Prasugrel (Efient®) with percutaneous coronary intervention for treating acute coronary syndromes (review of TA182): systematic review and economic analysis

ST segment elevation myocardial infarction

The objective of treatment for patients with STEMI is rapid and sustained revascularisation. 10 The recommended treatment for people with confirmed STEMI is immediate (primary) PCI to the occluded artery. 9,11 Clinical guidelines produced by National Institute for Health and Care Excellence (NICE) (CG1678) recommend coronary angiography with follow-on PCI (if indicated) as the preferred treatment for acute STEMI if presentation is within 12 hours of the onset of symptoms and primary PCI can be delivered within 120 minutes. When PCI facilities are not immediately available, treatment with thrombolysis (pharmacological reperfusion achieved through the use of ‘clot-busting’ drugs) should be considered. 12 When STEMI persists despite thrombolytic treatment, PCI (rescue) in an appropriately equipped unit should be considered. 8

Unstable angina/non-ST segment elevation myocardial infarction

The objective of treatment for patients with UA/NSTEMI is to alleviate pain and anxiety, prevent recurrences of ischaemia and prevent, or limit, progression to further acute MI. 1 NICE Clinical Guideline CG9413 recommends that people presenting with UA/NSTEMI are initially treated with aspirin and antithrombin therapy. Their risk of further cardiac events should then be assessed using a risk score measurement tool that predicts 6-month mortality, such as the Global Registry of Acute Cardiac Events (GRACE). 14 In addition to a GRACE14 score, additional factors should be considered, including full clinical history [age, previous MI, previous coronary artery bypass grafting (CABG)], physical examination (including measurement of blood pressure and heart rate), and resting 12-lead ECG and blood tests (troponin I or T, creatinine, glucose and haemoglobin). Table 1 is adapted from NICE CG9413 and describes the risk categories of future CV events assigned to risk scores.

Patients considered to be at intermediate to high risk should be offered coronary angiography and follow-on PCI (if appropriate) within 96 hours of admission. 15 Patients with UA/NSTEMI who are clinically unstable or at high ischaemic risk should be offered angiography as soon as possible. 13 Patients at low risk should be treated medically; however, if ischaemia is subsequently experienced or is demonstrated on ischaemia testing, coronary angiography and delayed PCI (if appropriate) should be offered. 13

British Cardiovascular Intervention Society Audit Data

The British Cardiovascular Intervention Society (BCIS) continuously audits interventional activity in the UK and the results are published annually. The most recent audit returns are for the year 2012. 17 The audit shows that there are currently 99 NHS PCI centres in the UK, almost double the number recorded in 2002. In 2012, 91,000 PCI procedures (for all indications) were carried out in the UK NHS, 27.4% in STEMI patients and 36.9% in UA/NSTEMI patients; the remainder were rescue or facilitated PCIs. A total of 24,631 PCIs for STEMI were conducted, the majority of which (23,842) were primary PCIs. The number of PCIs for STEMI has increased over time while the number of PCIs for UA/NSTEMI has remained stable.

Of patients referred for PCI in the UK in 2012, 74% were male and the average age was 64.9 years. 17 Approximately 20% had diabetes mellitus and 27% had had a previous MI. 17 One-quarter were current smokers and the majority (92%) were European. 17 It should be noted that these data are for an overall population of patients treated with PCI and, therefore, include patients other than those with ACS.

There are 85 NHS PCI centres in England and four in Wales. The total number of PCIs (all indications) performed in the NHS in England and Wales in 2012 was 75,217 and 3850, respectively. Almost 21,000 PCI procedures in England and 1000 in Wales were primary PCI procedures.

The BCIS audit data17 show that the number of PCIs performed in England and Wales has increased annually, although the rate of increase has slowed. In 2002, fewer than 30,000 procedures were carried out and, in contrast, almost 80,000 PCIs were conducted in 2012. The BCIS data describe the use of the radial artery (guidewire inserted through the wrist) as the access point for PCI. Radial access has risen to 65% of PCIs conducted in 2012 from 10% in 2004.

Intervention

The oral antiplatelet prasugrel, used within its licensed indication, is the focus of this review. The Summary of Product Characteristics (SPC) for prasugrel is available from the Electronic Medicines Compendium. 24

Prasugrel is a third-generation oral thienopyridine adenosine diphosphate receptor antagonist. It has a more rapid onset of action than clopidogrel as it requires only a single, relatively rapid metabolic step to produce the active agent (clopidogrel requires two steps). Prasugrel is prescribed as an adjunctive therapy to PCI to reduce platelet aggregation by irreversibly binding to P2Y₁₂ receptors. It is available as 5-mg or 10-mg film-coated tablets. Prasugrel is given (with aspirin) as a single 60-mg loading dose and then continued at 10 mg daily for up to 12 months.

Prasugrel is licensed in Europe25 to be co-administered with aspirin, for the prevention of atherothrombotic events in patients with ACS (STEMI and UA/NSTEMI) undergoing primary or delayed PCI. As stated in the SPC, the use of prasugrel in patients with a history of stroke or transient ischaemic attack (TIA) is contraindicated, whereas in older (≥ 75 years) patients prasugrel is generally not recommended. For patients who weigh < 60 kg, the 60-mg loading dose of prasugrel should be used followed by a maintenance dose of 5 mg. 24 The SPC further states that, in patients with UA/NSTEMI in whom coronary angioplasty is performed within 48 hours after admission, the loading dose of prasugrel should be given only at the time of PCI.

NICE guidance (TA18221) limits the use of prasugrel (co-administered with aspirin) in the NHS to people with ACS having PCI only when:

• immediate primary PCI for STEMI is necessary

• stent thrombosis has occurred during clopidogrel treatment

• the patient has diabetes mellitus.

In TA182,21 prasugrel was not recommended for patients with UA/NSTEMI who do not have diabetes mellitus or have not had a stent thrombosis following treatment with clopidogrel.

There is no patient access scheme in operation in the NHS for prasugrel.

The SPC for prasugrel highlights the increased bleeding risk for patients with ACS who are treated with prasugrel and aspirin. It is noted that the use of prasugrel in patients at increased risk of bleeding should be considered only when the benefits in terms of preventing ischaemic events are deemed to outweigh the risk of serious bleeding. 24

Current usage in the NHS

The decision paper26 presented to the Guidance Executive of NICE in June 2012 stated that the market share for prasugrel in terms of prescriptions had risen from 1% to 2% since 2011 and the monthly spend in the NHS had increased from approximately £400,000 to approximately £500,000. Data from the BCIS audit18 illustrate that prasugrel use has increased marginally between 2011 and 2012 (Table 3).

The current British National Formulary (BNF)27 list price of prasugrel for both 5-mg and 10-mg tablets is £47.56 per pack of 28 tablets. The current Drug Tariff28 list price of aspirin 75 mg is 0.82 pence per pack of 28 tablets.

Comparators

The stated comparators to prasugrel in the final scope issued by NICE7 are clopidogrel (generic) and ticagrelor, both in combination with low-dose aspirin.

