3 The manufacturer's submission
The Appraisal Committee (appendix A) considered evidence submitted by the manufacturer of bivalirudin and a review of this submission by the Evidence Review Group (ERG; appendix B).
3.1 The manufacturer's decision problem compared bivalirudin in combination with aspirin and clopidogrel against a strategy of heparin with glycoprotein inhibitor in combination with aspirin and clopidogrel. The population was adults with ST-segment-elevation myocardial infarction intended for primary PCI.
3.2 The main evidence for the clinical effectiveness of bivalirudin was from one prospective, dual-arm, single-blind, randomised multicentre study. The HORIZONS-AMI trial recruited 3602 patients with STEMI intended for a primary PCI and compared a strategy of bivalirudin with one using heparin plus glycoprotein IIb/IIIa inhibitors. Both treatment arms received aspirin and clopidogrel. The trial is ongoing and has at least 3 years of follow-up data available at this time.
3.3 Patients in the bivalirudin treatment group (n = 1800) received a 0.75 mg/kg intravenous bolus of bivalirudin followed by a 1.75 mg/kg/hour infusion that was continued uninterrupted until at least the end of the PCI. An optional post-procedural dose of 0.25 mg/kg/hour was recommended for up to 4 hours after the procedure; 6% (105/1749) of patients received this. A median of one vial (mean = 1.23 vials) of bivalirudin was used per patient (non-integer numbers of bivalirudin and glycoprotein IIb/IIIa inhibitor vials were increased to the next integer to cover wastage at patient level). Patients in the heparin with glycoprotein IIb/IIIa inhibitor treatment group (n = 1802) received 60 IU/kg heparin as recommended by the European Society of Cardiology guidelines for the management of patients with STEMI undergoing primary PCI. A glycoprotein IIb/IIIa inhibitor (either eptifibatide or abciximab) was given in accordance with its summary of product characteristics.
3.4 The ages of the patients were similar across the treatment groups, with slightly more older patients in the heparin with glycoprotein IIb/IIIa inhibitor group. The mean patient age was 61 years in both treatment groups. Slightly more than 75% of patients were men and more than 90% were white. The study was conducted at 123 sites in 11 countries; 57% of patients were enrolled in Europe and approximately 3% in the UK. Of the recruited population, 19 left the study because they withdrew consent and 28 were lost to follow-up.
3.5 There were two primary outcomes: a composite of major adverse cardiovascular events and major bleeding. Secondary outcomes were the components of major adverse cardiovascular events: death (all-cause mortality and cardiac mortality), reinfarction, target vessel revascularisation for ischaemia, and stroke. The rate of stent thrombosis was also assessed. Analyses were conducted on the intention-to-treat population and outcomes were presented for 30 days and for 1 year after PCI.
3.6 For the primary outcome of major adverse cardiovascular events, bivalirudin was not inferior to heparin plus glycoprotein IIb/IIIa inhibitor at 30 days and 1 year. In both treatment groups around 5.5% of patients had a major adverse cardiovascular event after 30 days and about 12% after 1 year. For the primary outcome of major bleeding, there was a statistically significant difference (p 0.0001) between the treatment groups at 30 days and at 1 year. At 30 days, major bleeding occurred in 5.1% of patients in the bivalirudin group and 8.8% of patients in the heparin plus glycoprotein IIb/IIIa inhibitor (comparator) group (p < 0.0001). At 1 year, major bleeding occurred in 5.8% of patients in the bivalirudin group and 9.2% of patients in the comparator group (p < 0.0001).
3.7 For the secondary outcomes of all-cause mortality and cardiac mortality, there were statistically significant differences between treatment with bivalirudin and treatment with heparin plus glycoprotein IIb/IIIa inhibitor. After 1 year of follow-up, all-cause mortality was 3.5% in the bivalirudin group and 4.8% in the comparator group (p = 0.037). The 1-year results for cardiac mortality were 2.1% for bivalirudin and 3.8% for comparator (p = 0.005). For the intention-to-treat population, 3-year all-cause mortality was significantly lower (p = 0.03) for bivalirudin (5.9%) than for the comparator (7.7%).
3.8 The proportion of patients who experienced a reinfarction was similar between the treatment groups, as was the proportion experiencing a stroke, and the differences were non-significant. A slightly higher percentage of patients in the bivalirudin group had a target vessel revascularisation than in the heparin plus glycoprotein IIb/IIIa inhibitor group (7.2% compared with 5.9% at 1 year), but this difference was not statistically significant.
3.9 The overall rate of any stent thrombosis at 30 days and at 1 year was identical in the bivalirudin and the heparin plus glycoprotein IIb/IIIa inhibitor groups. However, more stent thrombosis events occurred in the bivalirudin group within the first 24 hours of PCI.
