Alemtuzumab for treating highly active relapsing–remitting multiple sclerosis

The manufacturer provided clinical-effectiveness evidence, identified by systematic review, from:

• 2 phase III randomised controlled clinical trials: CARE‑MS I (n=581, median follow-up of 2 years), and CARE‑MS II (n=1,046, median follow-up of 2 years)

• 1 phase II randomised controlled clinical trial: CAMMS223 (n=334, maximum follow-up of 3 years extended by a follow-up period of 4 years from final alemtuzumab dose)

• 1 extension study: CAMMS03409 (n=1,322, median follow-up of 7.1 years), which enrolled people with active relapsing–remitting multiple sclerosis from the CAMMS223, CARE‑MS I and CARE‑MS II trials. In this study, patients previously randomised to the control group in CAMMS223, CARE‑MS I and CARE‑MS II received alemtuzumab and patients previously randomised to alemtuzumab in CAMMS223, CARE‑MS I and CARE‑MS II received further treatment with alemtuzumab, as needed.
  
  In addition, the manufacturer submitted a meta-analysis of the above-listed trials and a mixed treatment comparison to compare alemtuzumab with other disease-modifying treatments for active relapsing–remitting multiple sclerosis (see sections 3.7 and 3.8).

CARE‑MS I, CARE‑MS II and CAMMS223 compared the effectiveness of 12 mg alemtuzumab (with an additional arm receiving 24 mg per infusion in CAMMS223 only) with subcutaneous interferon beta‑1a (Rebif, small initial doses, gradually increasing to 44 micrograms 3 times weekly). All 3 trials included sites in the UK. All 3 trials specified the number of previous relapses patients must have had before they could enrol. For CAMMS223 this was at least 2 relapses in the previous 2 years. For CARE‑MS I and CARE‑MS II this was at least 2 relapses within the previous 2 years, with at least 1 within the previous year. CARE‑MS I and CAMMS223 included patients with an Expanded Disability Status Scale (EDSS) score between 0 and 3 (in which 0 means no disability and no signs of impairment in any functional system and 3 means unimpaired walking, but either moderate disability in 1 functional system or mild disability in 3 or 4 functional systems). CARE‑MS II included patients with an EDSS score between 0 and 5 (in which 5 means disability severe enough to impair normal daily activities and the person's ability to work a full day without special provisions, but they are still able to walk for 200 metres without aid or rest). All patients in CARE‑MS II had to have previously received disease-modifying treatment with beta interferon or glatiramer acetate for 6 months in the preceding 10 years (the inclusion criteria also specified that more than 1 multiple sclerosis relapse had to have occurred while receiving these treatments), whereas patients in CARE‑MS I and CAMMS223 did not.

The co-primary outcomes of the 3 trials were time to the onset of sustained accumulation of disability (specified as lasting for 6 months for CARE‑MS I and CARE‑MS II) and relapse rate. In the trials, patients were assessed quarterly using the EDSS to determine disability, and were assessed as needed for suspected relapses. Sustained accumulation of disability was defined as an increase lasting for 6 months of at least 1.5 points for people with a baseline EDSS score of 0, or 1.0 point for people with a baseline EDSS score of 1.0 or more. A relapse was defined as new or worsening neurological symptoms attributable to relapsing–remitting multiple sclerosis, lasting at least 48 hours, without fever, after at least 30 days of clinical stability, with an objective change on neurological examination. Data from CAMMS223 were analysed by intention to treat, and adjusted for country and baseline EDSS score, as prespecified in the statistical plan. In CARE‑MS I and CARE‑MS II only patients who had received at least 1 dose of trial medication were included in the analysis (that is, a modified intention-to-treat analysis). In CARE-MS II the analysis was also limited to patients who had followed the trial protocol (excluding patients who had not met all inclusion criteria). The results were adjusted for region.

In CARE‑MS I 8% of people in the alemtuzumab treatment group had disability lasting for 6 months, compared with 11.1% in the Rebif group. There was no statistically significant difference in the rates of disability lasting for 6 months between people taking alemtuzumab and people taking Rebif (hazard ratio [HR] 0.70, 95% confidence interval [CI] 0.4 to 1.23; p=0.22). In CARE‑MS II 12.7% of people in the alemtuzumab treatment group had disability lasting for 6 months, compared with 21.1% in the Rebif group. This corresponded to a statistically significant improvement of 42% with alemtuzumab (HR 0.58, 95% CI 0.38 to 0.87; p=008). In CAMMS223 alemtuzumab statistically significantly reduced the risk of sustained accumulation of disability lasting for 6 months by 75% compared with Rebif (HR 0.25, 95% CI 0.11 to 0.57, p<0.001). A separate extended follow-up study of CAMMS223 showed that over 5 years, alemtuzumab statistically significantly reduced the risk of sustained accumulation of disability lasting for at least 6 months by 69% compared with Rebif (HR 0.31, 95% CI 0.16 to 0.60, p=0.0005).

