Alirocumab for treating primary hypercholesterolaemia and mixed dyslipidaemia

3.2 ODYSSEY HIGH FH was a randomised, double‑blind study in 107 people with heterozygous‑familial hypercholesterolaemia whose LDL‑C levels were not adequately controlled with a maximally tolerated, stable, daily dose of statin with or without other LMT. Patients were randomised in a 2:1 ratio to either alirocumab 150 mg or placebo. The difference in mean percent change from baseline in LDL‑C level at 24 weeks was −39.1% (p<0.0001) with alirocumab compared with placebo.

3.3 ODYSSEY FH I was a randomised, double‑blind, study in 486 people with heterozygous‑familial hypercholesterolaemia whose LDL‑C levels were not adequately controlled with a maximally tolerated, stable, daily dose of statin with or without other LMT. Patients were randomised in a 2:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) or placebo. The difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −57.9% (p<0.0001) with alirocumab compared with placebo.

3.4 ODYSSEY FH II was a randomised, double‑blind study in 249 people with heterozygous‑familial hypercholesterolaemia whose LDL‑C levels were not adequately controlled with a maximally tolerated, stable, daily dose of statin with or without other LMT. Patients were randomised in a 2:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) or placebo. The difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −51.4% (p<0.0001) with alirocumab compared with placebo.

3.5 ODYSSEY COMBO I was a randomised, double‑blind study in 316 people with hypercholesterolaemia and established coronary heart disease or coronary heart disease risk equivalents whose LDL‑C levels were not adequately controlled with a maximally tolerated daily dose of statin with or without other LMT. Patients were randomised in a 2:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) or placebo. The difference in mean percent change from baseline in LDL‑C level at 24 weeks was −45.9% (p<0.0001) with alirocumab compared with placebo.

3.6 ODYSSEY COMBO II was a randomised, double‑blind, ezetimibe‑controlled, double‑dummy study in 720 people with hypercholesterolaemia and established coronary heart disease or coronary heart disease risk equivalents whose LDL‑C levels were not adequately controlled with a maximally tolerated daily dose of statin. Patients were randomised in a 2:1 ratio to either alirocumab (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) or ezetimibe 10 mg. The difference in mean percent change from baseline in LDL‑C level at 24 weeks was −29.8% (p<0.0001) with alirocumab compared with ezetimibe.

3.7 ODYSSEY LONG‑TERM was a randomised, double‑blind study in 2,341 people with non‑familial hypercholesterolaemia or established coronary heart disease/coronary heart disease risk equivalent, or people with heterozygous‑familial hypercholesterolaemia with or without coronary heart disease/coronary heart disease risk equivalents whose LDL‑C levels were not adequately controlled with a maximally tolerated daily dose of statin with or without other LMT. Patients were randomised in a 2:1 ratio to either alirocumab 150 mg or placebo. The difference in mean percent change from baseline in LDL‑C level at 24 weeks was −61.9% (p<0.0001) with alirocumab compared with placebo.

3.8 ODYSSEY OPTIONS I was a randomised, double‑blind study in 355 people with non‑familial hypercholesterolaemia or heterozygous‑familial hypercholesterolaemia and a history of coronary heart disease, risk of cardiovascular disease or diabetes with target organ damage, whose LDL‑C levels were not adequately controlled with atorvastatin 20 mg to 40 mg. Patients on a baseline regimen of atorvastatin 20 mg were randomised in a 1:1:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) with atorvastatin 20 mg, atorvastatin 40 mg, or atorvastatin 20 mg with ezetimibe 10 mg. Patients on a baseline regimen of atorvastatin 40 mg were randomised in a 1:1:1:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) with atorvastatin 40 mg, atorvastatin 80 mg, atorvastatin 40 mg with ezetimibe 10 mg, or rosuvastatin 40 mg. For patients having atorvastatin 20 mg, the difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −39.1% (p<0.0001) with alirocumab and statin (atorvastatin 20 mg) compared with statin (atorvastatin 40 mg) alone. The difference in mean percent change from baseline in LDL‑C level was −23.6% (p<0.0001) with alirocumab with statin (atorvastatin 20 mg) compared with ezetimibe with statin (atorvastatin 20 mg). For patients on atorvastatin 40 mg at baseline, the difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −49.2% (p<0.0001) with alirocumab and statin (atorvastatin 40 mg) compared with statin (atorvastatin 80 mg) alone. The difference in mean percent change from baseline in LDL‑C level was −32.6% (p<0.0001) with alirocumab and statin (atorvastatin 40 mg) compared with statin alone (rosuvastatin 40 mg). The difference in mean percent change from baseline in LDL‑C level was −31.4% (p<0.0001) with alirocumab with statin (atorvastatin 40 mg) compared with ezetimibe with statin (atorvastatin 40 mg).

