.

3.13 The company explained that some people with severe, highly active systemic lupus erythematosus routinely receive rituximab. No studies were identified that directly compared belimumab with rituximab. However, in a study that compared rituximab with placebo (EXPLORER) in patients with moderate-to-severe systemic lupus erythematosus disease activity, no statistically significant differences were reported in major or partial clinical responses between the rituximab group and the placebo group. In addition, the rituximab trial demonstrated no difference in secondary end points between the rituximab group and the placebo group over 52 weeks. The company stated that differences in the end points considered and the patient populations precluded any meaningful indirect comparison between the belimumab and rituximab studies.

Further evidence and proposed evidence collection provided by the company after the fourth committee meeting

3.14 The company acknowledged the uncertainties in the evidence base for belimumab and the assumptions used in the economic modelling. It considered that some of this uncertainty could be addressed by collecting real-world evidence that would inform a future review of the technology appraisal guidance. It submitted a proposal to collect additional evidence on belimumab over a minimum of 3 years through the British Isles Lupus Assessment Group (BILAG) registry, which has 29 specialist centres around the UK and has already enrolled over 200 patients with systemic lupus erythematosus. It stated that this would generate real-life data for belimumab as prescribed in UK clinical practice for the target population of patients with highly active systemic lupus erythematosus despite current standard of care.

3.15 The company's research proposal included details about sample size, data collection and timing of assessments:

  • It was anticipated that, based on current recruitment rates, around 9–11 patients receiving belimumab would enroll in the registry each month, with an estimated total of 300 patients after 3 years.

  • The data collected for belimumab would be consistent with those already being collected in the registry for biological treatments:

    • Efficacy data comprising clinical response (measured by BILAG Index 2004 and SLEDAI‑2K) and organ damage accrual (using the SLICC Damage Index and BILAG Index 2004).

    • Incidence of serious adverse events such as hospitalisation for infection, malignancy and death.

    • Patient-reported outcomes using EQ‑5D, SF‑36 and LupusQoL.

    • Other data such as previous and concomitant treatment.

  • Patients would be assessed before receiving a new biological treatment or before re-treatment. Further clinical assessment would be at 3, 6, and 12 months, and then annually.

  • Patients would complete questionnaires every 6 months for 2 years and annually thereafter.

3.16 The company considered that the data would address many areas of uncertainty that had been identified by the committee, such as treatment duration, type of standard of care in clinical practice in England, treatment benefit for the full range of disease manifestations, steroid-sparing effect, impact on quality of life, stopping rule adherence, maintenance of treatment effect, development of organ damage and safety data.

3.17 The company also submitted further supportive evidence, including organ damage and 5‑year safety data from 2 open-label, phase III extension studies that included people who had completed the BLISS‑52 and BLISS‑76 studies.

Cost effectiveness

3.18 A de novo decision-analytic model was developed by the company. The model is a micro-simulation that incorporates the interaction between patient characteristics, disease activity, medication (corticosteroid use), risk of organ damage development (a person with systemic lupus erythematosus could potentially develop damage in 12 different organ systems) and mortality. The company presented results on the target population, as well as the proportion of patients in the trial whose disease met the criteria in the marketing authorisation (hereafter referred to as the marketing authorisation population) and total trial populations. The model results presented here focus on the target population.

3.19 The health states in the model were informed by data from the BLISS trials, observational cohort data (the Johns Hopkins cohort, see section 3.21), and other data from the literature. A patient's baseline characteristics were simulated based on the pooled target population characteristics in the BLISS trials. The BLISS clinical trials were used to inform the likelihood of response at week 24 (based on a SELENA-SLEDAI score decrease of 4, this being the basis for the company's proposed continuation rule in which belimumab would be continued for people whose disease had such a response after 24 weeks), the change in SELENA-SLEDAI score up to week 52, the likelihood of discontinuation, and the effect of SELENA-SLEDAI score on utility and treatment costs. Data from the literature were used to inform the standardised mortality rate for a given SELENA-SLEDAI score, and quality-of-life and cost impacts of long‑term damage to each organ system.

