Ticagrelor for the treatment of acute coronary syndromes

For the comparison of ticagrelor plus aspirin with clopidogrel plus aspirin, the manufacturer identified 1 trial, the PLATO trial, an international, multicentre, randomised, double-blind, double-dummy, parallel group, phase 3 study. The trial evaluated the efficacy and safety of ticagrelor plus aspirin compared with clopidogrel plus aspirin over 12 months in people with ACS whose symptoms began up to 24 hours before their admission to hospital. In the trial, 18,624 adult patients with ACS with or without ST-segment elevation on electrocardiogram from 43 countries including 18 UK centres (n=281) were admitted to hospital, and randomised to either ticagrelor plus aspirin (n=9,333) or clopidogrel plus aspirin (n=9,291). In the ticagrelor group, patients received a loading dose of 180 mg of ticagrelor, then 90 mg twice a day. Patients randomised to clopidogrel received loading doses of 300 to 600 mg of clopidogrel, then 75 mg every day thereafter. Patients did not need loading doses of clopidogrel if they had taken clopidogrel before admission or had received clopidogrel after admission but before randomisation (median approximately 5 hours). In the time between admission and randomisation, 46% of patients in both the ticagrelor and clopidogrel groups received clopidogrel. All patients also received aspirin (in addition to ticagrelor or clopidogrel) with a loading dose of 325 mg, then 75 to 100 mg daily. Patients already taking aspirin did not need a loading dose of aspirin.

The primary end point was time to first event (a composite of myocardial infarction, stroke or death from vascular causes). The planned duration of treatment and follow-up was 12 months. If before this time 1,780 individuals had a primary end point event, then patients who had not yet been followed for 12 months would finish the study at their 6 or 9-month visit. At the end of the trial, 1,878 participants had experienced events, and the median duration of treatment was 9.1 months. Secondary end points included: myocardial infarction; stroke; death from vascular causes; death from any cause; a composite of myocardial infarction, stroke and death from any cause; and a composite of myocardial infarction, stroke, severe recurrent cardiac ischaemia, recurrent cardiac ischaemia, transient ischaemic attack, other arterial thrombotic events and death from vascular causes.

The results showed that the relative risk of experiencing a primary end point event was 16% lower in the ticagrelor group compared with the clopidogrel group (hazard ratio [HR] 0.84; 95% confidence interval [CI] 0.77 to 0.92; p<0.001). Of the components of the primary end point, randomisation to ticagrelor plus aspirin reduced the incidence of myocardial infarction (HR 0.84; 95% CI 0.75 to 0.95; p=0.005) and death from vascular causes (HR 0.79; 95% CI 0.69 to 0.91; p=0.001), but not of stroke (HR 1.17; 95% CI 0.91 to 1.52). Randomisation to ticagrelor plus aspirin reduced the absolute risk of experiencing the primary end point from 11.7% to 9.8% at 12 months (absolute risk reduction 1.9%) compared with clopidogrel plus aspirin.

The manufacturer explored the consistency of effects and safety end points in 25 pre-specified subgroups and 8 post-hoc subgroups. An analysis was conducted of the primary end point in several predefined subgroups. The manufacturer's submission stated that analyses showed statistically significant differences in treatment efficacy in 3 groups: geographic region; body weight above or below a gender-specific median; and use of lipid-lowering drugs at randomisation. The HRs by type of ACS at presentation – unstable angina, NSTEMI and STEMI – were 0.96 (95% CI 0.75 to 1.22), 0.83 (95% CI 0.73 to 0.94) and 0.84 (95% CI 0.72 to 0.98) respectively with a non-statistically significant test for interaction (p=0.41). The manufacturer presented 6 analyses in subgroups that included patients whose condition was managed invasively, managed medically, patients with STEMI, patients with diabetes, patients with genetic polymorphisms, and patients undergoing CABG. The results of these 6 subgroup analyses were generally consistent with the primary analysis.