Inclusion and exclusion criteria

Two reviewers (JG and NF) independently screened all titles and abstracts identified via searching and obtained full-paper manuscripts that were considered relevant by either reviewer (stage 1). The relevance of each study was assessed (JG/NF) according to the criteria set out below (stage 2), and studies that did not meet the criteria were excluded and their bibliographic details were listed alongside reasons for their exclusion. Any discrepancies were resolved by consensus and, when necessary, a third reviewer (AB) was consulted.

Quality assessment strategy

The quality of the clinical effectiveness studies was assessed independently by two reviewers (JG and KD) according to the Centre for Reviews and Dissemination at the University of York’s suggested criteria. 35 All relevant information is tabulated and summarised within the text of the report. Full details and results of the quality assessment strategy for clinical effectiveness studies are reported in Appendix 2.

Methods of data synthesis

The results of the clinical data extraction and clinical study quality assessment are summarised in structured tables and as a narrative description. An indirect treatment comparison of prasugrel with ticagrelor was planned.

Quantity and quality of research available

A total of 1940 titles and abstracts were screened for inclusion in the review of clinical effectiveness evidence. The process of study selection is shown in Figure 1. Titles excluded at stage 2 (n = 111) are listed in Appendix 3 along with reasons for their exclusion. The AG identified the pivotal trial (TRITON-TIMI 3836) discussed in TA18221 but did not identify any new trials for inclusion in the review.

FIGURE 1.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart.

At stage 2, the AG excluded four clinical trials. 37–40 One of the trials37 compared prasugrel with clopidogrel in a population of Asian patients with ACS undergoing PCI. This was excluded as it was considered to be a dose-ranging trial with a clopidogrel control. The trial recruited 719 patients and randomised them to one of three dosing regimens of prasugrel or standard clopidogrel according to patient weight and age (< 60 kg and > 70 years or vice versa). The primary outcome was platelet aggregation at 4 hours after the loading dose. Secondary outcomes included major adverse cardiac events and CABG and non-CABG thrombolysis in myocardial infarction (TIMI) bleeding at 30 days and 90 days. The study was not powered to detect differences between treatments on the secondary outcomes. The JUMBO-TIMI (Joint Utilization of Medications to Block Platelets Optimally – Thrombolysis in Myocardial Infarction 26)38 trial was similarly excluded. In this trial, patients (n = 904) undergoing PCI were randomised to one of three prasugrel dosing regimens or to clopidogrel and followed up for 30 days.

Two further excluded trials39,40 included relevant comparators and patient populations but had pharmacodynamic (platelet aggregation) parameters. The AG considered that the trial populations were too small and the length of follow-up too short (5 days and 1 hour) to provide data relevant to this review.

Assessment of clinical effectiveness

The AG’s systematic search of clinical effectiveness evidence yielded one relevant RCT (TRITON-TIMI 3836) for inclusion in the review. This trial was the pivotal trial discussed in TA18221 and the key elements of this RCT are summarised in Table 5. The TRITON-TIMI 3836 trial included 13,608 patients and was conducted in 30 countries. Patients received a loading dose of either prasugrel or clopidogrel (60 mg or 300 mg, respectively) followed by daily maintenance doses of 10 mg or 75 mg, respectively.

The results of the AG’s quality assessment of the TRITON-TIMI 3836 trial are presented in Appendix 2. Overall, the AG considers that the trial was robustly designed and of strong methodological quality.

As this report is an update of TA182,21 the AG has reproduced the original summary information for TRITON-TIMI 3836 in Appendix 4. The summary information presented includes:

• patient baseline characteristics (overall trial population)

• primary and secondary end point analyses (overall trial population)

• prespecified subgroup analyses for diagnosis, sex, age, diabetic status, type of stent implanted, use of glycoprotein IIb/IIIa receptor agonist, renal function (overall trial population)

• outcomes for STEMI patients (overall trial population)

• primary outcome for UA/NSTEMI, STEMI, all ACSs, patients with diabetes mellitus, patients with stents (overall trial population)

• outcomes for people with history of stroke/TIA

• outcomes for people > 70 years or weighing < 60 kg

• analyses of recurrent events following PCI (overall trial population).

A number of subgroup analyses relating to TRITON-TIMI 3836 have been published; the key publications are listed, along with a brief description, in Table 6. A more comprehensive list of associated publications is presented in Appendix 5 of this report. The paper by Wiviott (2011),42 which is directly relevant to this appraisal, focuses on a sub-population of patients from the TRITON-TIMI 3836 trial who are described as the ‘core clinical cohort’. This sub-population is discussed in TA18221 as the ‘target population’. The core clinical cohort comprises patients for whom prasugrel is licensed and who may be treated with the full recommended dose of prasugrel (60-mg loading dose followed by 10 mg daily). These patients have no history of stroke or TIA, are younger than 75 years and weigh more than 60 kg. The AG focuses on the clinical evidence relevant to this subgroup. The rationale for this focus is presented in Appendix 6.

The core clinical cohort42 comprised 10,804 patients (79%) from the randomised population of the TRITON-TIMI 3836 trial. The characteristics of the patients in the core clinical cohort and the overall trial population are described in Table 7. The proportions of patients quoted in Table 7 (taken from Wiviott et al. 42) are not presented by trial arm. However, Wiviott et al. 42 states that patients in the core clinical cohort randomised to prasugrel and clopidogrel were well matched and that 50% of the core clinical cohort was randomised to prasugrel. 42 The AG notes that the patients in the overall trial population and the core clinical cohort appear to be similar in terms of baseline characteristics. In TA182,21 the overall trial population of TRITON-TIMI 3836 was considered to be younger and less likely to have experienced a prior MI than patients in clinical practice in England and Wales.

Clinical efficacy in the core clinical cohort

The manufacturer submission (MS; Daiichi Sankyo Company Ltd/Eli Lilly and Company Ltd, 2013) and the Wiviott et al. 42 paper report the clinical outcomes for the core clinical cohort of patients from the TRITON-TIMI 3836 trial. It is emphasised by Wiviott et al. 42 that the core clinical cohort was identified in a post-hoc fashion defined by regulatory (EMA and the US Food and Drug Agency) criteria and should be considered as hypothesis generating.

The clinical efficacy outcomes for the core clinical cohort are presented in Table 8. For the primary composite end point of death from CV causes, non-fatal MI or non-fatal stroke, statistically significantly fewer events were recorded in the prasugrel arm (8.3%) than in the clopidogrel arm (11%) [hazard ratio (HR) = 0.74; 95% confidence interval (CI) 0.66 to 0.84; p < 0.0001]. Similarly, for the secondary composite end point (death from any cause, non-fatal MI, non-fatal stroke or non-CABG-related non-fatal TIMI major bleeding) statistically significantly fewer events were recorded in the prasugrel arm (10.2%) than in the clopidogrel arm (12.5%) (HR = 0.80; 95% CI 0.71 to 0.89; p < 0.001). The AG notes that the efficacy for both composite outcomes appears to be driven by the number of non-fatal MIs.

Statistically significant differences in favour of prasugrel were also reported for the outcomes of definite stent thrombosis (HR = 0.41, 95% CI 0.29 to 0.60; p < 0.001) and definite or probable stent thrombosis (HR = 0.44; 95% CI 0.31 to 0.62; p < 0.001). There were also statistically significantly fewer MIs in the prasugrel arm (6.7%) than in the clopidogrel arm (9.4%) (HR = 0.71, 95% CI 0.62 to 0.81; p < 0.001).