3.9 ODYSSEY OPTIONS II was a randomised, double‑blind study in 305 people with non‑familial hypercholesterolaemia or heterozygous‑familial hypercholesterolaemia and a history of coronary heart disease, risk of cardiovascular disease, or diabetes with target organ damage whose LDL‑C levels were not adequately controlled with rosuvastatin 10 mg to 20 mg. Patients on a baseline regimen of rosuvastatin 10 mg were randomised in a 1:1:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) with rosuvastatin 10 mg, rosuvastatin 20 mg, or rosuvastatin 10 mg with ezetimibe 10 mg. Patients on a baseline regimen of rosuvastatin 20 mg were randomised in a 1:1:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) with rosuvastatin 20 mg, rosuvastatin 40 mg, or rosuvastatin 20 mg with ezetimibe 10 mg. For patients having rosuvastatin 10 mg at baseline, the difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −34.2% (p<0.0001) with alirocumab and statin (rosuvastatin 10 mg) compared with statin (rosuvastatin 20 mg) alone. The difference in mean percent change from baseline in LDL‑C level (with possible up‑titration) was −36.2% (p<0.0001) with alirocumab and statin (rosuvastatin 10 mg) compared with ezetimibe and statin (rosuvastatin 10 mg). For patients on rosuvastatin 20 mg at baseline, the difference in mean percent change from baseline in LDL‑C level at 24 weeks (with possible up‑titration) was −20.3% (p=0.0453) with alirocumab and statin (rosuvastatin 20 mg) compared with statin (rosuvastatin 40 mg) alone. The difference in mean percent change from baseline in LDL‑C level was −25.3% (p=0.0136) with alirocumab with statin (rosuvastatin 20 mg) compared with ezetimibe with statin (rosuvastatin 20 mg).

3.10 ODYSSEY ALTERNATIVE was a randomised, double‑blind, ezetimibe‑controlled, double‑dummy study in 361 people with people with non‑familial hypercholesterolaemia or heterozygous‑familial hypercholesterolaemia with a moderate, high or very high cardiovascular risk and a history of intolerance to statin. Patients were randomised in a 2:2:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels), ezetimibe 10 mg or atorvastatin 20 mg. The difference in mean percent change from baseline in LDL‑C level was −30.4% (p<0.0001) with alirocumab compared with ezetimibe.

3.11 ODYSSEY MONO was a randomised, ezetimibe‑controlled, double‑blind study in 103 people with hypercholesterolaemia with a moderate cardiovascular risk. Patients were randomised in a 1:1 ratio to either alirocumab 75 mg (with up‑titration to alirocumab 150 mg at 12 weeks based on LDL‑C levels) or ezetimibe 10 mg. The difference in mean percent change from baseline in LDL‑C level at week 24 (with possible up‑titration) was −31.6% (p<0.0001) with alirocumab compared with ezetimibe.

3.12 The company undertook meta‑analyses of individual patient data for the mean percent change from baseline in calculated LDL‑C levels (on‑treatment) using a fixed‑effects model. In these analyses, alirocumab (with or without statins) was compared with a statin or ezetimibe (with or without statin). The meta‑analyses showed:

3.13 The company also provided information from an independent meta‑analysis of PCSK9 inhibitors (Navarese et al.). This showed a difference in mean percent change from baseline in LDL‑C level of −47.49% (95% confidence interval [CI] −69.64 to −25.35) and reduced all‑cause mortality and cardiovascular mortality with PCSK9 antibodies compared with control. The company stated that a large randomised controlled trial exploring the occurrence of cardiovascular events of alirocumab compared with placebo is expected to report in 2018.

3.14 The company provided safety information based on combined phase II and phase III studies. The company stated that the rate of treatment‑emergent adverse events (including serious adverse events) was similar between the alirocumab and control arms. It stated that there was no difference in the safety profile observed between alirocumab 75 mg and 150 mg. It also stated that discontinuation due to general allergic adverse events was infrequent but occurred in a higher percentage of the people having alirocumab.

3.15 The company estimated the risk of major adverse cardiovascular events (death from coronary heart disease, non‑fatal myocardial infarction, fatal or non‑fatal ischaemic stroke, or unstable angina requiring hospitalisation) by pooling phase III ODYSSEY trial data. The analysis showed a lower risk of a major adverse cardiovascular event with alirocumab compared with control (hazard ratio [HR] 0.81; 95% CI 0.52 to 1.25), although this was not statistically significant. A post‑hoc analysis from LONG‑TERM also showed a lower risk of major adverse cardiac events with alirocumab compared with placebo (HR 0.52; 95% CI 0.31 to 0.90).