Alemtuzumab statistically significantly reduced the relapse rate compared with Rebif: by 54.9% in CARE‑MS I (RR [rate ratio] 0.45, 95% CI 0.32 to 0.63, p<0.0001), by 49.4% in CARE‑MS II (RR 0.51, 95% CI 0.39 to 0.65, p<0·0001) and by 69% in CAMMS223 (RR 0.31, 95% CI 0.18 to 0.52, p<0.001). The extended follow-up study of CAMMS223 showed that, over 5 years, alemtuzumab statistically significantly lowered the rate of relapse by 66% compared with Rebif (RR 0.34, 95% CI 0.20 to 0.57, p<0.0001).

The manufacturer presented data from CARE‑MS II and CAMMS223 (and its separate study extension) to compare alemtuzumab with Rebif in a subgroup of people with rapidly evolving severe relapsing–remitting multiple sclerosis (size of subpopulation not available). The manufacturer pooled the results of the 12‑mg and 24‑mg alemtuzumab arms of CAMMS223 because it considered that the results in each arm were sufficiently similar to allow this. The manufacturer stated that the analyses showed that the effectiveness of alemtuzumab compared with Rebif in the rapidly evolving severe relapsing–remitting multiple sclerosis subgroup was comparable to or greater than that seen in the overall trial populations. The reduction of risk in sustained accumulation of disability lasting at least 6 months was 51% in CARE‑MS II (no p value reported) and 65% (p=0.036) in the pooled group of CAMMS223. The analysis also indicated a statistically significant reduction in relapse rates for alemtuzumab compared with Rebif, of 56% (p=0.0018) in the rapidly evolving severe relapsing–remitting multiple sclerosis subgroup of CARE‑MS II and of 81% (p<0.0001) in the pooled dose group of CAMMS223.

The manufacturer presented a mixed treatment comparison that compared alemtuzumab with each of the treatments in the decision problem (Rebif, intramuscular interferon beta‑1a [Avonex], interferon beta‑1b [Betaferon], glatiramer acetate, natalizumab and fingolimod). The manufacturer included 30 clinical trials identified in the systematic literature review, all of which recruited patients from the year 2000 onwards, and in which at least 80% of the patients had relapsing–remitting multiple sclerosis (the 'base-case mixed treatment comparison'). The manufacturer justified the year 2000 as an appropriate cut-off point because annualised relapse rates have fallen in recent years and because the diagnostic criteria used in multiple sclerosis trials have changed. The manufacturer provided a separate 'all years' analysis that, in addition, included trials recruiting patients before the year 2000. The outcomes in the base-case mixed treatment comparison were annualised relapse rate, proportion of patients who were relapse free, sustained accumulation of disability lasting for 3 months, sustained accumulation of disability lasting for 6 months, discontinuation of treatment rate and discontinuation of treatment rate because of adverse events. In the base-case mixed treatment comparison, alemtuzumab led to statistically significantly lower annualised relapse rates than the beta interferons and glatiramer acetate. For the 3‑month sustained accumulation of disability outcome, alemtuzumab was statistically significantly lower than Avonex, Betaferon and Rebif (44 micrograms); however, the difference between alemtuzumab and glatiramer acetate was not statistically significant. For the 6‑month sustained accumulation of disability outcome, alemtuzumab was statistically significantly lower than Rebif (44 micrograms). While the point estimates for alemtuzumab compared with glatiramer acetate favoured alemtuzumab, the difference was not statistically significant. The results of the mixed treatment comparison were considered confidential by the manufacturer and therefore cannot be reported here.

The manufacturer carried out 2 separate mixed treatment comparisons of alemtuzumab for the subgroups of patients with highly active relapsing–remitting multiple sclerosis despite beta interferon treatment (from CARE‑MS II) and rapidly evolving severe relapsing–remitting multiple sclerosis (from CARE‑MS I and II and CAMMS223). For the highly active relapsing–remitting multiple sclerosis despite beta interferon treatment subgroup, alemtuzumab had a lower annualised relapse rate than fingolimod; however, the difference was not statistically significant (HR 0.50, 95% CI 0.11 to 2.29). The 3‑month sustained accumulation of disability was lower with alemtuzumab than with fingolimod but the difference was not statistically significant (HR 0.65, 95% CI 0.11 to 3.72). For the rapidly evolving severe relapsing–remitting multiple sclerosis subgroup, alemtuzumab had a lower annualised relapse rate than natalizumab; however, the difference was not statistically significant (HR 0.69, 95% CI 0.11 to 4.53). The 6‑month sustained accumulation of disability was lower with alemtuzumab than with natalizumab, but the difference was not statistically significant (HR 0.78, 95% CI 0.06 to 10.83).

The manufacturer also presented a naïve indirect comparison of alemtuzumab compared with fingolimod and natalizumab for the subgroups of patients with highly active relapsing–remitting multiple sclerosis despite beta interferon therapy and patients with rapidly evolving severe relapsing–remitting multiple sclerosis respectively. The CARE MS-II study comparing alemtuzumab with active comparator (Rebif [44 micrograms]) showed that alemtuzumab had a greater treatment effect on 3‑month sustained accumulation of disability in people with highly active relapsing–remitting multiple sclerosis despite beta interferon treatment (HR 0.61, 95% CI 0.37 to 1.01) than fingolimod compared with placebo had in the FREEDOM study (HR 0.73, 95% CI 0.29 to 1.84). Studies comparing alemtuzumab with Rebif showed that alemtuzumab had a similar treatment effect on 6‑month accumulation of disability in people with rapidly evolving severe relapsing–remitting multiple sclerosis (CAMMS223 [HR 0.3, 95% CI 0.13 to 0.69], CARE MS-I [HR 0.83, 95% CI 0.28 to 2.42] and CARE MS-II [HR 0.47, 95% CI 0.17 to 1.32]) to natalizumab compared with placebo in the AFFIRM study (HR 0.36, 95% CI 0.17 to 0.76).