Efficacy across subgroups within the core clinical cohort

Wiviott et al. 42 present a forest plot that displays the relative effectiveness of prasugrel compared with clopidogrel across a range of subgroups within the core clinical cohort, including diagnostic group (UA/NSTEMI or STEMI), sex, age and diabetic status. The published forest plot is reproduced in Figure 2. The clinical effectiveness of prasugrel appears to be consistent across subgroups.

FIGURE 2.

Key subgroups for primary efficacy end point (core clinical cohort). BMS, bare metal stent; CrCI, creatinine clearance; DES, drug-eluting stent; DM, diabetes mellitus; GPI, glycoprotein inhibitor; RRR, relative risk reduction. a, p-value was not significant. Reprinted from Am J Cardiol, vol. 108, Wiviott SD, Desai N, Murphy SA, Musumeci G, Ragosta M, Antman EM, et al. , Efficacy and safety of intensive antiplatelet therapy with prasugrel from TRITON-TIMI 38 in a core clinical cohort defined by worldwide regulatory agencies, pp. 905–11, 2011, with permission from Elsevier.

Efficacy across time in the core clinical cohort

It is noted in Wiviott et al. 42 that, in the core clinical cohort, prasugrel was more effective than clopidogrel for the primary end point at 30 days as well as at the 15-month follow-up (Table 9).

Safety in the core clinical cohort

The key safety end point in the TRITON-TIMI 3836 trial was the rate of non-CABG-related TIMI major bleeding in the overall trial cohort at 15 months. The data for the safety end points at 15 months in the core clinical cohort are presented in Table 10. No statistically significant difference in non-CABG-related TIMI major bleeding was noted between patients in the prasugrel and clopidogrel arms; however, there was a significant difference in favour of clopidogrel when major and minor bleeding events were combined (3.0% vs. 3.9%) (HR = 1.26, 95% CI 1.02 to 1.57; p = 0.03).

Net clinical benefit

The analysis of the net clinical benefit outcome (death from any cause, non-fatal MI, non-fatal stroke or non-CABG-related non-fatal TIMI major bleeding) favoured the use of prasugrel in the core clinical cohort (12.5% in the clopidogrel group vs. 10.2% in the prasugrel group; HR 0.80, 95% CI 0.71 to 0.89; p < 0.001).

Health-related quality of life

Data relevant to HRQoL are available only for the TRITON-TIMI 3836 overall trial population and are not specific to the core clinical cohort. The HRQoL substudy was open to all TRITON-TIMI 3836 patients at participating sites in eight countries: the USA, Australia, Canada, Germany, Italy, Spain, the UK and France. HRQoL was evaluated using three instruments: (1) the Angina Frequency and Physical Limitations scales of the Seattle Angina Questionnaire; (2) the London School of Hygiene Dyspnoea Questionnaire; and (3) the European Quality of Life-5 Dimensions (EQ-5D) self-report questionnaire and the European Quality visual analogue scale. Assessments were taken at baseline and at days 30, 180, 360 and 450 (or last visit).

The HRQoL study recruited a much smaller sample than was initially planned (475 patients, compared with 3000 patients), and in TA18221 the representativeness of the substudy sample was considered to be unclear, as was the clinical utility of the results. Therefore, the AG was unable to draw any conclusions as to the HRQoL of patients treated with prasugrel or clopidogrel in the TRITON-TIMI 3836 trial. The results from the HRQoL study are presented in the MS.

Data relevant to key patient groups of the core clinical cohort

Specific clinical data relating to patients with STEMI, NSTEMI or diabetes mellitus in the core clinical cohort were not available from the MS. The AG notes from the forest plot in Figure 2 that the clinical effectiveness of prasugrel compared with clopidogrel was in evidence across the range of subgroups including STEMI, UA/NSTEMI and patients with and without diabetes. The manufacturer’s model enabled economic data pertaining to these patient groups to be extracted.

Clopidogrel loading dose: size

Timing of the clopidogrel loading dose

Clinical compared with non-clinical myocardial infarctions

Problems with an indirect comparison of the TRITON-TIMI 38

The key features of the TRITON-TIMI 3836 and PLATO33 trials are described in Table 12 (reproduced from the MS for TA236). 22 Both trials were conducted in an ACS population, use clopidogrel as a comparator and report the same primary composite efficacy end point (death from CV causes, non-fatal MI, or non-fatal stroke during the follow-up period).

Summary and critique of published indirect comparisons of prasugrel and ticagrelor

Four published indirect comparisons57–60 of prasugrel compared with ticagrelor were identified by the AG and the manufacturer during searching; the key features of these studies are described in Appendix 8. The quality of the four published indirect comparisons57–60 identified by the AG (and the manufacturer) was assessed using the assessment of multiple systematic reviews (AMSTAR)61 tool. The results are presented in Appendix 9.

The published indirect comparison of ticagrelor and prasugrel in patients with ACS conducted by Biondi-Zoccai et al. 57,62 was based on the results of the PLATO33 and TRITON-TIMI 3836 trials as well as on data from a 12-week dose-ranging trial that compared ticagrelor with clopidogrel in 990 patients with NSTEMI [Dose confIrmation Study assessing anti-Platelet Effects of AZD6140 vs. clopidogRel in non-ST segment Elevation myocardial infarction 2 (DISPERSE 2)]. 63 The total number of patients in the indirect comparison was 32,893. The results of the indirect comparison of prasugrel and ticagrelor demonstrated no statistically significant differences in overall death, non-fatal MI, non-fatal stroke, or their composite. 57 Prasugrel was associated with a significantly lower risk of stent thrombosis, and ticagrelor was associated with a significantly lower risk of any major bleeding and major bleeding associated with cardiac surgery. However, the risk of non-CABG-related major bleeding was similar for prasugrel and ticagrelor. The authors concluded that prasugrel and ticagrelor are superior to clopidogrel for ACS. The results of the indirect comparison suggest similar efficacy and safety of prasugrel compared with ticagrelor, whereas prasugrel appears more protective of stent thrombosis but causes more bleeding.

The AG’s main criticism of the indirect comparison in Biondi-Zoccai et al. 57 is that the findings are largely based on the outcomes of the TRITON-TIMI 3836 and PLATO33 trials. The substantial differences between the two trials (see Problems with an indirect comparison of TRITON-TIMI 38) render the results of the indirect comparison unreliable. The AG considers that results from the dose-ranging DISPERSE-263 trial make a negligible contribution to the results presented by Biondi-Zoccai et al. 57 as the length of follow-up was very short. The AG also notes that the published indirect comparison considered overall death (not CV death) as part of the primary composite end point.