3.10 Investigators reported comparable rates of adverse events between the treatment groups, with a trend towards fewer events in the bivalirudin group and significantly reduced rates of drug-related adverse events (p < 0.0001). Significantly fewer thrombocytopenia events occurred in the bivalirudin group (1.4%) than in the heparin plus glycoprotein IIb/IIIa inhibitor group (3.9%; p < 0.0001).
3.11 Subgroup analyses were presented for all-cause mortality and for major bleeding. For all-cause mortality, the results for bivalirudin were more favourable for most subgroups that were analysed, but differences were not statistically significant. Treatment with bivalirudin was associated with fewer incidents of major bleeding than treatment with heparin plus glycoprotein IIb/IIIa inhibitor.
3.12 In the HORIZONS-AMI trial, radial arterial access was used for 5.9% (214/3597) of patients overall, whereas brachial or femoral access was used for most patients. In clinical practice in England and Wales a higher level of radial access is expected. Because of the small numbers of patients undergoing a primary PCI with radial arterial access, the manufacturer was not able to carry out a statistically meaningful direct comparison of radial arterial access and other routes of access. Bivalirudin was associated with a statistically significant reduction in non-access site (organ) bleeding, with a relative risk of 0.684 (95% confidence interval 0.507 to 0.922) compared with heparin plus glycoprotein IIb/IIIa inhibitor.
3.13 The manufacturer submitted an economic analysis that employed an economic model based primarily on the analysis of patient-level data from the HORIZONS-AMI trial. The model has two parts. The first part is a decision-tree structure that covers the initial reperfusion to the end of a specified follow-up period (1 year in the base case). The tree splits into branches to allow for different events to occur: no relevant complications; minor bleed; major bleed; ischaemic stroke; myocardial infarction; repeat revascularisation; or death. The tree is coupled with a two-state (alive/dead) Markov model to account for subsequent survival over a 39-year horizon using an annual cycle length.
3.14 The model is primarily affected by the differences in all-cause mortality between the two treatments observed in the HORIZONS-AMI trial.
3.15 Absolute and relative clinical event risks and most resource use parameters were derived from the raw data of the HORIZONS-AMI intention-to-treat population. The relative risks of treatment with bivalirudin compared with heparin plus a glycoprotein IIb/IIIa inhibitor for major bleed, minor bleed, ischaemic stroke, repeat myocardial infarction, repeat revascularisation and death were 0.643, 0.536, 1.057, 0.817, 1.124 and 0.710, respectively. Average life expectancy assumed in the model was 11.26 years. Radial access use of 42.5% was assumed.
3.16 Costs of bivalirudin treatment were based on the costs of initial angiography, initial revascularisation and hospital care, anticoagulant medications, management of treatment-related adverse events and long-term follow-up. Unit costs for interventional procedures and associated treatment costs were based on NHS reference costs. Medication usage was based on usage in the HORIZONS-AMI trial. Treatment with the most expensive glycoprotein IIb/IIIa inhibitor (abciximab) was assumed in more than 70% of cases in which glycoprotein IIb/IIIa inhibitor was administered. The cost of heparin treatment was considered insignificant and was omitted from the model. The treatment cost per patient with bivalirudin was £412 compared with £573 for heparin plus glycoprotein IIb/IIIa inhibitor.
3.17 Two utility values were used in the model for the period after the initial STEMI event: a value of 0.683 was applied in the first year only and 0.718 was applied for each subsequent year of life. The utility values were obtained from a review of the literature. Costs and health outcomes were discounted at 3.5%.
3.18 In the base-case analysis the bivalirudin strategy dominated the heparin plus glycoprotein IIb/IIIa inhibitor strategy because it was cheaper and more effective, with total costs of £12,843 and £13,110, and total QALYs of 6.256 and 6.166, respectively. The manufacturer's deterministic sensitivity analysis showed that the results of the model were robust to a number of parameters, including changes in the relative risks of death and major bleed in the bivalirudin strategy, life expectancy, and type of glycoprotein IIb/IIIa inhibitor used.
3.19 Probabilistic sensitivity analysis presented for the base-case 1-year analysis showed that the bivalirudin strategy was dominant (that is, it was cost-saving and showed a QALY gain) in 9924 (99.2%) of 10,000 ICER results at the £20,000 per QALY threshold.
3.20 When the manufacturer presented an 'extreme case scenario' combining several unfavourable assumptions (only eptifibatide but no abciximab use; 100% radial arterial access; no difference in length of initial hospital stay), the ICER was £4106 per QALY gained (with a difference in cost of £367 and a difference in QALYs of 0.089). In all other scenarios, the bivalirudin strategy remained dominant.