3.16 The ERG noted that evolocumab was not considered to be a relevant comparator by the company because it was still under assessment by NICE. It noted that there were no head‑to‑head trials of alirocumab compared with evolocumab.

3.17 The ERG stated that although it had identified missing terms in the company's search strategy which may have affected its overall sensitivity, it generally considered the company's searches to be fit for purpose.

3.18 The ERG noted that the LDL‑C reduction with alirocumab compared with control was rapid and persistent throughout follow‑up.

3.20 The company submitted a Markov model based on the modelling approaches developed for NICE guidelines on lipid modification and familial hypercholesterolaemia, and technology appraisal guidance on ezetimibe for the treatment of primary (heterozygous-familial and non-familial) hypercholesterolaemia, ticagrelor for the treatment of acute coronary syndromes and rivaroxaban for preventing adverse outcomes after acute management of acute coronary syndrome. The cycle length was 1 year and a half cycle correction was applied. An annual discount rate of 3.5% was applied to costs and health effects. The model had a lifetime time horizon and was conducted from a NHS and personal social services perspective.

3.21 The company's model consisted of 12 mutually exclusive health states:

3.22 The baseline characteristics (age, sex, percentage of patients with diabetes and minimum LDL‑C level) for each population were informed by UK data from the Health Improvement Network (THIN) database, patient characteristics from ODYSSEY trials and the UK National Familial Hypercholesterolaemia audit.

3.23 The baseline probabilities of cardiovascular death in all post‑acute coronary syndrome and post‑ischaemic stroke health states were adjusted to account for the higher risk of future events associated with recurrence of cardiovascular events.

3.24 Alirocumab was given in line with its marketing authorisation. The patient population was modelled according to severity of hypercholesterolaemia (by baseline LDL‑C levels) before starting treatment. Baseline cardiovascular risk (calculated using THIN data) was adjusted by LDL‑C level using a log‑linear relationship between the absolute LDL‑C observed in statin studies and cardiovascular events using the Cholesterol Treatment Trialists' Collaboration (CTTC) meta‑analysis of statins. The company used the difference in mean percent change of alirocumab compared with alternatives based on estimates from specific clinical trials and meta‑analyses. The model assumed that the relative reduction in LDL‑C for alirocumab was constant across all subgroups.

3.25 In the absence of cardiovascular events data from the clinical trials for alirocumab, the company used LDL‑C reduction as a surrogate to link to cardiovascular events. In its base‑case analysis, the company chose the Navarese meta‑analysis of 24 randomised controlled trials (n=10,159) to provide the rate at which the risk of a cardiovascular event declines with a reduction in LDL‑C levels. This was because it preferred estimates from PCSK9 inhibitor studies rather than estimates from statin studies (such as CTTC), because they better reflected the population who will have alirocumab. By assuming a log‑linear relationship between LDL‑C levels and cardiovascular events, the company estimated the risk reduction for cardiovascular mortality as rate ratios (RRs): 0.64 per 1.0 mmol/l reduction in LDL‑C rate (95% CI 0.40 to 1.04) and 0.64 for myocardial infarction (95% CI 0.43 to 0.96). The risk reduction for coronary revascularisation and ischaemic stroke was assumed to be the same as other non‑fatal cardiovascular events.

3.26 Transition probabilities were based on Kaplan–Meier analyses from an observational retrospective cohort analysis using the THIN database of people with established cardiovascular disease, diabetes, familial hypercholesterolaemia or chronic kidney disease. This was used to calculate 1‑year cardiovascular risk probabilities. Transition probabilities for the primary prevention (heterozygous‑familial) population were based on the Dutch lipid criteria for people with heterozygous‑familial hypercholesterolaemia, because the patient characteristics from THIN were not representative of this population. For the secondary prevention (heterozygous‑familial) population (see section 3.19), some patient characteristics (such as rate of diabetes and age) were different from known prevalence. To address this, the company used data from Mohrschladt (2003) in its base‑case analysis for this population.

3.27 Non‑cardiovascular death probabilities in the model increased in accordance with age and sex using UK life tables. Probability of cardiovascular events also increased with age, in line with published data.

3.28 Age‑adjusted utility values for the primary prevention (heterozygous‑familial) population were calculated using Health Survey for England (HSE) data for people with no history of cardiovascular disease, multiplied by the disutility associated with cardiovascular events taken from Ara and Brazier (2010). Baseline utilities in the model were as follows: non‑fatal myocardial infarction 0.765, unstable angina 0.765, acute coronary syndrome 0.765, ischaemic stroke 0.775.