In a pooled analysis of CARE‑MS I, CARE‑MS II and CAMMS223 results, most patients reported at least 1 adverse event, the majority of which were mild or moderate in severity. The most common adverse events were headache, rash, fever and multiple sclerosis relapse. The incidence of serious adverse events as reported at the end of the trials from the European Public Assessment Report (EPAR) was 18.3% in both the alemtuzumab and comparator arms. Independent investigators considered that the adverse events were related to alemtuzumab in 7.1% of all patients receiving 12 mg alemtuzumab and to Rebif in 1.6% of all patients receiving Rebif. The most frequently reported serious adverse events in the alemtuzumab 12 mg group were multiple sclerosis relapse (6.1%), pneumonia (0.4%), autoimmune thrombocytopenia (0.4%), gastroenteritis (0.4%), appendicitis (0.4%) and hives (0.4%). Four people developed idiopathic thrombocytopenic purpura. More thyroid-related adverse events were observed in the alemtuzumab arm of the trial (16.6%) than in the Rebif arm (5.2%). Thyroid-related adverse events were observed in 36.2% (at 4 years) and 44.7% (at 8 years) of patients in the alemtuzumab 12 mg/day group. The highest incidence of thyroid-related adverse events was observed between 24 and 42 months after the first treatment cycle. Other serious adverse events observed throughout the clinical trials included infections and renal disease. With the exception of thyroid disorders, administering more than 2 treatment cycles of alemtuzumab did not result in increased frequencies of common adverse events or clinically important events which had not already been observed. Eight people died during the clinical trials; 7 of these people had received alemtuzumab, and the EPAR states that the investigator judged that 3 deaths were possibly or likely to have been related to alemtuzumab treatment.

The manufacturer assessed health-related quality of life during the phase II and III trials using the Short-Form Health Survey (SF‑36), the Functional Assessment of Multiple Sclerosis (FAMS) and the EuroQoL‑5 Dimension-5 Level (EQ‑5D-5L) questionnaire. In CARE‑MS I and II, patients completed the SF‑36 at baseline, at month 12, at month 24, and at early discontinuation of treatment. In CARE‑MS I and II, the FAMS and EQ‑5D-5L were assessed at baseline and every 6 months thereafter until month 24 or early discontinuation of treatment. In CAMMS223, patients completed the SF‑36 every 6 months for 3 years, but not the FAMS or EQ-5D-5L. The manufacturer pooled the EQ-5D-5L utility scores from the CARE MS I and II trials in the alemtuzumab and Rebif (44 micrograms) arms at baseline and 24 months by EDSS score. The difference in mean utility values between patients with the same EDSS scores at baseline and at 24 months showed no consistent trend in either the alemtuzumab or the Rebif arms. The results were provided by the manufacturer as commercial in confidence.

To assess the cost effectiveness of alemtuzumab the manufacturer submitted a multi-state Markov model reflecting the course of multiple sclerosis and the effect of treatment with alemtuzumab or the comparators defined in the decision problem (that is, Rebif, Avonex, Betaferon, glatiramer acetate, natalizumab and fingolimod). The model incorporated health states for the type of multiple sclerosis (relapsing–remitting or secondary progressive) and for disease severity defined by the level of disability (EDSS scores ranging from 0 [normal neurological examination] to 9 [confined to bed]). Patients with active relapsing–remitting multiple sclerosis entered the model at EDSS 0 up to EDSS 7 (an EDSS of 7 and above means patients have lost the ability to walk on their own). EDSS 10 represented death from multiple sclerosis. In each cycle, patients remained in the same state, progressed to a worse state (moving to a better state was not possible), transferred to a state reflecting secondary progressive multiple sclerosis, or died. The model assumed that when a patient progressed from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis, their EDSS score increased by 1 point. The manufacturer chose a cycle length of 1 year, and a lifetime time horizon of 50 years. Patients entering the model had a mean age of 39.3 years, and there were approximately 3 times as many women as men. The analyses used an NHS and personal social services perspective and a 3.5% discount rate on costs and health effects. Most patients received only 2 courses of alemtuzumab, but the model included re-treatment for some patients in year 3, in years 6 to 9 and in year 10 or above (the manufacturer labelled the rates of re-treatment as commercial in confidence and so they cannot be presented here).