3.20 The patient entered the model in which their lifetime history of systemic lupus erythematosus was simulated, based on the BLISS trial data. A patient's characteristics were 'cloned' so that the same modelled 'patient' entered both standard care plus belimumab 10 mg/kg (hereafter referred to as belimumab) and standard care only (hereafter referred to as standard care) treatment paths and then worked through the model. For a patient entering the model assigned to either belimumab or standard care, it was first determined whether the patient survived for that year. A surviving patient on belimumab could then either continue with belimumab treatment or discontinue treatment. The treatment discontinuation rate was calculated from the BLISS trial data. Patients discontinued treatment after week 24 if they did not have an improvement in SELENA-SLEDAI score of 4 points or more. An annual discontinuation rate in patients whose disease responded to treatment was estimated to be 8% per year.

3.21 Prediction models based on data from the Johns Hopkins cohort were used to predict change in adjusted mean SLEDAI score (which is used as a proxy for SELENA-SLEDAI score), average corticosteroid dose per year, risk of organ damage and risk of death. The Johns Hopkins cohort reported data on a large population of patients with systemic lupus erythematosus from Baltimore, Maryland, US, of whom 93% were women, 52% were white and 38% were black. Analyses were conducted on a dataset of 1,282 people, with follow‑up of greater than 2 years and data after 1992. Mean age at diagnosis was 33 years and mean SLEDAI score at first visit was 3.32. Few people with SLEDAI scores of 10 or more remained at the end of this observational study.

3.22 In the first year of the simulation, the effects on disease activity as observed in the BLISS trials were applied, measured by SELENA-SLEDAI score. A linear regression model based on data from the BLISS trials was used to predict the change in SELENA-SLEDAI score at 52 weeks. For subsequent cycles, disease activity was predicted using regression equations based on the natural history data from the Johns Hopkins cohort. Because the baseline characteristics from the Johns Hopkins cohort were different from the patient characteristics in the pooled BLISS trials (patients in the Johns Hopkins cohort had lower disease activity than those in the BLISS trials), the company adjusted the constant in the regression to obtain a better fit to the data.

3.23 Corticosteroid use was calculated based on a regression equation from the Johns Hopkins cohort, with disease activity as measured by mean SLEDAI score as the sole independent variable. For each organ system contained within the SLICC Damage Index, the probability of damage during that year was calculated based on the patient's characteristics and disease activity at that time. The company also developed a survival model using the Johns Hopkins cohort, adjusting it by standardised mortality ratios from the literature. Average costs and utilities calculated from regression analyses were assigned to a patient's health state for that particular year. Costs and utilities were recorded together with clinical outcomes for that patient. Time was then increased by 1 year and the process was repeated for the lifetime of the patient. These yearly cycles continued until a patient died. Utilities and costs were discounted at 3.5%. An NHS and personal social services perspective was adopted. Adverse events were not included in the model.

3.24 The baseline quality of life assumed in the cost-effectiveness analysis was determined by a regression equation (which accounted for age, family origin and SELENA-SLEDAI score), which was derived from the BLISS trials. Disutility multiplier values for each type of organ damage were identified from a search of the literature. These disutility multipliers were applied to the utility score if a patient developed organ damage in the model cycle. Costs in the analysis were limited to direct medical costs and costs associated with disease activity and long‑term organ damage. Total resource use varied according to disease severity and was determined using a linear regression analysis. A literature search was conducted to identify the cost of organ damage. All costs were inflated to 2010 values. The base case considered only the additional acquisition costs for belimumab. Because belimumab is given in addition to standard care, it was assumed that the costs for standard care treatments would be the same for people on belimumab as for those not on belimumab and so were not included. The administration cost of £126 for belimumab was calculated based on 2 hours of senior hospital staff nurse time (£63 per hour): 1 hour for the infusion and another 1 hour for patient preparation and monitoring post-infusion. It was assumed that the first year annual cost of treatment and administration of belimumab was £10,918 and, in subsequent years, was £10,138, based on a cost of belimumab of £114.30 for a 120‑mg vial and £381 for a 400‑mg vial. At the time of submission, the vial price for belimumab had not been finalised, so the expected vial list price was used in the base-case analyses. The effect on cost effectiveness of a maximum expected vial price for both the 120‑mg and 400‑mg vials was investigated in a scenario analysis. The inclusion of a cost for standard care and different costs of administration were also explored in scenario analyses.