The publication by Passaro et al. 59 presented a simplified network meta-analysis graph to improve the communicative value of the analysis undertaken by Biondi-Zoccai et al. 57 The analysis excluded the dose-ranging DISPERSE-263 trial and instead included the outcomes from the Clopidogrel in Unstable angina to prevent Recurrent Events (CURE)64 trial in which clopidogrel was compared with placebo in 12,562 patients with NSTEMI who were largely managed medically (only 21% of patients were treated with PCI). No rationale was given for the inclusion of the CURE64 trial. The AG assumes that the reason for inclusion was to enable the authors to expand the treatment network. The conclusions of this analysis concurred with those of Biondi-Zoccai et al. ,57 with the exception that no difference in major bleeding between prasugrel and ticagrelor was indicated. 59

As stated previously, the AG does not consider it appropriate to compare the results of the TRITON-TIMI 3836 and PLATO33 trials owing to their inherent differences.

The meta-analysis conducted by Chatterjee et al. 58 was intended to compare prasugrel and ticagrelor in patients with ACS or those undergoing coronary intervention for the same, or for significant coronary artery disease, by conducting a network meta-analysis. 58 Four studies, comprising a total of 34,126 patients, were included: PLATO,33 TRITON-TIMI 38,36 DISPERSE-263 and JUMBO-TIMI 26. 38 The JUMBO-TIMI 2638 trial was a dose-ranging Phase II trial comparing prasugrel with clopidogrel in 900 patients intended for PCI. The follow-up was limited to 30 days. Chatterjee et al. 58 found no difference in CV mortality or rates of MI among patients undergoing PCI but stated that CABG-related bleeding was lower with prasugrel than with ticagrelor. The authors concluded that prasugrel may be more effective than ticagrelor for preventing stent thrombosis and recurrent ischaemic events and warn that the credibility of any indirect comparison hinges on the similarity of the included trials and point to the differences in the patient populations included in the TRITON-TIMI 3836 and PLATO33 trials (randomised at presentation for PCI and randomised at presentation to the treatment centre, respectively). The authors acknowledge that this increases the likelihood of heterogeneity and recommend that a head-to-head trial of prasugrel and ticagrelor should be carried out.

The AG is of the opinion that the results of the TRITON-TIMI 3836 and PLATO33 trials have made a major contribution to the Chatterjee et al. 58 analysis and do not consider it appropriate to compare these two trials. The AG also considers that the length of follow-up of the DISPERSE-238 and JUMBO-TIMI 2638 trials was too short to provide data relevant to the current appraisal.

The work published by Steiner et al. 60 was intended to indirectly compare prasugrel, ticagrelor, high-dose clopidogrel and standard-dose clopidogrel in patients scheduled for PCI by undertaking a network meta-analysis from 14 eligible studies (48,982 patients). All studies are described in Appendix 7. The three largest studies are TRITON-TIMI 38,36 a substudy from the PLATO trial (PLATO-INVASIVE56) and CURRENT-OASIS 7 PCI. 45 These trials included patients with ACS and contributed almost 90% of patients in the analysis, whereas the other studies included stable or mixed study populations. A subgroup analysis was conducted on patients with ACS and treated with PCI using data from five studies: TRITON-TIMI 38,36 PLATO,33 CURRENT-OASIS 7,45 Han et al. 65 and DOSER. 66 This subgroup analysis corroborated the overall findings of the review which were that, for the majority of outcomes, there was no superiority of either prasugrel or ticagrelor and that prasugrel was associated with a significantly lower risk than ticagrelor for stent thrombosis but an increased risk of major or minor bleeding.

The AG is of the opinion that the overall network meta-analysis is not relevant to this review as the majority of included trials comprise stable or mixed study populations and are of short duration with primarily pharmacodynamics outcomes. The results of the ACS PCI subgroup are largely based on the comparison of the TRITON-TIMI 3836 and PLATO33 trials; the AG has previously stated this comparison to be inappropriate. The three other trials included in the subgroup analysis (CURRENT-OASIS 7,46 Han et al. 65 and DOSER66) compare high-dose clopidogrel with standard-dose clopidogrel and are of too short a duration to be of relevance to the current appraisal.

Search strategy

This review is an update of an existing review; however, searching was not date limited. In addition to searching the MS for relevant references, the following databases were searched for economic evaluations of prasugrel:

• Ovid MEDLINE(R) (1946 to August Week 3 2013)

• Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations (searched 30 August 2013)

• NHS EED (searched 30 August 2013)

• EMBASE (1974 to 30 August 2013).

The results were entered into an EndNote X5 library (and the references were deduplicated electronically). Full details of the search strategy are presented in Appendix 1.

Inclusion and exclusion criteria

At stage 1, two reviewers (ABol and SB) independently screened all titles and abstracts. Full paper manuscripts of any titles and abstracts that were considered relevant by either reviewer were obtained when possible. At stage 2, the relevance of each study was assessed (ABol and SB) according to the criteria set out in Table 13. Studies that did not meet the criteria were excluded. Any discrepancies were resolved by consensus and, when necessary, a third reviewer was consulted.

Data extraction and quality assessment strategy

In the AG’s review protocol,67 data relating to both study design and quality were planned to be extracted by two reviewers (ABol and SB) into an Excel spreadsheet (Microsoft Excel 2010; Microsoft Corporation, Redmond, WA, USA). It was also planned that all economic evaluations identified for inclusion in the review would be quality assessed according to the Drummond et al. 68 10-point checklist. However, no studies were identified for inclusion in the AG’s review.

Results: quantity and quality of research available

After deduplication of 1449 references, a total of 1230 titles and abstracts were screened for inclusion at stage 1. Of these 1230 references, 1117 were immediately excluded because they did not include prasugrel as an intervention or a comparator. At stage 2, inclusion criteria were applied to 113 references. During stage 2, 98 references were excluded, leaving a possible 15 references available for potential inclusion and these are listed Table 14. Of the 15 potentially eligible references, none of the papers met the full inclusion criteria that were set by the AG.

The review carried out by the AG picked up the three studies69,72,73 that the manufacturer had identified for inclusion in the review of cost-effectiveness evidence presented in the MS. Two of these studies69,73 were carried out from a US perspective and the third study72 employed the model that was submitted to NICE for the evaluation of prasugrel in 2009 (TA18221); all three studies69,72,73 were therefore excluded from the review by the AG.

Conclusions of the Assessment Group’s cost-effectiveness literature review

The AG did not identify any published papers that met the inclusion criteria for the review.

Overview of manufacturer’s submitted model

Table 16 describes NICE’s reference case checklist and provides the manufacturer’s assessment of how the submitted economic model matches NICE’s checklist.

In summary, the manufacturers have submitted the same economic model that they previously presented during the original appraisal of prasugrel for the treatment of ACS with PCI (TA182). 21 However, some aspects of the submitted model have been updated in the light of feedback generated during the original appraisal of prasugrel (TA182). 21 These revised aspects are:

• use of sensitivity analysis encompassing the entire population as opposed to a ‘typical’ patient profile

• removal of the functionality that allowed the user to choose to model 15 months of treatment (as the licence is only for 12 months)

• conduct of scenario analysis using the ERG’s suggestions for utility values, amended long-term relative risk (RR) of mortality and reduced incidence of non-fatal MI

• use of the generic (reduced) price of clopidogrel

• updated costs.