To estimate the rate of disease progression in people with relapsing–remitting multiple sclerosis, the manufacturer used a matrix to represent the natural history transition and disability progression in people who were not receiving disease-modifying therapies. The manufacturer chose the London Ontario dataset, a longitudinal observational study from 1989, to populate the natural history transition matrix. Since no data for patients with an EDSS state of 0 were available in this dataset, the manufacturer obtained transition probabilities for an EDSS 0 from the placebo arms of 2 trials (TOWER and TEMSO) that compared teriflunomide with placebo for treating multiple sclerosis. The manufacturer based the population entering the model on the average demographic profile of patients in the UK Risk Sharing Scheme, in which 85.8% have relapsing–remitting multiple sclerosis, the mean EDSS of patients with relapsing–remitting multiple sclerosis is 3.1, and the mean EDSS of patients with secondary progressive multiple sclerosis is 5.5.

To model the effect of treatment with alemtuzumab on relapsing–remitting multiple sclerosis, the manufacturer applied the hazard ratios for the outcome of disability sustained for 3 months compared with placebo from the base-case mixed treatment comparison (see section 3.7) to the natural history matrix. Separately, the manufacturer considered treatment effects on relapse rate and severity (whether or not the relapse leads to hospitalisation). In the base case, the manufacturer assumed that patients discontinue treatment when they convert from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis, or progress to EDSS 7. After discontinuing treatment, patients were assumed to receive best supportive care only. The manufacturer's model assumed that no patient who received alemtuzumab ever discontinued treatment, while patients could discontinue comparator treatments (and subsequently receive best supportive care). The manufacturer also assumed that the treatment effect of alemtuzumab did not change over time (even during years when patients did not receive alemtuzumab) until a patient reached EDSS 7 or converted to secondary progressive multiple sclerosis. On entering EDSS 7 the benefits of alemtuzumab stopped, independent of the number of courses of alemtuzumab given. In each cycle patients could stop using comparator treatments, discontinue treatment after reaching EDSS 7, or experience relapse or adverse events. The probability of death was dependent on the EDSS state (the higher the EDSS score, the higher the risk of death), age and sex.

The manufacturer's model applied health state utility values to each of the EDSS states. Although the manufacturer collected EQ‑5D data in the CARE‑MS I and II trials, it did not use these data in the model as they were not available at the time of submission. Instead, the manufacturer obtained health state utility values from Orme et al. (2007), a UK survey of health-related quality of life in (EQ‑5D) in people with multiple sclerosis. Utility values decreased as EDSS scores increased, with the exception of the utility value for EDSS state 3, which was lower than EDSS 4. EDSS states 8 and 9 had negative utility values, indicating states that are considered to be worse than being dead. The manufacturer applied disutilities for a relapse, to caregivers, and for adverse events. The manufacturer obtained the value for the disutility of relapse from Orme et al. (2007), and the value for the disutility of relapse leading to hospitalisation from a US study (Prosser et al. 2003). To estimate disutility to caregivers, the manufacturer used values taken from Gani et al. (2008), and to estimate the time spent caring for the patient, the manufacturer used Orme et al. (2007). Disutility values applied for each adverse event were annualised based on the published literature. The manufacturer also took into account how long each adverse event lasted, and whether it was specific to treatment. The adverse events included infusion-associated reactions, bronchitis, herpes zoster, urinary tract infections, autoimmune thyroid-related adverse events, nephropathies, idiopathic thrombocytopenic purpura, other cytopenias and vomiting.

The model used NHS reference costs and the payment-by-results tariff to estimate the costs of administration, monitoring and adverse events associated with each treatment. The manufacturer assumed that monitoring of patients previously treated with alemtuzumab lasts for up to 12 years. The manufacturer derived some costs from the literature: health state costs (including direct medical costs and direct non-medical costs) from a UK study (Tyas et al. 2007), and the costs associated with relapse from a study from the Republic of Ireland (Dee et al. 2012). For a sensitivity analysis, the manufacturer used an alternative UK study (Karampampa et al. 2012) to derive health state costs, although the manufacturer provided only natural history costs aggregated for EDSS states 0 to 3, 4 to 6 and 7 to 9, rather than costs for individual EDSS states. The manufacturer validated the resource use and costs it applied in the model using clinical experts. The cost of one of the comparators, fingolimod, includes a simple discount patient access scheme agreed with the Department of Health. However, the manufacturer did not know how large the discount was, and therefore could not use it in its base-case analysis. Instead, the manufacturer explored different prices of fingolimod in sensitivity analyses, using a range of assumed discounts.

The manufacturer's submission presented the total life years gained, the total quality-adjusted life years (QALYs) and the total costs resulting from the economic model for alemtuzumab and Rebif (44 micrograms). Treatment with alemtuzumab was associated with 18.62 life years, which equated to 4.03 QALYs, at a total cost of £499,347. Treatment with Rebif (44 micrograms) was associated with 18.38 life years, which equated to 2.85 QALYs, at a total cost of £489,354.

The manufacturer conducted a fully incremental analysis, calculating the incremental QALY gains and costs for all treatment options and ordered by increasing costs. The treatments included alemtuzumab, glatiramer acetate, Rebif (22 micrograms), Rebif (44 micrograms), Avonex, and Betaferon. The manufacturer also included fingolimod and natalizumab in its incremental analysis, although it acknowledged that these drugs have marketing authorisations only for use in highly active relapsing–remitting multiple sclerosis despite beta interferon treatment and rapidly evolving severe relapsing–remitting multiple sclerosis. When compared in this incremental analysis, the probabilistic estimates of the incremental cost-effectiveness ratios (ICERs) suggested that:

• Alemtuzumab dominated Betaferon, fingolimod (without applying a patient access scheme discount), fingolimod (assuming a patient access scheme price of £13,000 per year), and natalizumab. (A treatment dominates other treatments when it is less expensive and more effective.)

• Rebif (44 micrograms) and Rebif (22 micrograms) were extendedly dominated by alemtuzumab. (A treatment is extendedly dominated when its ICER is higher than that of the next, more effective, option when compared with a common baseline.)

• The ICER for alemtuzumab compared with glatiramer acetate was £7,017 per QALY gained. The manufacturer's deterministic results were similar with an ICER of £8,924 per QALY gained for alemtuzumab compared with glatiramer acetate.

Using the results of the subgroup mixed treatment comparisons (see section 3.8), the manufacturer compared alemtuzumab with fingolimod and with natalizumab for the highly active relapsing–remitting multiple sclerosis despite beta interferon treatment and the rapidly evolving severe relapsing–remitting multiple sclerosis subgroups, respectively. For both analyses, alemtuzumab dominated the respective comparator.

The manufacturer conducted one‑way sensitivity analyses, which showed that the cost effectiveness of alemtuzumab was most sensitive to the hazard ratios reflecting the comparative effectiveness of alemtuzumab compared with placebo for sustained disability progression, disease costs, and the discontinuation rate of Rebif (44 micrograms). Alemtuzumab continued to dominate all comparators except glatiramer acetate, except when the manufacturer varied the hazard ratios for disability progression. When the manufacturer applied the upper limit of the 95% confidence interval around the sustained accumulation of disability hazard ratio for alemtuzumab from the manufacturer's mixed treatment comparison, the resulting ICER for alemtuzumab compared with Rebif (44 micrograms) was £1,200,973 per QALY gained. With the lower limit of the 95% confidence interval, alemtuzumab dominated Rebif (that is, had the lowest total treatment costs for the greatest clinical gain of all treatments in the analysis).

The manufacturer also tested how sensitive the results were to which mixed treatment comparison it used, by using the 'all years' data instead of the 'base-case' mixed treatment comparison and by only including trials in which 100% of patients had relapsing–remitting multiple sclerosis (rather than the base-case mixed treatment comparison, in which trials with at least 80% of patients with relapsing–remitting multiple sclerosis were included). When trials from 'all years' in which at least 80% of patients had relapsing–remitting multiple sclerosis were included, the deterministic ICER for alemtuzumab compared with glatiramer acetate increased from £8,924 to £9,982 per QALY gained. When the manufacturer included trials from all years in which the percentage of the population with relapsing–remitting multiple sclerosis was 100% the ICER for alemtuzumab compared with glatiramer acetate increased to £27,434 per QALY gained. When the manufacturer used the mixed treatment comparison including trials after the year 2000 in which 100% of patients had relapsing–remitting multiple sclerosis, the ICER for alemtuzumab compared with glatiramer acetate was £10,822 per QALY gained.

The manufacturer conducted a number of scenario analyses using Rebif (44 micrograms) as the comparator, but not glatiramer acetate, with the justification that Rebif (44 micrograms) was the standard treatment for active relapsing–remitting multiple sclerosis. In the best-case scenario alemtuzumab dominated Rebif and in the worst case scenario the ICER for alemtuzumab compared with Rebif was £20,388 per QALY gained. The manufacturer developed other scenarios based on:

• sourcing the baseline characteristics from the CARE‑MS trials rather than from the UK Risk Sharing Scheme (the ICER for alemtuzumab compared with Rebif was £869 per QALY gained)

• using costs related to the natural history of multiple sclerosis from Karampampa et al. (2012) rather than Tyas et al. (2007); alemtuzumab dominated Rebif

• using natural history transition probabilities assuming that the population only included people with active relapsing–remitting multiple sclerosis, instead of all people with relapsing–remitting multiple sclerosis (the ICER for alemtuzumab compared with Rebif was £8,597 per QALY gained)

• assuming long-term waning of treatment effect by 25% or 50% after year 5 for all treatments, instead of assuming that the beneficial effect of alemtuzumab does not wane (the ICERs for alemtuzumab compared with Rebif were £13,956 and £20,388 per QALY gained, respectively)

• assuming that treatment with alemtuzumab does not influence the probability of relapses or hospitalisation (the ICER for alemtuzumab compared with Rebif was £14,517 per QALY gained)

• using the trial data (pooled CARE‑MS I and CARE‑MS II) for the transition probabilities instead of using values sourced from the literature (alemtuzumab dominated Rebif).

The ERG reviewed the manufacturer's model and economic systematic review. The ERG commented that the structure of the economic model was appropriate for multiple sclerosis and consistent with previous economic evaluations of treatments for multiple sclerosis, and that the methods of analysis were appropriate and conformed to NICE methodological guidelines.