3.25 The model showed lower disease activity for patients on belimumab than in patients on standard care only, which led to decreased corticosteroid dose and decreased risk of organ damage, and contributed to a difference in mortality risk. The model predicted that patients on belimumab live longer than those on standard care. Although a decreased duration of damage was shown for organs on which belimumab has a large effect (cardiovascular, pulmonary and renal), the duration of damage for other organ systems is increased because of the prolonged life expectancy.

3.26 The model predicted that patients treated with belimumab, in the target population, live on average 2.9 years longer (34.9 compared with 31.9 years), have a reduction in average adjusted mean SLEDAI score, have a reduced cumulative monthly corticosteroid dose and similar total SLICC organ damage score compared with those on standard care only. Treatment with belimumab provided an estimated additional 1.1 life years and 0.8 quality-adjusted life years (QALYs; both discounted values). For both treatment groups, the organ damage costs were the highest expense. In total, the organ damage costs were lower for patients treated with belimumab. The costs related to disease activity were similar in the 2 treatment arms. Because of their increased life expectancy and the cost of belimumab treatment, costs were higher for patients receiving belimumab than for those on standard care.

3.27 For the target population, not including the patient access scheme, total costs were £157,291 for belimumab and £105,366 for standard care. Total QALYs were 10.61 for belimumab compared with 9.81 for standard care. The incremental costs were therefore £51,925, and the incremental QALYs 0.806. This resulted in an incremental cost-effectiveness ratio (ICER) of £64,410 per QALY gained. The probabilistic sensitivity analysis results showed that at a threshold of £30,000 per QALY gained, there is a 0% probability that belimumab is cost effective compared with standard care.

3.28 In comparison, the ICER for the marketing authorisation population was £66,170 per QALY gained (undiscounted life years gained of 2.1 years, reflecting a difference in estimated survival of 35.0 compared with 32.8 years). The ICER for the total trial population (which included a wider population than that specified in the marketing authorisation) was £82,909 per QALY gained.

3.29 In sensitivity analyses conducted in the target population analysis, factors affecting cost effectiveness were: the treatment effect regression to estimate the effect of belimumab after 52 weeks, the size of the company's adjustment to the constant of the disease activity prediction equation, the probability of discontinuation, the effect of the adjusted mean SLEDAI score on mortality, and the natural history models for pulmonary and renal involvement. Scenario analyses were conducted, with resulting ICERs ranging from £50,114 to £77,707 per QALY gained. Removing the continuation rule increased the ICER to £72,207 per QALY gained, and increased vial prices of £127.80 for the 120‑mg vial and £426 for the 400‑mg vial (the maximum expected vial price) resulted in an ICER of £71,297 per QALY gained.

3.30 The patient access scheme comprises a simple discount, which was accepted by the Department of Health and incorporated into the analysis of belimumab compared with standard care. An ICER with the patient access scheme was provided. However, the level of the discount and the results from the economic analysis incorporating the patient access scheme are commercial in confidence.

3.31 A comparison of the costs of belimumab and rituximab, taking into account the patient access scheme, was also provided by the company. The company calculated the cost of rituximab from the administration schedule used in EXPLORER. A course of rituximab was 1,000 mg, provided on days 1, 15, 168 and 182. The total drug cost of rituximab was £6,985 per year. The cost of belimumab per year was commercial in confidence.

Further evidence submitted by the company after the first committee meeting

3.32 In response to consultation, the company presented long‑term efficacy and safety trial data from the open-label, phase II extension study (LBSL99; Petri et al. 2011) for belimumab, which suggested continued efficacy with belimumab and safety over a 6‑year follow‑up period. Patients with seropositive disease treated with belimumab showed sustained improvement in disease activity and a decline in BILAG scores and flares over 6 years, accompanied by reductions in corticosteroid use and autoantibody levels. The abstract provided by the company showed a mean reduction in corticosteroid use of 4.7 mg per day, an average reduction of 34.4% from the baseline dose, by the end of 6 years of follow‑up. The company calculated that based on 6‑year follow‑up the annual discontinuation rate was approximately 13% in this trial.

3.33 As well as further clinical evidence, in response to consultation the company submitted additional cost-effectiveness evidence using the same assumptions as in the original base-case model, but incorporating a maximum treatment duration of 6 years and using the confirmed list price for belimumab. The company's revised base case resulted in an ICER of £47,342 per QALY gained, with an incremental cost of £28,705 and incremental QALYs of 0.61. In a scenario analysis conducted by the company on the revised base-case analysis, the continuation rule for belimumab was changed from a SELENA-SLEDAI score of greater than or equal to 4 to greater than or equal to 6 and the health effects discount rate lowered from 3.5% to 1.5%. These scenarios had the effect of lowering the ICER to £40,863 and £31,988 per QALY gained respectively. When both scenarios were applied together, they lowered the ICER to £27,807 per QALY gained, with an incremental cost of £20,766 and incremental QALYs gained of 0.747.

3.34 The company stated that the change from an unlimited treatment duration to a maximum of 6 years was made in response to comments in the appraisal consultation document about the need to align the use of belimumab more closely with how clinicians would consider using belimumab in clinical practice. While recognising the lack of any direct evidence about optimal treatment duration, the company supported the use of belimumab for up to 6 years with the newly available long‑term data for belimumab from the phase II extension study (see section 3.32). The company also explained that other treatments for systemic lupus erythematosus, such as immunosuppressants, are prescribed for 2–5 years to maintain suppression of disease activity. The company stated that it believed that 6 years of treatment with belimumab was long enough for the benefits of belimumab on controlling high disease activity to have an important impact on reducing long‑term morbidity.

3.35 According to the company it was appropriate to use NICE's clarification to section 5.6 of the guide to the methods of technology appraisal 2008 on the discounting of health benefits in special circumstances for a number of reasons; namely, the nature of systemic lupus erythematosus and the fact that belimumab has been shown to result in clinically important reductions in disease activity, and has the potential to provide important long‑term benefits including reduced organ damage, reduced use of high-dose corticosteroids, along with their associated risks, and consequently improved survival. Therefore, the company considered that the discount rate of 1.5% for health effects rather than the 3.5% normally applied in technology appraisals was appropriate. Furthermore, the company stated that by applying a continuation rule at 24 weeks of a SELENA-SLEDAI score greater than or equal to 6 rather than 4, a more efficient use of NHS resources could be made.

Further evidence provided by the company after the appeal

3.36 After the appeal, the company submitted a further revised base case using a revised patient access scheme. This used the same assumptions as in the original base-case model (that is, lifetime treatment with a continuation rule of a SELENA-SLEDAI score greater than or equal to 4), but incorporated an annual discontinuation rate of 13% and an administration cost of belimumab of £154, reflecting the infusion administration cost used in NICE's technology appraisal guidance on tocilizumab for the treatment of rheumatoid arthritis. The total costs were £146,056 for belimumab and £105,366 for standard care. Total QALYs were 10.47 for belimumab compared with 9.81 for standard care. The incremental costs were therefore £40,691, and the incremental QALYs 0.66. This resulted in an ICER of £61,328 per QALY gained. ICERs with the patient access scheme were also provided. These were marked commercial in confidence because of the confidential nature of the patient access scheme.

3.37 The company also presented scenario analyses including a maximum treatment duration of 6 and 10 years, alternative treatment continuation rules and a range of treatment discontinuation rates. The alternative discontinuation rates included the 8% discontinuation rate used in the original base case, and a variable discontinuation rate of 13% up to year 5 and a 30% rate afterwards. The company stated that the variable discontinuation rate more closely represented the distribution of treatment durations likely to be prescribed in clinical practice for people in the target population. The company's additional evidence also included a cost comparison between rituximab and belimumab.

Further evidence provided by the company after the fourth committee meeting

3.38 The company submitted additional economic analyses for belimumab compared with standard care in which the base-case maximum treatment duration was 3 or 5 years. They incorporated a treatment continuation rule at 6 months of a change in SELENA-SLEDAI score of 4 or more, a discontinuation rate of 8% in year 1 and 11.7% thereafter and assumed that people who stopped having belimumab reverted to average SELENA-SLEDAI score for standard care. Scenario analyses that included different continuation rules were also provided. The base-case results and those of the scenario analyses are confidential and cannot be presented here. The company noted that the base-case ICERs for belimumab compared with standard care with a maximum treatment duration of 3 or 5 years were lower than those it had previously submitted.

3.39 The company asserted that a robust comparison of the relative efficacy of belimumab and rituximab was not feasible because of a lack of direct or indirect data from randomised controlled trials.

ERG's critique of the company's original submission

3.40 The ERG stated that the marketing authorisation population and the target population that formed the focus of the submission were subgroups identified from post-hoc analyses aimed at identifying people with the greatest response to belimumab. The ERG noted that, according to clinical opinion, the SELENA-SLEDAI (a component of the SRI and 1 of the measures used to identify people in the target population) is not commonly used to define high disease activity in clinical practice.

3.41 The ERG commented that, although both trials included adults with active autoantibody-positive systemic lupus erythematosus, the population in BLISS‑76 is more likely to be similar to that of England and Wales than that of BLISS‑52, so the results from BLISS‑76 are more likely to be generalisable to the UK. This was because the differences in geography and family origin between the patients in the trials were considered to potentially affect the results of the trials as well as reflecting differences in clinical practice. The ERG stated that, for the target population, the results from the BLISS‑52 trial were more favourable for belimumab than those from BLISS‑76, and BLISS‑52 provided more patients to the pooled target population than BLISS‑76 (55% compared with 45%). Therefore, results favourable to belimumab for the pooled target population were more strongly driven by the contribution from the BLISS‑52 target population. The ERG, therefore, had concerns about the relevance of the pooled results for patients in England and Wales.

3.42 The ERG highlighted that information on SLEDAI and SF‑36 changes in the rituximab EXPLORER trial were available, and that randomised controlled trials for both rituximab and belimumab recorded BILAG score changes.

3.43 The ERG considered that the company's model was complex, though generally well constructed. It noted that the model conformed to the NICE reference case and that the longer‑term effects of systemic lupus erythematosus had been modelled well, using the Johns Hopkins cohort. An ERG cross-check of the probabilistic modelling for the target population resulted in a central estimate for the ICER of £65,530 per QALY gained.

3.44 The ERG commented that there was a lack of clarity around the reasons for patients' discontinuation of belimumab, the derivation of the 8% annual discontinuation rate among patients showing a response to belimumab at week 24, and whether extrapolation using this value was reasonable. Sensitivity analyses by the company showed that a low discontinuation rate, such as 2%, increased the ICER for belimumab to £85,893 per QALY gained, whereas a higher discontinuation rate, such as 14%, reduced the ICER to £54,518 per QALY gained.

3.45 The ERG stated that the model assumed that patients whose disease had not responded to belimumab by week 24 (a third of patients) experienced the average SELENA-SLEDAI score seen with standard care. Thus the actual experience of the patients whose disease had not responded to belimumab was not used in the model. There were approximately an equal proportion of patients whose disease had responded to standard care (with an average change in SELENA-SLEDAI score of −6.9) and whose disease had not responded to standard care (with an average change in SELENA-SLEDAI score of −1.1) at week 24 in the pooled target population. The average change in SELENA-SLEDAI score for patients receiving standard care was therefore estimated to be −4.1, while the average change in SELENA-SLEDAI score for the patients whose disease had not responded to belimumab was −0.9 at week 24. The ERG considered that the company's assumption that patients whose disease does not respond to belimumab have the same change in SELENA-SLEDAI scores as all patients receiving standard care is likely to overestimate the average impact on SELENA-SLEDAI scores in the belimumab arm, both between week 24 and 52 and beyond week 52, leading to an underestimation of the ICER.