3.21 The ERG noted that there is a lack of comparison of bivalirudin with heparin alone, but the use of heparin alone is not common UK practice. The ERG considered that the manufacturer's submission provided a thorough account of the only available RCT.
3.22 The ERG noted that the RCT reported data on the licensed dose of bivalirudin in patients undergoing PCI for STEMI. The RCT differed from standard UK practice in that pre-procedural heparin was used for most patients. Within the bivalirudin group, patients treated with pre-procedural heparin had a lower rate of major adverse cardiovascular events than those who did not receive pre‑procedural heparin. This pattern was not seen in the comparator group. However, there was no significant interaction between treatment group and pre-procedural heparin (p = 0.1060). The ERG noted that it is unclear how the RCT results would be reflected in practice, given the lack of pre-procedural heparin in standard UK practice. The RCT also differed from standard UK practice in using predominantly femoral rather than radial arterial access. The ERG thought that because access site bleeding is less common with radial than femoral arterial access, the benefit seen in reduced bleeding from bivalirudin is likely to be lower in practice than in the RCT.
3.23 Overall the ERG considered that the model structure employed by the manufacturer addresses the scope of the decision problem. The choice of intervention and comparator was appropriate. The ERG noted that the model applies the same relative risk of events with bivalirudin treatment to all patients in the bivalirudin group, regardless of whether they had undergone a primary PCI. However, this assumption did not alter the conclusions from the model.
3.24 The ERG noted that the decision tree part of the model treats clinical events as being mutually exclusive, although in reality they are not. For example, some patients could experience both a bleed and a stroke during the initial period after STEMI. The ERG noted that the same utility values are applied to both treatment strategies, such that any difference in health-related quality of life in the model is driven by differences in survival between the treatment groups. Utility decrements arising from complications following reperfusion were not included in the base-case analysis. However, the ERG noted that only negative health effects associated with repeat revascularisation would have an unfavourable impact on the cost effectiveness of bivalirudin.
3.25 The ERG identified some minor issues and discrepancies on costs in the model, but none of these had any significant impact on the results of the model.
3.26 The ERG was satisfied that the results of the manufacturer's model, which suggest that a bivalirudin-based intervention dominates heparin plus glycoprotein IIb/IIIa inhibitor, were robust to sensitivity analyses. Bivalirudin remained dominant across most sensitivity analyses and in cases in which it was more effective and more expensive than heparin plus glycoprotein IIb/IIIa inhibitor, the ICER for bivalirudin remained below £5000 per QALY gained.
3.27 The ERG conducted exploratory analyses to check that the manufacturer's model was robust. This included sensitivity analysis concerning the long-term Markov model to assess the impact of uncertainty around longer-term outcomes for patients after reperfusion. In this analysis the ERG found that the structure of the manufacturer's model and the use of a Markov component means that provided use of bivalirudin is associated with an improved survival rate and lower cost at 1 year, it will always dominate heparin plus glycoprotein IIb/IIIa inhibitor over longer time horizons.
3.28 The ERG investigated the following two issues:
the impact of assuming 2 vials of bivalirudin per patient instead of 1.23
the impact of using prasugrel in the comparator group rather than clopidogrel.
3.29 The ERG found that when the assumption of mean vial use was changed from 1.23 vials to 2 vials per patient, bivalirudin remained the dominant strategy.
3.30 The ERG stated that for investigating the impact of using prasugrel in the comparator arm rather than clopidogrel, there are no data directly comparing bivalirudin plus clopidogrel, with heparin plus glycoprotein IIb/IIIa inhibitor plus prasugrel. The ERG also found that there is a lack of data for treatment with heparin plus glycoprotein IIb/IIIa inhibitor plus prasugrel. There are two ongoing trials of bivalirudin plus prasugrel. The first is comparing bivalirudin plus prasugrel with clopidogrel plus heparin plus abciximab. The second trial is comparing bivalirudin plus prasugrel with clopidogrel plus heparin.
3.31 The ERG referred to details contained within the manufacturer's submission for NICE technology appraisal 182 on prasugrel for the treatment of acute coronary syndromes with PCI, which showed that prasugrel results in an additional per patient drug cost of £162 compared with clopidogrel. Costs of hospitalisation in each group were estimated to be similar. The ERG found that if this cost difference was applied to the heparin plus glycoprotein IIb/IIIa inhibitor group in the model, bivalirudin (given with clopidogrel) would remain cost-saving. However, evidence relating to the relative impact of bivalirudin plus clopidogrel versus prasugrel plus heparin and glycoprotein IIb/IIIa inhibitor is not available.
3.32 Full details of all the evidence are in the manufacturer's submission and the ERG report, which are available from www.nice.org.uk/guidance/TA230