The ERG stated that the manufacturer systematically reviewed the literature to populate its transition matrix and reflect the natural history for disability progression for patients not receiving a disease-modifying treatment. The ERG did not find any data more appropriate than the London Ontario data identified by the manufacturer, but commented that the manufacturer did not fully explore the uncertainty around the natural history of multiple sclerosis. In light of previous technology appraisals, the ERG suggested that it would have been more appropriate to explore alternative sources of data.

The ERG evaluated the results of the economic model outputs as compared with published literature. The ERG noted that the manufacturer compared the results at the end of year 2, but no further. As there was no validation beyond 2 years, uncertainty remains as to the validity of longer-term outcomes.

The ERG stated that the manufacturer had performed appropriate structural sensitivity analyses, but had not conducted a sensitivity analysis that varied the rate of disease progression for patients receiving best supportive care only, or the rate of progression from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis.

The ERG identified weaknesses and uncertainty in the manufacturer's economic analysis. The ERG stated that basing the starting model population on the UK Risk Sharing Scheme instead of the clinical trial populations introduced uncertainty into the model, because these populations did not have the same baseline characteristics, particularly with regard to the distribution of initial EDSS states. The ERG commented that the conversion rate used for patients moving from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis in the model was too high, because it did not reflect the people receiving first-line treatment for relapsing–remitting multiple sclerosis. The ERG also stated that the London Ontario estimates for disease progression for patients not taking disease-modifying treatments did not allow EDSS scores to improve. Trial-based transition probabilities were available that allowed EDSS scores to improve, although the ERG commented that using the trial data could pose problems as it reflected a short period of time. The ERG explored the impact of changing these assumptions in their exploratory analyses.

The ERG presented a 'preferred' base case that included alternative characteristics for the patient population, and a different progression rate from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis. The ERG also conducted a series of sensitivity analyses to test uncertainties.

In all its exploratory analyses, the ERG compared alemtuzumab with Rebif (44 micrograms; instead of glatiramer acetate as used in the manufacturer's fully incremental analysis). The ERG made this change because Rebif (44 micrograms) was the direct comparator in the clinical trials and was the most efficacious comparator in the manufacturer's mixed treatment comparison. Using the baseline characteristics for the populations in CARE‑MS I and CARE‑MS II, the ERG calculated that the ICER for alemtuzumab compared with Rebif (44 micrograms) would decrease from £8,445 (manufacturer's base case comparing alemtuzumab with Rebif (44 micrograms) to £2,869 per QALY gained. The ERG also applied a conversion rate of 15 years from relapsing–remitting multiple sclerosis to secondary progressive multiple sclerosis (instead of the 10 to 11 years used by the manufacturer), as used in NICE's technology appraisal guidance on teriflunomide for treating active relapsing–remitting multiple sclerosis. This had the effect of reducing the ICER to £3,100 per QALY gained for alemtuzumab compared with Rebif (44 micrograms). The ERG's preferred approach combining these 2 changes resulted in alemtuzumab dominating (being less costly and more effective than) Rebif (44 micrograms), with a cost saving of £852 per QALY gained.

The ERG tested its preferred base case for alemtuzumab compared with Rebif (44 micrograms) in sensitivity analyses, including:

• reducing by 50% the transition probabilities to more severe health states from the London Ontario dataset (alemtuzumab dominated Rebif [44 micrograms])

• using quality-of-life utility values (upper and lower confidence intervals from the Orme et al. 2007 data used in the manufacturer's model; for both, alemtuzumab dominated Rebif [44 micrograms])

• using disease health state costs from Karampampa et al. (2012) and Biogen et al. (2007); alemtuzumab dominated Rebif (44 micrograms) for Karampampa et al.; for Biogen et al., the ICER for alemtuzumab compared with Rebif (44 micrograms) was £4,654 per QALY gained

• reducing the cost of a relapse that results in hospitalisation from £6,146 to £3,039 (the ICER for alemtuzumab compared with Rebif [44 micrograms] was £1,013 per QALY gained)

• applying a waning of treatment effect for alemtuzumab of 75% for year 10 and beyond, or 75% from year 6 to year 9 and 50% from year 10 and beyond (the ICERs for alemtuzumab compared with Rebif [44 micrograms] were £1,815 and £7,319 per QALY gained, respectively)

• varying the proportion of patients receiving additional alemtuzumab treatment at year 3 (60%) and years 5 and beyond (the ICER for alemtuzumab compared with Rebif [44 micrograms] was £8,336 per QALY gained)

• applying the results from the 'all years' mixed treatment comparison (alemtuzumab dominated Rebif [44 micrograms])

• using the outcome of sustained accumulation of disability lasting for 6 months from the mixed treatment comparison (instead of 3 months) to calculate the disease transition probabilities (alemtuzumab dominated Rebif [44 micrograms]).

The ERG also explored the cost effectiveness of alemtuzumab for the treatment‑naive and treatment‑experienced subgroups separately, using the ERG preferred base case, the relative risk for annualised rate of relapse, and a sustained accumulation of disability lasting 3 months for alemtuzumab. Using the treatment-naive group data from CARE‑MS I, the ERG's preferred base case (that is, where alemtuzumab dominated Rebif [44 micrograms], see section 3.29) changed to an ICER of £6,392 per QALY gained for alemtuzumab compared with Rebif (44 micrograms). When the ERG used the CAMMS223 data, alemtuzumab dominated Rebif (44 micrograms). Alemtuzumab also dominated Rebif (44 micrograms) when the ERG pooled data from the 2 trials. For the treatment-experienced group, using effectiveness data from CARE‑MS II, the ICER was £2,854 per QALY gained for alemtuzumab compared with Rebif (44 micrograms).