3.46 The ERG noted that a higher adjusted mean SLEDAI score was associated with an increased probability of the patient dying and of a patient developing particular organ involvement. The economic modelling did not take into account a patient's history before entry into the trial and this may also have exaggerated the impact that changes in SELENA-SLEDAI score had on the adjusted mean SLEDAI score for belimumab compared with standard care, with the likely result that the base-case ICER was an underestimate. This is potentially important when comparing the Johns Hopkins cohort, in which most patients had SELENA-SLEDAI scores of less than 10, with the target population, who all had scores of greater than 10 at baseline.

3.47 The ERG stated that the reason for adjusting the Johns Hopkins cohort survival model by standardised mortality ratios from the literature was unclear and may have tended to exaggerate the impact of the individual covariates within the Johns Hopkins cohort survival model. Unpublished data from a UK study obtained by the ERG also suggested that the standardised mortality ratios used by the company may not accurately represent a UK cohort. An exploratory analysis using the lower standardised mortality ratios derived from the UK study increased the ICER by approximately £6,000, to £70,860 per QALY gained.

3.48 The ERG highlighted that the constant in the SELENA-SLEDAI change regression equation from the Johns Hopkins data was originally 2.0577 but was adjusted by the company to 3.0 to improve the fit to belimumab trial data after week 52. Sensitivity analyses by the company showed that using the original value of the constant term increased the ICER by approximately £29,000, to £93,654 per QALY gained.

3.49 The ERG considered the impact of using different administration costs than those used in the model (£126). The ERG's exploratory analysis found that, if costs were in line with those from a previous appraisal of another intravenous monoclonal antibody drug (NICE's technology appraisal guidance on tocilizumab for the treatment of rheumatoid arthritis), which had a similar duration of administration and an administration cost of £154, then the ICER would increase by approximately £2,500 to £66,907 per QALY gained. If the full day-case cost was used (£432), then the ICER would be higher by approximately £27,000, at £91,699 per QALY gained.

3.50 The ERG completed an exploratory analysis that used the estimates from the individual BLISS trials in the disease activity regression equation rather than the pooled estimate. This analysis demonstrated that the economic model was not particularly sensitive to the use of estimates from the individual trials. Using BLISS‑76 as the source of the regression increased the ICER by approximately £2,000, to £66,318 per QALY gained.

Critique by the ERG of the company's new evidence provided after the first appraisal committee meeting

3.51 The ERG commented on the new evidence provided by the company about the long‑term corticosteroid sparing effect of belimumab. The ERG noted that the basis of the calculations was not clear and the ERG questioned whether the average baseline corticosteroid use was calculated for the same patients in whom corticosteroid use was estimated at 6 years. The ERG stated that the company proposed that the corticosteroid sparing effect, together with other belimumab benefits such as reduced flare frequency, would reduce the development of organ damage and would therefore translate into long‑term benefit. However, the ERG stated that data are only available for 6 years, which indicates that there is a substantial degree of uncertainty over whether the effects observed in the data would translate into longer‑term effects.

3.52 The ERG reviewed and critiqued the company's additional economic analysis submitted after consultation. The ERG noted that the company's revised base-case model was based on 6 years maximum treatment duration, while the original model had some patients on treatment for 40 years. The ERG considered that the maximum duration of belimumab treatment was uncertain because clinical opinion is likely to vary. The ERG stated that the company's revised base-case model also assumed that while the SELENA-SLEDAI scores for the patients at the end of year 6 revert to scores expected for patients receiving standard care, the adjusted mean SLEDAI score continues to show benefit, which could indicate a sustained reduction in organ damage in the treatment arm. The ERG also noted that, given an annual discontinuation rate of 8% (as in the original submission) or the rate observed in the phase II extension study (13% annual discontinuation rate), if a maximum treatment duration of 6 years was imposed, a considerable number of patients benefiting from belimumab would have treatment withdrawn. The ERG calculated that, of 339 patients receiving belimumab at the end of the second year of treatment in the phase II extension study, 167 were still receiving treatment at the end of the sixth year. The ERG commented that the company did not address tapering-off rules, the issue of potential rebound phenomena, the ethical considerations of withdrawing treatment, or the possibility of reintroducing treatment and the effect of this on cost effectiveness.