The manufacturer reported adverse events from the PLATO study, specifically bleeding, dyspnoea and ventricular pauses. There was no statistically significant difference in the primary safety end point of 'major' bleeding between the ticagrelor plus aspirin and clopidogrel plus aspirin groups (11.6% versus 11.2% respectively; p=0.43), or in the end point of bleeding defined by the Thrombolysis in Myocardial Infarction (TIMI) scale. Both were analysed according to which treatment a patient took, rather than to which a patient had been randomised; these findings were consistent across all major subgroups. Patients randomised to ticagrelor experienced more overall major and minor bleeding (HR 1.11; 95% CI 1.03 to 1.20; p=0.008) as well as more major bleeding not related to CABG (HR 1.19; 95% CI 1.02 to 1.38; p=0.03). Intracranial bleeding was more common in the ticagrelor plus aspirin group than in the clopidogrel plus aspirin group, with fatal intracranial bleeding statistically significantly more common in the ticagrelor plus aspirin group (HR not reported; p=0.02). Fatal bleeding excluding intracranial bleeding was statistically significantly more common in the clopidogrel plus aspirin group (HR not reported; p=0.03). There was no difference between the 2 groups in relation to overall fatal bleeding (0.3% in each group). Patients randomised to ticagrelor experienced dyspnoea statistically significantly more often than patients taking clopidogrel (13.8% versus 7.8% respectively; p<0.001). More patients taking ticagrelor plus aspirin discontinued treatment because of dyspnoea than patients taking clopidogrel plus aspirin (0.9% versus 0.1% respectively; p<0.001). Holter monitoring detected more ventricular pauses of 3 seconds or longer during the first week in the ticagrelor plus aspirin group than in the clopidogrel plus aspirin group, but these occurred infrequently at 30 days of treatment and were rarely associated with symptoms. Patients treated with ticagrelor had statistically significantly greater increases from baseline in levels of serum uric acid and serum creatinine compared with those on clopidogrel (p<0.001 for both events throughout the study).

The manufacturer identified no trials directly comparing ticagrelor plus aspirin with prasugrel plus aspirin. Instead, the manufacturer identified 2 trials comparing prasugrel plus aspirin with clopidogrel plus aspirin that provided data for an indirect comparison: the PLATO trial (ticagrelor plus aspirin compared with clopidogrel plus aspirin) and TRITON-TIMI 38, which compared prasugrel plus aspirin with clopidogrel plus aspirin in patients (n=13,608) with ACS and scheduled PCI. The manufacturer took the view that the trials were not comparable and, by inference, a comparison between prasugrel and ticagrelor based on these trials was inappropriate and should be viewed with caution. The manufacturer noted that the PLATO and TRITON-TIMI 38 trials were similar in many ways, both including populations with ACS, both comparing the intervention plus aspirin to clopidogrel plus aspirin, and both sharing the same primary end point. However, there were important differences in the use of PCI and medical management, in the size and timing of the loading dose of clopidogrel, and in assessing myocardial infarction. Although the manufacturer considered the indirect comparison inappropriate, it cited a published paper based on the PLATO and TRITON-TIMI 38 trials that showed no statistically significant differences in the occurrence of myocardial infarction, stroke, death from any cause or the composite of these outcomes between the 2 drugs. Ticagrelor plus aspirin was associated with a statistically significantly lower risk of major bleeding and major bleeding specifically associated with bypass grafting than prasugrel plus aspirin. The risk of major bleeding not related to CABG did not differ between patients taking prasugrel and those taking ticagrelor.

The PLATO trial included a pre-specified sub-study of health economics and quality of life that evaluated the health-related quality of life for ticagrelor plus aspirin compared with clopidogrel plus aspirin. Investigators administered the EuroQual 5D (EQ-5D) questionnaire to 8,840 patients at discharge from hospital for the index ACS event, again at 6 months, and at the end of treatment in all countries where a version of EQ-5D in the country's official language was available. No differences in any of the items on the EQ-5D were found between the ticagrelor plus aspirin group and the clopidogrel plus aspirin group.

The manufacturer did not identify any publications that evaluated the cost effectiveness of ticagrelor for the treatment of ACS. The manufacturer developed a new economic model, informed by 9 existing economic evaluations. For the health economics evaluation of ticagrelor plus aspirin compared with prasugrel plus aspirin, the manufacturer presented the results of a published indirect comparison of the TRITON-TIMI 38 trial and the PLATO trial, conducted by an independent group.

The manufacturer constructed a 2-part cost−utility model with a 1-year decision tree to model effectiveness based on data from the PLATO study, and a Markov model to extrapolate costs and benefits to a lifetime horizon (40 years), and to incorporate major clinical events. Patients in the model had ACS (STEMI, NSTEMI or unstable angina) and included patients whose condition was managed medically or with PCI or CABG; the model therefore reflected the marketing authorisation for ticagrelor. The model compared ticagrelor plus aspirin with clopidogrel plus aspirin.