The ERG also carried out exploratory analyses for the subgroup with highly active relapsing–remitting multiple sclerosis despite beta interferon treatment, and the subgroup with rapidly evolving severe relapsing–remitting multiple sclerosis. In these analyses, the ERG used its preferred base case for a slower progression to secondary progressive multiple sclerosis for the rapidly evolving severe relapsing–remitting multiple sclerosis subgroup, and different patient characteristics for the highly active relapsing–remitting multiple sclerosis despite beta interferon treatment subgroup. These changes had only minimal effect on the model results, and alemtuzumab continued to dominate fingolimod and natalizumab.

The manufacturer provided a revised base-case analysis using the Committee's preferred assumptions, as requested in the appraisal consultation document which did all of the following:

• used sustained accumulation of disability lasting 6 months as the primary outcome measure of the mixed treatment comparison

• used the 'all years' mixed treatment comparison adjusted for baseline relapse rates to estimate disease progression and withdrawal rates

• used the intention-to-treat analyses developed for the CAMMS223, CARE‑MS I and CARE‑MS II trials adjusted for baseline Expanded Disability Status Scale (EDSS) only (unadjusted for country or region)

• used the EQ-5D-5L utility scores pooled from the CARE MS I and II trials comparing alemtuzumab with Rebif (44 micrograms)

• used data on the natural history and progression of disability from the placebo arms of the TOWER and TEMSO trials to allow for improvements in patients' EDSS states

• incorporated the deaths observed in the trials into the model

• assumed that the efficacy for alemtuzumab began waning at 3 or 5 years

• used additional costs of other licensed treatments for active relapsing–remitting multiple sclerosis after failure of alemtuzumab

• used a time-dependent rate of re-treatment for the costs of alemtuzumab

• removed the mid-cycle correction for the costs of alemtuzumab

• increased the number of monitoring and neurology visits for patients treated with alemtuzumab as well as visits for monitoring after restarting alemtuzumab treatment

• used the lower health state costs used in the ERG's analyses

• used costs associated with adverse effects of treatment including renal failure, renal transplantation, dialysis and death

• used baseline characteristics from the alemtuzumab trials rather than from the UK Risk Sharing Scheme to populate the economic model.
  
  The manufacturer applied the Committee's preferences in individual analyses (see section 3.34) and also combined them into one analysis (see section 3.35).

In the manufacturer's individual analyses using the baseline characteristics of patients in the alemtuzumab trials in the model instead of the UK Risk Sharing Scheme, alemtuzumab dominated glatiramer acetate. For each of the other individual analyses, the resulting probabilistic ICERs for alemtuzumab compared with glatiramer acetate remained below £20,000 per QALY gained with the exception of the analyses exploring the impact of waning effectiveness of alemtuzumab and its comparators. For these analyses, the manufacturer presented 2 scenarios; the first assumed a decreasing efficacy for both alemtuzumab and the comparators over time, and the second assumed decreasing efficacy only for alemtuzumab. When the manufacturer assumed that the treatment effectiveness for both alemtuzumab and its comparators was reduced from 100% to 75% from year 3 to year 5 after treatment, and then to 50% from year 6 onward, the manufacturer's incremental analyses showed that glatiramer acetate dominated Avonex, Betaferon and Rebif (44 micrograms) and that the ICER for alemtuzumab compared with glatiramer acetate was £23,432 per QALY gained. When the manufacturer assumed that the effectiveness of alemtuzumab was reduced from 100% to 75% from year 3 to year 5, followed by a reduction to 50% from year 6 onward (while the efficacy of alemtuzumab's comparators remained unchanged) glatiramer acetate dominated Avonex, Betaferon and Rebif (44 micrograms) and the ICER for alemtuzumab compared with glatiramer acetate was £30,657 per QALY gained.

The manufacturer presented a fully incremental analysis combining each of the Committee's preferred assumptions including the 2 scenarios in which the effectiveness of treatments wanes over time. When the manufacturer assumed that the effectiveness for both alemtuzumab and its comparators was reduced from 100% to 75% from year 3 to year 5, followed by a reduction to 50% efficacy from year 6 onward, the manufacturer's incremental analyses showed that glatiramer acetate dominated Avonex, Betaferon and Rebif (44 micrograms) and that the ICER for alemtuzumab compared with glatiramer acetate was £13,636 per QALY gained. When the manufacturer assumed that the treatment effectiveness for alemtuzumab was reduced from 100% to 75% from year 3 to year 5, followed by a reduction to 50% efficacy from year 6 onward (while the efficacy of its comparators remained unchanged at 100%) glatiramer dominated Avonex, Betaferon and Rebif (44 micrograms) and the ICER for alemtuzumab compared with glatiramer acetate was £24,472 per QALY gained.