3.53 The ERG evaluated the continuation rule used in the analyses. The ERG observed that changing the continuation rule so that a minimum SELENA-SLEDAI improvement of 6 is needed to continue treatment reduces the benefits the patients get from belimumab, but it accordingly reduces costs and the ICER by a greater proportion than when a continuation rule of a minimum SELENA-SLEDAI improvement of 4 is applied.

3.54 The ERG noted that the company suggested that belimumab treatment for systemic lupus erythematosus should be appraised using a 1.5% discount rate for health benefits. The ERG noted that the evidence presented showed a beneficial response to belimumab lasting at least 6 years in an appreciable population of patients. The ERG noted that the company considered that this early effect of belimumab, together with the observed 34% reduction in corticosteroid usage, would translate into long‑term benefit by reducing the development of organ damage. The ERG commented that the extent to which short‑term benefits translated into longer‑term benefits was uncertain and presented data showing that in the economic modelling 63% of the incremental QALY gain (undiscounted) was accrued within 30 years.

3.55 The ERG completed additional analyses, applying a lifetime treatment duration and a maximum 6‑year treatment duration. For both of these, separate scenarios were modelled that assumed no continuation rule at 24 weeks, a continuation rule at 24 weeks of SELENA-SLEDAI score greater than or equal to 4 and a continuation rule at 24 weeks of SELENA-SLEDAI score greater than or equal to 6. These analyses also assumed an annual discontinuation rate of 13% after 24 weeks and an administration cost of £154, as had been used in previous appraisals of intravenous monoclonal antibody treatments for rheumatoid arthritis. Benefits and costs were discounted at 3.5%. Analyses were presented both with and without the patient access scheme.

3.56 Assuming a lifetime treatment duration for belimumab, the ICERs without the patient access scheme were £90,002, £61,193 and £53,744 per QALY gained for the scenarios assuming no continuation rule at 24 weeks, a continuation rule at 24 weeks for a SELENA-SLEDAI score of greater than or equal to 4, and a continuation rule at 24 weeks for a SELENA-SLEDAI score of greater than or equal to 6 respectively. The incremental costs in these scenarios were £57,526, £40,499 and £31,878 respectively and incremental QALYs 0.639, 0.662 and 0.593 respectively. ICERs with the patient access scheme were provided. These were marked commercial in confidence because of the confidential nature of the patient access scheme.

3.57 Assuming a maximum 6‑year treatment duration for belimumab, the ICERs without the patient access scheme were £70,942, £47,382 and £42,108 per QALY gained for the scenarios assuming no continuation rule at 24 weeks, a continuation rule at 24 weeks for a SELENA-SLEDAI score of greater or equal to 4 and a continuation rule at 24 weeks for a SELENA-SLEDAI score of greater or equal to 6 respectively. The incremental costs in these scenarios were £37,888, £26,300 and £21,104 respectively and incremental QALYs 0.534, 0.555 and 0.501 respectively. ICERs with the patient access scheme were provided. These were marked commercial in confidence because of the confidential nature of the patient access scheme.

Critique by the ERG of the company's further evidence provided after the appeal

3.58 The ERG reviewed the company's revised base case and confirmed that no structural changes were made to the model originally submitted and that only inputs were changed. It noted that, in 1 scenario, an error in the calculations resulted in an incorrect value, which when corrected raised the ICER slightly. The ERG stated that many of its previous concerns about the economic model had not been addressed in the additional evidence and that, if justified, these would increase the estimates of cost effectiveness. The ERG also highlighted the sensitivity of the model to estimates of pulmonary damage and uncertainties in these costs. With respect to the comparison of costs presented by the company, the ERG noted that there was still no research data that would allow for either a direct or an indirect comparison between belimumab and rituximab. The ERG noted that, in a recent systematic review of rituximab therapy in systemic lupus erythematosus, most observational studies used less rituximab than used in the rituximab EXPLORER trial.