The 1-year decision tree contained 4 mutually exclusive health states: non-fatal myocardial infarction, non-fatal stroke, death from any cause, and no further event. The Markov model included 6 states: non-fatal myocardial infarction, post-myocardial infarction, non-fatal stroke, post-stroke, death, and no further event. Non-fatal myocardial infarction and non-fatal stroke were tunnel states, which allowed for a worse prognosis the first year after a non-fatal event compared with second and subsequent years. After the first year following a non-fatal event, patients proceeded to 1 of 4 mutually exclusive health states: post-myocardial infarction, post-stroke, death or no further event. Costs and health outcomes were discounted at 3.5%. The Markov model used a half-cycle correction to adjust for simulated costs and outcomes. The model did not permit a patient to discontinue treatment for any reason other than death.

In the model, costs, life years, and quality-adjusted life years (QALYs) accrued beyond the first year of treatment with ticagrelor or clopidogrel; however, the model assumed that the beneficial effect of ticagrelor does not persist beyond 1 year. This means that the transition probabilities between states in the Markov model were the same for both treatment arms; the only difference between treatment arms was the number of patients who started the Markov model in each state, which was based on the output of the 1-year decision tree. Adverse events (notably bleeding) were not included in the structure of the model, but the increased costs and decreased health-related quality of life associated with adverse events recorded in PLATO (as part of PLATO-HECON) were included in the first year (decision tree) of the model. The manufacturer assumed that adverse events including bleeding and dyspnoea have no lasting effects beyond the 12-month duration of the trial. To model the incidence of cardiovascular complications beyond 1 year (in the Markov component of the model), the manufacturer assumed a constant probability of 3.15% per year for non-fatal myocardial infarction and 1.02% per year for non-fatal stroke. The risk of death from MI after the index event (STEMI, NSTEMI or unstable angina) was assumed to be the same as that of death at least 1 year after the index ACS event.

For the 1-year decision tree, the manufacturer used a parametric time-to-event survival model with a Weibull distribution to determine the baseline risk (that is, the risk of cardiovascular events and death in the clopidogrel group). The manufacturer then applied HRs reflective of the effectiveness of ticagrelor from the PLATO study to this baseline risk to determine the risk in patients taking ticagrelor. Using data in the 1-year decision tree derived from the PLATO study, the manufacturer estimated from patients in the clopidogrel group age-adjusted event rates (myocardial infarction, stroke, death from any cause and death from vascular causes) for a UK population with ACS (mean age of PLATO patients, 62.2 years; reported age of UK patients with ACS in 2009 to 2010, 69.7 years). In the Markov model, the transition probabilities from the no event health state to each of the non-fatal myocardial infarction or non-fatal stroke health states were estimated from a study that the manufacturer commissioned from the Myocardial Ischaemia National Audit Project and the General Practice Research Database. The probabilities of transitioning between all other health states were based on relative risks applied to the probability of death in standard life tables.

The manufacturer used the 12-month cohort (patients who were eligible for a 12-month follow-up) in the PLATO-HECON study to calculate the utility accrued in the study and reported it as the average utility value for a patient over the 12-month period using the EQ-5D. The manufacturer performed a literature search to assess the relationship between utility values in the PLATO study and in the literature. The lower values from the literature were used in sensitivity analyses. The utility scores from both the PLATO-HECON substudy and the literature were adjusted downwards by 0.0328 to better reflect the patient population that would be treated in UK clinical practice. In addition, because utility decreases with age, the manufacturer applied a utility decrement of 0.004 in the Markov model to each cycle beyond the first year.

The costs for the generic drugs clopidogrel and aspirin were taken from the NHS Drug Tariff, November 2010. The cost of the drugs used in the economic evaluation were: aspirin 28-pack, £0.82; clopidogrel 30-pack, £3.40; and ticagrelor 28-pack, £54.60. The PLATO-HECON substudy measured resource use and determined costs for all patients participating in the PLATO study by recording admissions to hospital, interventions, investigations, blood products, re-operations due to bleeding, and use of concomitant or study drugs to estimate total healthcare costs associated with ticagrelor and clopidogrel. Resource use included costs from randomisation to the time of discharge from hospital, as well as after discharge from hospital to the end of the PLATO study. The manufacturer also included in sensitivity analyses the costs of a visit to the GP and of a blood test to check renal function, as stipulated in the SPC for ticagrelor.

In its deterministic base case (40-year time horizon), the manufacturer's model estimated that ticagrelor provides an incremental health gain of 0.108 QALYs compared with clopidogrel, at an incremental cost of £379, resulting in an incremental cost-effectiveness ratio (ICER) of £3,521 per QALY gained. The manufacturer also presented results using time horizons of 1 year, 5 years, 10 years and 20 years: the ICER differed substantially from the base-case ICER only when using the 1-year time horizon, with an ICER of £33,764 per QALY gained. The manufacturer also presented base-case ICERs for the subgroups of ACS specified in the scope, which were £2,551 per QALY gained for STEMI, £5,217 per QALY gained for NSTEMI and £5,310 per QALY gained for unstable angina.

The manufacturer carried out deterministic sensitivity analyses to the base case and showed the effects of changing 43 model parameters. Only the change to the costs of the 'no further event' health state impacted substantially on the results. When the cost of the 'no further event' health state for ticagrelor plus aspirin was set to its lowest, ticagrelor plus aspirin dominated clopidogrel plus aspirin (that is, ticagrelor plus aspirin was more effective and less expensive than clopidogrel plus aspirin), whereas when the cost of the clopidogrel plus aspirin 'no further event' health state was set to its lowest, the ICER was £21,000 per QALY gained. Changes in all other parameters did not increase the ICER beyond £7,620.

The manufacturer ran scenario analyses for 0% and 6% discount rates, using published rather than PLATO-derived utility values, removing the 0.0328 downwards utility adjustment and removing the age-related decrease in utility per cycle. The results of the scenario analyses showed that the ICER for ticagrelor plus aspirin compared with clopidogrel plus aspirin ranged from £2,358 to £4,699 per QALY gained.

The cost-effectiveness acceptability curve showed that at £5,000 per QALY gained, the probability of ticagrelor plus aspirin being cost effective compared with clopidogrel plus aspirin was 76.6%. At £20,000 per QALY gained, the probability of ticagrelor plus aspirin being cost effective compared with clopidogrel plus aspirin was 99.9%.

The manufacturer's submission also provided results for ticagrelor plus aspirin compared with prasugrel plus aspirin for the subgroup receiving PCI, based on the results of a published indirect comparison of the PLATO and TRITON-TIMI 38 trials. Because of the small proportion of patients who participated in the TRITON-TIMI 38 substudy of quality of life (EQ-5D was collected in only 461 of 13,608 patients at baseline), the model incorporated utility information from the literature, rather than from the substudy. If costs from the PLATO-HECON substudy were not available, the manufacturer used NHS reference costs in the analysis for prasugrel plus aspirin. The manufacturer obtained the cost of prasugrel from MIMS, October 2010. The analysis of ticagrelor plus aspirin compared with prasugrel plus aspirin resulted in an incremental cost of £227, incremental QALYs of 0.065 and an ICER of £3,482 per QALY gained, with a 40-year time horizon. The manufacturer stated that the results of the indirect comparison should be viewed with caution because of the problems associated with the indirect comparison of ticagrelor plus aspirin with prasugrel plus aspirin discussed in section 3.6.

The ERG conducted a literature search and agreed that the PLATO trial was the only trial relevant to the decision problem. The ERG considered that the PLATO trial was well designed with robust processes for randomisation and blinding. It noted that compliance and deviations in protocol were similar across treatment arms. Although only 281 patients in the PLATO trial were from centres in the UK, the ERG considered that they were not dissimilar to other European participants. The ERG also noted that participants in the PLATO trial were younger than patients with ACS in England and Wales, but that the manufacturer's model accounted for this difference.

The ERG noted that for patients with STEMI not undergoing PCI, NICE recommends dual antiplatelet therapy (clopidogrel plus aspirin) for at least 4 weeks (from NICE's previous guideline on MI: secondary prevention, now replaced by NICE's guideline on myocardial infarction: cardiac rehabilitation and prevention of further cardiovascular disease). From statements in the 'clinical need and practice' and 'evidence and interpretation' sections of NICE's technology appraisal guidance on drug-eluting stents for the treatment of coronary artery disease, the ERG concluded that standard practice for STEMI should include dual antiplatelet therapy for 3 months for patients undergoing revascularisation with bare-metal stents and 12 months for patients undergoing revascularisation with drug-eluting stents.

The ERG considered that the PLATO trial reflects current clinical practice and that all patients received antiplatelet treatment at a clinically appropriate dose. The ERG was satisfied with the manufacturer's means of categorising adverse events from bleeding. The ERG expressed concerns about the components of the primary efficacy end point in the PLATO trial. Firstly, the primary end point was inconsistent with the concept that all components of an end point should be of similar importance to patients. For example, the average utility values from the 12-month cohort in the PLATO-HECON study used in the manufacturer's model differed by end point and were 0.246 for death from a vascular cause, 0.812 for myocardial infarction and 0.736 for stroke. Secondly, the primary end point was inconsistent with the concept that all components of an end point occur with similar frequencies. For example, in 18,624 participants there were 795 vascular deaths, 1,097 myocardial infarctions and 231 strokes during the median 9.1-month follow-up in the PLATO study. Thirdly, the primary end point was inconsistent with the concept that the effect of a treatment should have an effect of similar magnitude and direction on all components of a primary end point. For example, in the PLATO study, the HR for stroke (non-significantly higher with ticagrelor) differed from those for myocardial infarction and death from vascular causes (significantly lower with ticagrelor). The ERG concluded that the results of the overall composite end point should be interpreted cautiously. The ERG also noted that the manufacturer excluded 'silent' myocardial infarctions (defined as new or presumed pathological Q waves on ECG in the absence of symptoms). The ERG considered that the secondary end points and their components reflected those used in other cardiovascular trials.

The ERG noted that the manufacturer tested whether the effectiveness and safety of ticagrelor plus aspirin compared with clopidogrel plus aspirin differed across 25 pre-specified and 8 post-hoc subgroups, without adjustment for multiple comparisons. The ERG expressed concern about the large number of subgroups and potential overemphasis of any statistically significant results from these analyses, which might have occurred by chance alone. With these caveats noted, the ERG observed that the regional analysis showed that in the USA, patients randomised to ticagrelor plus aspirin did worse than those randomised to clopidogrel plus aspirin.

For patients with STEMI who receive bare-metal stents, the ERG highlighted concerns about the comparator treatment included in the economic evaluation. It interpreted NICE's technology appraisal guidance on drug-eluting stents for the treatment of coronary artery disease as stating that dual antiplatelet therapy for 3 months was standard practice for patients undergoing revascularisation with bare-metal stents, whereas for patients undergoing revascularisation with drug-eluting stents, NICE's previous guideline on MI: secondary prevention (now replaced by NICE's guideline on myocardial infarction: cardiac rehabilitation and prevention of further cardiovascular disease) recommends dual antiplatelet therapy for 12 months. Another concern of the ERG was that the manufacturer treated the STEMI group as a homogeneous population and estimated a single ICER. By contrast, the ERG believed that STEMI has 4 distinct populations differing by treatment: STEMI with medical management, STEMI revascularised with drug-eluting stent, STEMI revascularised with bare-metal stent and STEMI with other intervention (for example, CABG).

The ERG stated that as the trial was designed to test the efficacy of 12 months of treatment, all patients should have been treated for 12 months. The ERG noted that the PLATO trial design did not involve uniform duration of treatment, instead, the protocol stipulated that patients could leave the study at their 6- or 9-month visit if a predetermined number of primary end-point events had occurred by that time. Approximately 44% of patients were followed up for 12 months in the trial. This increased the uncertainty in the estimates of effectiveness at the conclusion of the trial, which in turn was the prime driver of the Markov model and, therefore, the long-term benefits for patients.

The ERG noted that the model featured 2 separate paths. In 1 path, after first presentation with ACS, patients may have a subsequent non-fatal myocardial infarction at any time during the decision-tree part of the model and remain in the non-fatal myocardial infarction health state to the end of the decision-tree part of the model, then progress to the 'post-myocardial infarction' state for all remaining cycles until death (whether from cardiovascular or non-cardiovascular causes). Similarly, patients may instead have a non-fatal stroke as their first event (after the initial presentation with ACS) during a cycle, and then progress to the post-stroke state until death. The ERG considered that this structure does not represent reality, because it does not allow patients to have more than 1 myocardial infarction, more than 1 stroke, or both myocardial infarctions and strokes in their lifetime following their initial presentation with ACS, and that this may bias future costs and benefits. The ERG also noted that the model simplified the natural history of treated cardiovascular disease by keeping constant the transition probability of previously event-free patients (since initial treatment for ACS) experiencing a non-fatal myocardial infarction or stroke throughout the long-term Markov model. The modelling ignored the increase in risk associated with other factors, notably, increasing age. The ERG considered that this omission may have led to the manufacturer's model inaccurately estimating future events, costs and progressive changes in the outcomes and quality of life of patients.

The ERG was concerned that the model applied an average utility score for the first year, whereas clinical experience showed that ACS patients experience an initial decline in utility that steadily improves. Therefore, the ERG noted that the ICER at 12 months may be an underestimate.

The ERG noted that in the manufacturer's submission the subgroups of interest in the economic evaluation did not reflect the subgroups of interest in the clinical section. The ERG could not verify the estimates of clinical effectiveness used in the manufacturer's model ascribed to ticagrelor in patients with NSTEMI or unstable angina. The ERG also noted that the manufacturer considered the subgroup with unstable angina as a homogeneous group whereas, in clinical practice in England and Wales, physicians typically categorise patients into lowest, low, intermediate, high and highest risk groups using the Global Registry of Acute Coronary Events (GRACE) classification and treat them accordingly.

The ERG noted that the manufacturer adjusted the age of the modelled patients to reflect the UK population with ACS. The ERG noted potential problems with the methods chosen by the manufacturer, which may have led to inaccuracies. The ERG established that these inaccuracies represented an 8% underestimate of benefits from ticagrelor plus aspirin compared with clopidogrel plus aspirin and suggested that the ICER presented by the manufacturer may be an overestimate.

The ERG acknowledged that use of healthcare resources was estimated in the model using data from an imbedded health economic study, which collected details of hospital care received by patients during the PLATO trial. For the purposes of the model, only data for those patients in the 12-month cohort were included. This cohort comprised of patients who, based on timing of enrolment, had the potential to receive 12 months treatment with ticagrelor. The ERG also noted that for each patient category in the model, the resources used by each patient were calculated separately for each treatment arm, and these were multiplied by a corresponding unit cost and totalled for an estimated hospital-care cost per patient for the first 12-month period. The ERG had some concerns relating to this type of resource analysis, and conducted a combined analysis of resource use (taking the ticagrelor and clopidogrel groups together), making some adjustments for double-counting of costs. Results suggested that the health state costs with ticagrelor were £100 lower (rather than £371 lower, as in the manufacturer's base-case) than the health state costs of clopidogrel, which would have the effect of doubling the estimated ICER at the 1-year time horizon.

The ERG noted that the manufacturer's base-case analysis estimated costs for the study drugs assuming 100% use in the trial period, despite evidence of deaths before the end of follow-up, treatment withdrawals, and poor adherence in some participants. The ERG instead incorporated data on drug use from the PLATO trial and noted that this reduced the average cost of both ticagrelor and clopidogrel substantially, and the difference in drug costs of ticagrelor plus aspirin compared with clopidogrel plus aspirin reduced from £651 to £507 per patient. Applying the ERG's amended age adjustment, resource use, and costs of study drugs to the manufacturer's model resulted in a 42% increase in the manufacturer's ICER for the 1-year time horizon from £36,177 to £51,204 per QALY gained. However, the ERG emphasised that both the incremental costs and additional benefits associated with ticagrelor plus aspirin compared with clopidogrel plus aspirin were very small at longer time horizons, and subject to considerable uncertainty.

The ERG conducted a wide-ranging sensitivity analysis, calculating overall deterministic cost-effectiveness estimates for all combinations of 4 long-term variables – survival gain at 12 months, life expectancy at 12 months, the mean long-term utility value and the mean long-term discounted cost per patient year. The most favourable ICERs for ticagrelor plus aspirin are £3,407 per QALY gained for all patients, £3,551 per QALY gained for the STEMI group, £3,350 per QALY gained for the NSTEMI group and £3,405 per QALY gained for the group with unstable angina. Incorporating the least favourable combination of assumptions resulted in an estimated ICER for ticagrelor plus aspirin below £20,000 per QALY gained for each of the specified populations compared with 12 months' clopidogrel plus aspirin treatment. The central estimates from these sensitivity analyses were £7,897 per QALY gained for all patients, £8,872 per QALY gained for the STEMI group, £7,215 per QALY gained for the NSTEMI group and £9,131 for the subgroup with unstable angina.

The ERG noted that there are no head-to-head trial data comparing ticagrelor plus aspirin with prasugrel plus aspirin. With regard to the indirect comparison of ticagrelor plus aspirin with prasugrel plus aspirin, the ERG considered that any comparison of the PLATO and TRITON-TIMI 38 trials posed problems. The ERG agreed with the manufacturer that sufficient clinical evidence is not yet available for a credible indirect comparison of ticagrelor plus aspirin compared with prasugrel plus aspirin for patients with ACS. It concluded that the effectiveness and safety of ticagrelor compared with prasugrel remains unknown.

Full details of all the evidence are in the manufacturer's submission and the ERG report.