The model was developed with the principle of simulating the TRITON-TIMI 3836 trial outcomes as closely as possible. There are two main phases to the model: the active treatment phase, which spans the duration of the clinical trial, and the post-treatment phase, which extrapolates outcomes and costs beyond events that took place during the treatment phase, up until death or lifetime horizon (base case 40 years). Within the trial period, there is an opening 3-day period, modelled using a decision tree, followed by 12 cycles, each of 1 month, up to 12 months. The transitions were time dependent. Long-term mortality was based on adjustment of population life tables to reflect prognostic implications of the events modelled over the short term. The model also permits some costs to accumulate after the end of the trial period.

Figure 3 shows the state transition diagram for the Markov model element of the TRITON-TIMI 3836 study. Patients enter the model at the point of the index ACS event, immediately prior to undergoing PCI. Exit occurs at death, or at completion of the model time horizon.

FIGURE 3.

Schema of manufacturer’s model. Note: the lightly dashed lines leading to ‘bleed end point event’ are intended to highlight that these do not represent transitions to health states that continue to impart prognostic effects in terms of long-term mortality, or permanent utility decrements. Patients remain in their origin states following bleed events, except when the event is fatal. Temporary utility decrements are applied at the time of major non-fatal bleeds. Rehospitalisation occurs in all states at rates determined by current and past clinical events. MI, myocardial infarction. Reproduced from MS.

Parameters and values

The parameters and values used in the economic model are displayed in Table 17.

Sources of evidence used to inform and develop the model

The TRITON-TIMI 3836 trial was the key source of clinical evidence described in the MS. Non-trial sources of clinical evidence were also identified via literature reviews to inform assumptions regarding additional clinical inputs, long-term extrapolation of mortality and HRQoL.

Population

The populations described in the economic model reflect the patients enrolled in TRITON-TIMI 3836 (details presented in Table 20).

Interventions and comparators

The economic evaluation compares prasugrel in combination with aspirin and clopidogrel in combination with aspirin, at licensed doses. Consistent with both the TRITON-TIMI 3836 trial and the SPC,24 prasugrel is initiated with a single 60-mg loading dose and then continued at 10 mg once a day for up to 12 months in combination with aspirin (75–325 mg). Clopidogrel was initiated with a single 300-mg loading dose and then continued at 75 mg once a day in combination with aspirin for 12 months.

The manufacturer considered that a formal indirect comparison between prasugrel and ticagrelor was inappropriate and no economic analysis of this comparison has been presented in the MS.

Perspective, time horizon and discounting

The perspective for outcomes reflects all the direct health effects, whereas the perspective used for costs is that of the NHS. Outcomes are expressed in terms of life-years and quality-adjusted life-years (QALYs) gained. The time horizon is set at 40 years and, in line with the NICE Guide to the Methods of Technology Appraisal,90 both costs and benefits are discounted at 3.5%. A half-cycle adjustment was performed for both costs and outcomes (attributing events on the basis of average patient exposure over the course of each cycle).

Health-related quality of life

Although the TRITON-TIMI 3836 trial included a HRQoL substudy, the manufacturer reports that it was not possible to provide robust HRQoL estimates owing to the very small numbers of patients with events included within the analysis. The manufacturer, therefore, conducted a systematic review of the literature to identify HRQoL studies relevant to the modelled trial population. The MS (p. 102) includes details of the methods used in the systematic review. Mean utility decrements for ACS (0.049) and stroke/MI (0.052) were taken directly from a US study,87 which was designed to produce a specific list of preference weights for use in economic evaluations; the study used the US version of the EQ-5D.

To calculate utility weights for use in the economic evaluation, background UK population norms (free of disease) which vary by age and sex, as described by Kind et al. ,91 were applied to all patients in the trial. The utility decrements for ACS and stroke/MI were then used alongside these background utility estimates. Finally, the MS assumed that, for a major bleed, a decrement of 25% of the population (utility) norm was applicable for a 14-day period (25% decrement equates to a 0.007 utility toll).

Resources and costs

The key categories of cost estimates in the MS are related to (1) hospitalisations and (2) drug costs. Key cost parameter assumptions are presented in Table 21.

Cost-effectiveness results

Five different subgroups are considered, namely (1) the whole ACS licensed population (excluding prior stroke/TIA), (2) the ACS core population (excluding those with prior stroke/TIA and patients weighing < 60 kg or aged ≥ 75 years), (3) the UA/NSTEMI licensed population (excluding those with prior stroke/TIA), (4) the STEMI licensed population (excluding those with prior stroke or TIA) and (5) the ACS–diabetes licensed population (excluding those with prior stroke or TIA). The base case ICERs generated by the manufacturer’s model for these five subgroups are presented in Table 24.

Sensitivity analyses

A probabilistic sensitivity analysis was not undertaken. Univariate (one-way) sensitivity analysis was conducted by the manufacturer for selected model parameters, namely discounting, haemorrhage utility decrement, MI and stroke utility decrements, hospitalisation episodes, treatment duration, RR for all-cause mortality (post-trial phase) and time horizon. The results of the one-way sensitivity analysis are shown in Table 25.

Critique of submitted economic model

The AG’s critique of the manufacturer’s submitted economic model is the same as the original critique presented by the ERG during the original appraisal of prasugrel (TA182). 21 The AG and the ERG are the same academic research group.

As outlined in section 8.2.4.1 of the MS, at the time of the original appraisal, the ERG suggested amendments to the manufacturer’s economic model in the following six main areas:

• life table calculations, which need to allow for competing risks

• differences in discounting approaches

• treatment costs, which should reflect usage and pack wastage

• alternative utility values (i.e. those derived from the HODAR database)

• reduced incidence of non-fatal MIs such that the underlying rate of MIs is 50% that recorded in the TRITON-TIMI36 trial

• amended long-term RRs of mortality by ignoring the initial impact of ACS prior to TRITON-TIMI-related events (i.e. ignoring the sources from Rosengren et al. 83).

The AG agrees with the manufacturer that the first three points mentioned above lead to non-significant changes in the size of the ICER. The manufacturer carried out a scenario analysis to determine the effect of the remaining three amendments suggested by the ERG.

The impact of this scenario analysis on the results for the relevant subgroups is presented in Table 26.

The results of the manufacturer’s scenario analysis show that, when comparing prasugrel with clopidogrel, all relevant ICERs remained within the £20,000 to £30,000 per QALY gained threshold.

However, the AG is of the opinion that the basic structure of the manufacturer’s economic model still requires further refinement. The main focus of the AG’s critique is the manufacturer’s projection of long-term survival. The AG’s specific concerns are outlined in detail in Independent economic assessment: results and Independent economic assessment: discussion.

In summary, the AG developed its own economic model for the following reasons:

• The long-term model phase in the manufacturer’s submitted economic model was considered to be unsatisfactory and potentially not sufficiently reliable to generate a realistic representation of 39-years of follow-up.

• The manufacturer’s decision model projects long-term (years 2–40) costs and outcomes solely in terms of mortality hazard rates fixed after 1 year, and takes no account of the effects of accumulating experience of CV events and disability.

• The AG considered it appropriate to develop an economic model using the most reliable clinical evidence available and, therefore, preferred to use 3-year clinical data from the CAPRIE92 (Clopidogrel versus Aspirin in Patients at Risk of Ischaemic Events) trial instead of 15-month data from the TRITON-TIMI 3836 trial.

• To fulfil the remit stated by NICE and to review fully the guidance for prasugrel issued in TA182,21 the AG was required to compare four patient subgroups (STEMI without diabetes mellitus, STEMI with diabetes mellitus, NSTEMI without diabetes mellitus and NSTEMI with diabetes mellitus). The structure of the decision model submitted by the manufacturer does not readily facilitate modelling these four subgroups in terms of cost-effectiveness.

Background and modelling rationale

The manufacturer of prasugrel has chosen to resubmit the same decision model previously employed for the NICE single technology appraisal (STA) of prasugrel in 2009. 21 This model comprised two distinct phases:

• a short-term statistical model of the data from the TRITON-TIMI 3836 clinical trial (up to 15 months follow-up)

• a long-term model projecting survival and hospitalisation of patients alive at the end of the first phase up to a maximum of 40 years.

In the ERG’s report prepared as part of the STA process, particular concern was expressed about the structure of this model. The ERG concluded that the initial phase of the model generated reliable outcome estimates:

Comparison of the mortality rate (all causes) obtained by Kaplan–Meier analysis of TRITON-TIMI 38 data (supplied by the manufacturer) with corresponding rates generated by the model at 30 days and 12 months indicate a good correspondence for treatment with clopidogrel and with prasugrel for all specified populations.

Greenhalgh et al. 2009, section 5.5.293

However, the long-term model phase was considered by the ERG to be less satisfactory and potentially not sufficiently reliable to generate a realistic representation of a further 39 years of follow-up:

In the long-term component of the submitted model there is an assumption that differences established between the prasugrel and clopidogrel arms of the TRITON-TIMI 38 trial will be preserved indefinitely at the level observed at the end of the trial. However, there is no reason to believe that further serious nonfatal events will not continue to occur to patients in both cohorts, and if events occurring during the trial are presumed to influence later survival, then it is also likely that any such events in subsequent periods will also have important effects. Since active treatment with clopidogrel or prasugrel will have ceased, it can be expected that event rates will be similar in both arms. As a result of this process it is likely that over time the disease history of patients will converge, and therefore any initial advantage for either treatment will be progressively attenuated. This effect would have become evident in the model results if the long-term model had been structured to reflect changes in health states over time.

Greenhalgh et al. 2009, section 5.5.393

As these serious concerns have not been addressed by the manufacturer in the model submitted for this reappraisal of prasugrel, the AG has developed a new decision model. The AG’s model accepts the manufacturer’s statistical model for the initial phase (up to 12 months), but replaces the long-term projection with a more detailed structure that provides an improved representation of subsequent CV events, accumulating patient histories, alteration in health states and associated care costs, as well as patient HRQoL.

Patient populations

The AG has structured its decision model to accommodate four mutually exclusive subgroups of the core clinical cohort population (i.e. all ACS patients excluding those with a history of TIA or stroke, those with body weight of < 60 kg or those aged > 75 years):

• ACS patients treated with PCI for STEMI and with diagnosed diabetes

• ACS patients treated with PCI for STEMI and without diagnosed diabetes

• ACS patients treated with PCI for UA or NSTEMI and with diagnosed diabetes

• ACS patients treated with PCI for UA or NSTEMI and without diagnosed diabetes.

These were the groups considered by the ERG to be important in the development of the final 2009 guidance related to prasugrel (TA18221) and they therefore form an appropriate basis for this review of the existing guidance.

Treatment options

No suitable clinical evidence has been identified that can provide the basis for a reliable comparison between prasugrel and ticagrelor. The AG model, therefore, has been developed as a simple comparison between dual antiplatelet therapy for 12 months from index PCI with either clopidogrel in combination with low-dose aspirin or prasugrel in combination with low-dose aspirin.

Model design and structure

The AG for this review also acted as AG for the reappraisal of clopidogrel and modified-release dipyridamole for the prevention of occlusive vascular events. That reappraisal was an update of NICE guidance TA9094 and resulted in the publication of TA210,95 which was issued in December 2010. In TA210,95 NICE made recommendations concerning the use of clopidogrel and modified-release dipyridamole for the prevention of occlusive vascular events. As part of the TA21095 guidance development process, the AG developed a detailed decision model to estimate the long-term health care and outcomes expected for patients receiving different strategies of long-term preventative treatment. The model took the form of an individual patient simulation. It was calibrated mainly using data provided by the manufacturer of clopidogrel from the CAPRIE92 clinical trial, supplemented with data provided by the manufacturer of dipyridamole from the PROFESS96 clinical trial and some additional published trial results. The additional data included follow-up results for 3 years from the start of preventative therapy. Supplementary details are provided in Appendix 6.

The AG has concluded that the MI subpopulation model used in the development of TA21095 (the TA21095 model), which was based largely on CAPRIE92 trial data, addresses very similar issues to those that are of concern to this review of TA182. 21 The AG’s clinical advisor has confirmed that CAPRIE92 data are an appropriate trial source for extrapolating long-term vascular events and that no better source has become available since 2010.

However, there is a significant practical drawback to using the individual patient simulation approach that was employed in the TA21095 model, namely the extended run times involved in generating model results, especially when carrying out probabilistic sensitivity analyses. The AG has therefore re-engineered the TA21095 model, and the current AG model for prasugrel employs a long-term Markov chain, which operates for up to 39 years of follow-up beyond the first 12 months of treatment with clopidogrel or prasugrel. This re-engineering has necessitated some compromises to the fully flexible logic of the TA21095 model, which allowed each patient to experience any number of occlusive vascular events at any time in any year. However, the frequency of these events is low, and restricting the Markov model to 12-month cycles and allowing only one event per cycle is unlikely to have a noticeable effect on the evaluation of treatments. In theory, the number of events per patient may be marginally understated, along with the related treatment costs and disutilities; however, as these apply in the same way to both arms of the evaluation, the impact on the assessment of comparative cost-effectiveness is believed to be negligible.

The annual transition matrix for the AG model is shown in Table 27. The matrix shows how the health state of a patient is altered depending on the type of vascular event suffered during the year and the most severe previous event experienced, including whether or not the patient had suffered a severely disabling stroke (modified Rankin Scale score 3–5).

Patients enter the long-term model with the average number of vascular events experienced in the first 12 months following the index PCI event, estimated by the manufacturer’s short-term statistical model, apportioned between the first four states [None, MI(1)ND, Stroke(1)ND and Stroke(1)D] (see Table 27). The model then traces the long-term accumulating event history separately for males and females within each of the four subpopulations, using sex-specific parameter values (Table 28).

Assessment of uncertainty

A univariate sensitivity analysis has been performed on all model variables subject to uncertainty, and results are presented in the form of ‘torpedo’ diagrams ranking the 20 variables subject to greatest uncertainty in terms of influence on the deterministic estimated ICER per QALY gained for prasugrel compared with clopidogrel, as measured after 40 years’ follow-up.

A probabilistic sensitivity analysis has been carried out, using 1000 simulations and employing a standardised set of random variables selected to ensure full coverage of the uncertainty domain (sometimes referred to as orthogonal sampling), and incorporating correlated random variables as necessary.

Parameter sources and values

All the parameter values used in the Markov model for event incidence risk (Table 29), event fatality rates (Table 30) and RRs remain unchanged from those previously described in the AG report for the development of NICE guidance TA 210,95 with the exception of the RR applying to patients with/without diabetes (Table 31).

Cost of medication

The cost of dual antiplatelet therapy in the first year, and the cost of continuing low-dose aspirin thereafter, is detailed in Table 32. Both clopidogrel and prasugrel usage has been adjusted to reflect actual usage in the clinical trial. The cost of a loading dose of 300-mg clopidogrel or 60-mg prasugrel is included.

Resource use estimation

Health-care costs and health-related utility values are applied for both time spent in each health state and as discrete single-event costs and disutilities.

Discounting costs and outcomes

Both costs and outcomes were discounted annually at 3.5%. Univariate sensitivity analyses were carried out using discount rates of 0% and 6% for both costs and outcomes.

Time horizon

The model generates results annually at the end of each year from trial randomisation. However, deterministic results are reported at 1, 5, 10, 20 and 40 years, and probabilistic results at 5 and 40 years.

Key modelling assumptions

Validation and quality assurance

The AG’s long-term model has been cross-matched against the original individual patient model to ensure all formulae have been correctly implemented. In addition, check totals have been incorporated into each annual application to ensure that any discrepancies in patient totals, health state totals and event totals are readily identifiable. The starting values for the long-term model have been matched to the manufacturer’s model at 12 months for accuracy.

ST segment elevation myocardial infarction: diabetes subgroup

Deterministic results are detailed in Table 35 (life-years), Table 36 (QALYs), Table 37 (costs) and Table 38 (ICERs). The ICER at the end of the first year is high, owing to the inclusion of the full additional cost of treatment with prasugrel, while only modest health gains have accrued from the reduced incidence of MIs. Over time the estimated ICER decreases steadily, suggesting that incremental benefit continues to accrue over subsequent decades while incremental cost increases at a slower rate. The ICER for prasugrel compared with clopidogrel falls below £30,000 per QALY gained after 5 years.

Figure 4 displays the results of univariate sensitivity analysis, indicating that uncertainty from individual model parameters has a modest influence on the magnitude of the ICER in this subgroup: the discount rates for costs and outcomes cause the largest changes, but the ICER remains within the range £1000 to £2500 per QALY gained.

FIGURE 4.

Univariate sensitivity analysis: 20 most important parameters in determining the ICER for STEMI patients with diabetes.

Probabilistic analysis at the 40-year follow-up horizon for this subgroup yields a higher estimated ICER (£1732 per QALY gained) derived from very small incremental cost and QALY estimates (+£515 and +0.297, respectively). The scatterplot (Figure 5) and CEAC for this subgroup (Figure 6) indicate the relative cost-effectiveness of prasugrel despite the long-term erosion of incremental differences over time.

FIGURE 5.

Probabilistic sensitivity analysis scatterplot of prasugrel compared with clopidogrel for STEMI patients with diabetes.

FIGURE 6.

Cost-effectiveness acceptability curve of prasugrel compared with clopidogrel for STEMI patients with diabetes.

ST segment elevation myocardial infarction: no diabetes subgroup

Deterministic results are detailed in Table 39 (life-years), Table 40 (QALYs), Table 41 (costs) and Table 42 (ICERs). The ICER at the end of the first year is high, owing to the inclusion of the full additional cost of treatment with prasugrel, whereas only modest health gains have accrued from the reduced incidence of MIs. Over time the estimated ICER decreases steadily, suggesting that incremental benefit continues to accrue over subsequent decades while incremental cost increases at a slower rate. The ICER for prasugrel compared with clopidogrel falls below £30,000 per QALY gained at 10 years.

Figure 7 displays the results of univariate sensitivity analyses, which indicate that uncertainty from the discounting rate for outcomes has the largest impact on the estimated ICER (ranging between £4000 and £9000 per QALY gained). Other individual model parameters have only a modest influence on the magnitude of the ICER in this subgroup.

FIGURE 7.

Univariate sensitivity analysis: 20 most important parameters in determining the ICER for STEMI patients without diabetes.

Probabilistic analysis at the 40-year follow-up horizon for this subgroup yields a higher estimated ICER (£7073 per QALY gained) derived from small incremental cost and QALY estimates (+£609 and +0.086, respectively). The scatterplot (Figure 8) and CEAC for this subgroup (Figure 9) indicate the relative cost-effectiveness of prasugrel despite the long-term erosion of incremental differences over time.

FIGURE 8.

Probabilistic sensitivity analysis scatterplot of prasugrel compared with clopidogrel for STEMI patients without diabetes.

FIGURE 9.

Cost-effectiveness acceptability curve of prasugrel compared with clopidogrel for STEMI patients without diabetes.

Unstable angina/non-ST segment elevation myocardial infarction: diabetes subgroup

Deterministic results are detailed in Table 43 (life-years), Table 44 (QALYs), Table 45 (costs) and Table 46 (ICERs). The ICER at the end of the first year is high, owing to the inclusion of the full additional cost of treatment with prasugrel, whereas only modest health gains have accrued from the reduced incidence of MIs. Over time, the estimated ICER decreases steadily, suggesting that incremental benefit continues to accrue over subsequent decades, while incremental cost increases at a slower rate. The ICER for prasugrel compared with clopidogrel falls below £30,000 per QALY gained after 5 years.

Figure 10 displays the results of univariate sensitivity analyses, which indicate that uncertainty from event incidence and fatality rates have the largest effect on the estimated ICER (ranging between –£1000 and £400 per QALY gained). Other individual model parameters have only a modest influence on the magnitude of the ICER in this subgroup.

FIGURE 10.

Univariate sensitivity analysis: 20 most important parameters in determining the ICER for UA/NSTEMI patients with diabetes.

Probabilistic analysis at the 40-year follow-up horizon for this subgroup confirms that prasugrel dominates clopidogrel with a small net cost saving and positive incremental benefit (–£120 and +0.191, respectively). The scatterplot (Figure 11) and CEAC for this subgroup (Figure 12) indicate the cost-effectiveness of prasugrel despite the long-term erosion of incremental differences over time.

FIGURE 11.

Probabilistic sensitivity analysis scatterplot of prasugrel compared with clopidogrel for UA/NSTEMI patients with diabetes.

FIGURE 12.

Cost-effectiveness acceptability curve of prasugrel compared with clopidogrel for UA/NSTEMI patients with diabetes.

Unstable angina/non-ST segment elevation myocardial infarction: no diabetes subgroup

Deterministic results are detailed in Table 47 (life-years), Table 48 (QALYs), Table 49 (costs) and Table 50 (ICERs). The ICER at the end of the first year is high, owing to the inclusion of the full additional cost of treatment with prasugrel, whereas only modest health gains have accrued from the reduced incidence of MIs. Over time the estimated ICER decreases steadily, suggesting that incremental benefit continues to accrue over subsequent decades while incremental cost increases at a slower rate. The ICER for prasugrel compared with clopidogrel falls below £30,000 per QALY gained after 10 years.