3.59 The ERG reviewed 7‑year data from the belimumab phase II extension study and noted that these new data suggested that there may have been an error in the company's calculation of the discontinuation rate of 13% from the 6‑year data. The ERG stated that, based on the numbers available, and applying a least-squares exponential fit rather than a simple average to the 7‑year data, the annual discontinuation rate for the phase II extension study should be 11.7%.

3.60 The ERG presented pairwise ICERs without incorporating the patient access scheme for belimumab compared with standard care for 3 scenarios that assumed:

  • an annual discontinuation rate of 11.7%

  • lifetime treatment with belimumab

  • an administration cost of belimumab of £154

  • discounting health benefits (that is, QALYs) by 3.5% per annum.

3.61 For a scenario assuming no continuation rule, the ERG estimated an ICER of £68,986 per QALY gained. For a scenario assuming a continuation rule of SELENA-SLEDAI score of greater than or equal to 4, the ERG estimated an ICER of £59,946 per QALY gained. For a scenario assuming a continuation rule of SELENA-SLEDAI score greater than or equal to 6, the ERG estimated an ICER of £52,517 per QALY gained. ICERs with the patient access scheme were provided. These are commercial in confidence because of the confidential nature of the patient access scheme.

Critique by the ERG of the company's further evidence provided after the fourth committee meeting

3.62 The ERG provided a review of the economic analyses provided by the company. It confirmed that the analyses had been conducted correctly (that is, as described by the company), but queried the relevance of the proposed maximum treatment durations. It also presented additional scenario analyses, including the company's base case combined with a lifetime maximum treatment duration and an assumption that patients whose condition does not respond to belimumab revert to the same SELENA-SLEDAI score as those whose condition did not respond to standard care. The results of these exploratory analyses are confidential and cannot be presented here.

NICE Decision Support Unit work after the appeal

3.63 NICE commissioned the DSU to carry out additional work to identify estimates of annual discontinuation rates for people with systemic lupus erythematosus whose disease responds to belimumab treatment (taking into account its marketing authorisation describing continuous use). A survey questionnaire was developed by the DSU and sent to 41 UK lupus experts who had experience of treating people with systemic lupus erythematosus, but not necessarily with belimumab. The survey questionnaire asked the experts to provide estimates of annual discontinuation rates for people whose disease responded to belimumab treatment at 24 weeks based on a continuation rule of a SELENA-SLEDAI score greater than or equal to 4, or greater than or equal to 6.

3.64 Of the 41 experts invited, 14 (34.1%) responded but only 3 (7.3%) completed the survey questionnaire (either in part or in full). The most common reason given for non‑completion was that clinical experience with belimumab is limited because it is not currently recommended by NICE or the Scottish Medicines Consortium. Therefore, the experts found it difficult to provide reliable estimates of long‑term discontinuation rates. All 3 participating experts had experience of treating people with lupus and 2 had experience with belimumab. Two of the 3 experts suggested an initial SELENA-SLEDAI response (that is, a SELENA-SLEDAI score of greater than or equal to 4 or greater than or equal to 6) would lead to different long‑term discontinuation rates and all 3 experts believed discontinuation rates would increase over time. The DSU stated that the results of the survey showed variability in the responses provided by the 3 experts on the expected discontinuation rates with belimumab treatment. The DSU concluded that limited confidence could be placed on the estimates provided in the survey questionnaire compared with those already available from the BLISS and phase II extension studies because of the low completion rate.

NICE DSU work after the fourth committee meeting

3.65 The DSU provided a review of the company's proposed evidence collection using advice in the guide to methods of technology appraisal 2013 and a framework developed by Claxton et al. (2012). It found that the proposal accurately described the BILAG registry and would provide some data that could be used to resolve the committee's uncertainty and might be particularly valuable in collecting data on adherence to the proposed stopping rule (because this is not being used in any other country). It noted that several other ongoing studies described in the company's submission would provide data to address some of the other uncertainties identified by the committee.

3.66 Full details of all the evidence are in the committee papers.

  • National Institute for Health and Care Excellence (NICE)