For the subgroup of rapidly evolving severe relapsing–remitting multiple sclerosis, when comparing alemtuzumab with natalizumab, the manufacturer used the Committee's preferred assumptions (see section 3.33) with the exception of using the results of the 'all years' mixed treatment comparison adjusted for baseline relapse rate because the manufacturer had not identified a relapse rate for natalizumab. When the manufacturer applied the Committee's assumptions individually, alemtuzumab dominated natalizumab for all but 1 scenario. In that scenario, the manufacturer assumed that the treatment effect for alemtuzumab waned beyond 3 years after treatment with alemtuzumab, while assuming that the treatment effect for natalizumab remained constant over the lifetime of the model. The ICER for this scenario was £236,172 per QALY gained. When the manufacturer combined all the Committee's preferred assumptions, alemtuzumab dominated natalizumab in the subgroup of people with rapidly evolving severe relapsing–remitting multiple sclerosis.

For the subgroup of patients with relapsing–remitting multiple sclerosis with high disease activity despite beta interferon treatment, when comparing alemtuzumab with fingolimod, the manufacturer used the Committee's preferred assumptions (see section 3.33) with the exception of 2 assumptions for which the manufacturer did not identify data. These 2 assumptions were sustained accumulation of disability lasting 6 months as a primary outcome measure in the mixed treatment comparison, and using the results of the 'all years' mixed treatment comparison adjusted for baseline relapse rate. When the manufacturer applied the Committee's assumptions individually, alemtuzumab dominated fingolimod in all scenarios. When the manufacturer combined all the Committee's assumptions, alemtuzumab continued to dominate fingolimod.

For the same subgroup, that is, patients with relapsing–remitting multiple sclerosis with high disease activity despite beta interferon treatment, and when comparing alemtuzumab with fingolimod, the manufacturer explored additional scenarios. The manufacturer combined all of the Committee's preferred assumptions (see section 3.33) and:

• assumed that the hazard ratios from patients with relapsing–remitting multiple sclerosis with high disease activity despite beta interferon treatment in CARE MS II, which reflected the effectiveness of alemtuzumab compared with Rebif (44 micrograms) to delay disability (sustained accumulation of disability at 3 months) and annual relapse rates, were equivalent to what would have been expected had alemtuzumab been compared with placebo

• incorporated these assumptions together with the hazard ratios from the subgroup of patients with relapsing–remitting multiple sclerosis with high disease activity despite beta interferon treatment in the FREEDOMS trial which compared fingolimod with placebo

• assumed that the patient access scheme price for fingolimod (the details of which were not available to the manufacturer of alemtuzumab) was £13,000 and

• applied either the utility values from CARE-MS I and II or those from the placebo arms of the TEMSO study (teriflunomide versus placebo) combined with the utility values reflecting relapses from Orme et al.
  
  The ICERs resulting from these analyses for alemtuzumab compared with fingolimod in patients with relapsing–remitting multiple sclerosis with high disease activity despite beta interferon treatment were £7,089 per QALY gained using the CARE-MS I and II trial utility results and £17,232 per QALY gained using the utility results from the placebo arms of the TEMSO study combined with the Orme study utility decrements.

The ERG confirmed that the additional evidence presented by the manufacturer reflected the Committee's requests for additional analyses. The ERG confirmed that it could calculate the manufacturer's deterministic ICERs both in the individual analyses and in the Committee's preferred combined analysis but, owing to time constraints, it could only verify a sample of the probabilistic results presented by the manufacturer. The ERG noted that the manufacturer had incorrectly estimated the ICERs in its fully incremental analyses because the manufacturer compared the treatment with the next less costly treatment, even when the next less costly treatment was dominated.

The ERG reviewed the manufacturer's response to the ACD which focused on both subgroups reflecting patients with high disease activity. The ERG noted that while the manufacturer's mixed treatment comparison provided evidence of the effectiveness of alemtuzumab in these subgroups, there remained a number of uncertainties with these data: the evidence network depended on the teriflunomide trials which included either the overall relapsing–remitting multiple sclerosis population (TENERE) or a subgroup of previously treated patients as a proxy for highly active relapsing–remitting multiple sclerosis despite beta interferon treatment (TEMSO and TOWER); inconsistencies in the definitions of the subgroups in each of the trials; differences between the patient populations included in the trials; and that the mixed treatment comparison was heavily dependent on indirect evidence to complete the evidence network.

The ERG conducted exploratory analyses that assumed that alemtuzumab and fingolimod were equally effective, and that alemtuzumab and natalizumab were equally effective. To do this, the ERG applied the hazard ratios for annual relapse rates and sustained accumulation of disability from the mixed treatment comparison for alemtuzumab compared with placebo and fingolimod and natalizumab each compared with placebo. The ERG applied hazard ratios for alemtuzumab to fingolimod and natalizumab (and vice versa) in the respective subgroups, and also applied the midpoint hazard ratio between alemtuzumab and either fingolimod or natalizumab. In the subgroup of high disease activity despite beta interferon treatment when comparing alemtuzumab with fingolimod, the resulting ICERs were £4,460, £14,788 and £8,942 per QALY gained, respectively. In the subgroup of rapidly evolving severe relapsing–remitting multiple sclerosis, alemtuzumab dominated natalizumab in each scenario.