Figure 13 displays the results of univariate sensitivity analyses, which indicate that uncertainty from discounting rates, and event incidence and fatality rates have the largest effect on the estimated ICER (ranging between £2500 and £6500 per QALY gained). Other individual model parameters have only a modest influence on the magnitude of the ICER in this subgroup.

FIGURE 13.

Univariate sensitivity analysis: 20 most important parameters in determining the ICER for UA/NSTEMI patients without diabetes.

Probabilistic analysis at the 40-year follow-up horizon for this subgroup yields a lower estimated ICER of £4154 per QALY gained, derived from small incremental cost and QALY estimates (+£212 and +0.051, respectively). The scatterplot (Figure 14) and CEAC for this subgroup (Figure 15) indicate the relative cost-effectiveness of prasugrel despite the long-term erosion of incremental differences over time.

FIGURE 14.

Probabilistic sensitivity analysis scatterplot of prasugrel compared with clopidogrel for UA/NSTEMI patients without diabetes.

FIGURE 15.

Cost-effectiveness acceptability curve of prasugrel compared with clopidogrel for UA/NSTEMI patients without diabetes.

Clinical effectiveness

This review focused on the health outcomes of the subgroup of patients discussed in TA18221 and for whom the full dose of prasugrel is licensed, namely the core clinical cohort (i.e. patients without a history of TIA or stroke, those with body weight of < 60 kg or those aged > 75 years). This group of patients constituted 79% of the overall population of TRITON-TIMI 38. 36 In the core clinical cohort, all non-bleeding clinical outcomes of the TRITON-TIMI 3836 trial favoured the use of prasugrel compared with clopidogrel. These findings held over time and across subgroups of patients, including those with STEMI and UA/NSTEMI. There was a statistically significant difference in event rates in favour of clopidogrel when major and minor bleeding rates were combined.

A clinical comparison of prasugrel with ticagrelor was not carried out by the AG (or the manufacturer of prasugrel). There were two reasons for this. First, there was no direct RCT evidence comparing prasugrel with ticagrelor and, second, it was not possible to conduct an indirect comparison as there were irreconcilable differences between the two pivotal trials33,36 (including timing and dosing of clopidogrel and assessment of MI). Thus, the comparative effectiveness and safety of prasugrel compared with ticagrelor still remain unknown.

Cost-effectiveness

In the AG’s independent economic model, the outcomes of the TRITON-TIMI 3836 trial population were simulated as four mutually exclusive subgroups: STEMI without diabetes mellitus, STEMI with diabetes mellitus, NSTEMI without diabetes mellitus and NSTEMI with diabetes mellitus. This approach has allowed the AG to reconsider the strength of evidence underlying the previous NICE guidance21 that excluded patients from treatment with prasugrel if they had not suffered a STEMI event, or had not been diagnosed with diabetes. The new model confirmed that, using a £20,000 to £30,000 per QALY gained threshold, within 5–10 years, it appears likely that prasugrel is a cost-effective treatment option when compared with clopidogrel for all four subgroups.

Strengths and limitations of the assessment

The main strength of this review is that, despite some remaining areas of uncertainty, the case for prasugrel compared with clopidogrel appears to have been strengthened. The results of the AG’s independent economic model confirm the cost-effectiveness of prasugrel compared with clopidogrel, at a threshold of £20,000 to £30,000 per QALY gained, for key groups of patients with ACS who are to be treated with PCI. The structure of the AG’s model differs from the model developed by the manufacturer in that it uses the most up-to-date clinical evidence available (from the CAPRIE92 trial) and compares four patient subgroups (STEMI without diabetes mellitus, STEMI with diabetes mellitus, NSTEMI without diabetes mellitus and NSTEMI with diabetes mellitus). A particular strength of the AG’s economic model is that is provides assessments at specific time periods within the modelled time horizon of 40 years.

Both the AG and the manufacturer demonstrate the cost-effectiveness of prasugrel compared with clopidogrel at a threshold of £20,000 to £30,000 per QALY gained. However, the AG acknowledges that any long-term modelling exercise is vulnerable to major assumptions about the continuation of early health outcome gains, and it is noted that both the manufacturer’s and the AG’s models rely on extrapolating relatively short-term results from beyond the end of the trial to a further 40 years.

Since TA182,21 the patent for clopidogrel has expired. In TA182,21 the assessment of the cost-effectiveness of prasugrel was based on the non-generic price of clopidogrel using the economic model submitted by the manufacturer of prasugrel. A key strength of this update is that the AG has been able to reassess the cost-effectiveness of prasugrel compared with clopidogrel using the generic price of clopidogrel in an independent economic model.

The clinical effectiveness and cost-effectiveness findings of the report are limited by the nature of the available clinical evidence. Since TA182,21 no new clinical evidence has become available to support the use of prasugrel compared with clopidogrel. In the short-term, all clinical effectiveness data used in the model were derived from a single RCT (TRITON-TIMI 38). 36 In the longer term, all clinical effectiveness data used in the model were primarily derived from a single RCT (CAPRIE). 92 The AG notes that both RCTs recruited large numbers of patients and were well conducted and well reported.

The AG notes that, although the TRITON-TIMI-3836 trial was considered to be of a robust design, the majority (93%) of the trial population was white Caucasian. This does not negate the findings of this report, but it must be considered as a limitation to the applicability of the recommendations.

Uncertainties

The three areas of uncertainty noted by the AC for TA18221 were reconsidered in this review. These centred on the generalisability of the TRITON-TIMI 3836 trial results to patients in clinical practice in the UK. The AG is of the opinion that the clinical evidence for the equivalence of a 300-mg loading dose of clopidogrel (administered in TRITON-TIMI 3836) with a 600-mg loading dose (often given in clinical practice in the UK) remains uncertain. Similarly, the importance of timing of the administration of the loading dose of clopidogrel on patient outcomes remains unresolved and differs between the TRITON-TIMI 3836 trial and clinical practice in the NHS in England and Wales. The AG considers that the case for the clinical effectiveness of prasugrel compared with clopidogrel in preventing MIs of all types and sizes appears to be robust.

Part of the remit for this review was to consider the efficacy of prasugrel compared with ticagrelor for patients with ACS who are to be treated with PCI. As no head-to-head trial has been conducted comparing these two treatments, the AG considered the possibility of an indirect treatment comparison using data from the TRITON-TIMI 3836 and PLATO33 trials; however, the AG concluded that the key differences between the two trials made any comparison unreliable. Thus, the comparative clinical effectiveness and safety of prasugrel compared with ticagrelor remains unknown. However, the AG is aware of a RCT that commenced recruiting patients in September 2013. 103,104 The Intracoronary Stenting and Antithrombotic Regimen: Rapid Early Action for Coronary Treatment (ISAR-REACT) 5103,104 trial is designed to assess whether or not ticagrelor is superior to prasugrel in patients with ACS and planned invasive strategy. The primary outcome is the composite of death, MI or stroke at 12 months in a planned patient population of 4000. The results of the ISAR-REACT 5103,104 trial will allow a formal comparison of the efficacy of prasugrel compared with ticagrelor.

1946 to June Week 1 2013

1974 to 2013 June 18