3.29 Age‑adjusted utility values for the secondary prevention (heterozygous‑familial), high‑risk cardiovascular disease and recurrent events/polyvascular disease (non‑familial) populations were calculated using HSE data for people with no history of cardiovascular disease, multiplied by the disutility values associated with a coronary cardiovascular health state (cardiovascular event more than 1 year ago) taken from Ara and Brazier 2010. Utility multipliers in the model were: primary prevention of heterozygous‑familial hypercholesterolaemia 1 (assumed), secondary prevention of heterozygous‑familial hypercholesterolaemia 0.924, acute coronary syndrome (0 to 12 months) 0.765, history of ischaemic stroke 0.822, acute coronary syndrome (13 to 24 months) 0.924, coronary heart disease 0.924, peripheral arterial disease 0.924, and polyvascular 0.854. Disutilities for further cardiovascular events in the model were applied to the secondary prevention (heterozygous‑familial) population baseline utilities.

3.30 Initial costs of treatment for hypercholesterolaemia and cardiovascular events were based on the cost of hospitalisation, follow‑up care and medication. Drug acquisition costs for the comparators were taken from the January 2015 edition of the British National Formulary. The cost of the background therapy was weighted by the proportion of the cohort using the statin sources from market research data. The cost of alirocumab incorporated the patient access scheme.

3.31 Health state costs were based on the NICE guideline on lipid modification and costs for the first 3 years after a cardiovascular event were taken from the British national formulary, the NHS Drug Tariff, NHS reference costs, PSSRU unit costs, and the NICE guideline on stroke rehabilitation in adults. The costs for each health state were as follows:

3.32 The ERG stated that in terms of face validity, the company's model structure and transition probabilities were plausible. However, the ERG noted that the company's model omitted the transient ischaemic attack and stable angina health states and that it had limited capacity to capture multiple cardiovascular event histories. It also stated that the company omitted treatment‑emergent adverse events from the model. The ERG also noted that the secondary prevention (heterozygous‑familial) population (see section 3.19) using Mohrschladt had a higher cardiovascular risk compared with data from THIN. It was unable to verify the most appropriate risk without another external data source. The ERG believed that the company's use of THIN for cardiovascular event and transition probabilities was appropriate because QRISK2 risk estimates were not valid for the high cardiovascular risk population.

3.33 Although the ERG accepted the company's decision to use an LDL‑C threshold of 3.36 mmol/l for people with high‑risk cardiovascular disease, it noted that Jameson et al. reported a mean LDL‑C of 2.13 mmol/l in people with cardiovascular disease having atorvastatin in primary care in the UK. It also noted that a large proportion of people in THIN were having low‑intensity statins and may not have been on optimal statin treatment. The ERG stated that the mean baseline LDL‑C levels used by the company may not have been applicable to people having maximally tolerated statins and that it considered the company's mean LDL‑C levels to be uncertain.

3.34 The ERG had several comments about the company's assumptions used to scale the estimated effect of alirocumab to cardiovascular events:

3.35 The ERG stated that the company assumed 100% treatment continuation and compliance over the entire time horizon. It noted that the high compliance was in line with ODYSSEY (approximately 98%) and that the assumption was consistent with the NICE guideline on lipid modification and the technology appraisal guidance on ezetimibe for the treatment of primary (heterozygous-familial and non-familial) hypercholesterolaemia.

3.36 The ERG stated that the company's health state utility values were calculated and implemented appropriately. However, it had several comments on the costs used in the model:

3.42 The company undertook a number of probabilistic sensitivity analyses, stating that the uncertainty in the results reflected the wide confidence intervals from preliminary PCSK9 inhibitor outcomes data.

3.43 The company also undertook deterministic sensitivity analyses to explore the upper and lower bounds of the confidence interval or by varying selected inputs by an arbitrary ±20%. The ICERs for all populations were most sensitive to changes in the relationship of LDL‑C level to cardiovascular events and annual cardiovascular risk.

3.44 The company conducted subgroup analyses by LDL‑C level:

3.45 The company conducted a range of scenario analyses:

3.54 The ERG stated that the company's new evidence was consistent with the committee's preferred modelling assumptions.

3.55 The ERG agreed with the company's assertion that the rate ratios used to estimate the relationship between LDL‑C levels and cardiovascular outcomes for alirocumab were different from the rate ratios used in the NICE technology appraisals of ezetimibe and evolocumab. It noted that: