3.1.1. Clinical introduction

Aspirin was the first anti–platelet agent to be investigated and has been prescribed for many years for patients at risk of vascular ‘events’ such as heart attacks (myocardial infarction, hereafter referred to as MI) and strokes. It blocks cyclooxygenase, an enzyme involved in the pathway of prostaglandin and thromboxane synthesis, agents which are highly vasoactive and prothrombotic. Platelets do not synthesise new cyclooxygenase once exposed to aspirin and so its effect persists for the life of each inhibited platelet.

Given the widespread acceptance of the use of aspirin in current practice the GDG limited the evidence search to systematic reviews to determine the evidence for people with UA or NSTEMI and asked the following clinical question:

‘What is the efficacy and safety of aspirin therapy in the medical management of patients with UA or NSTEMI compared to placebo?’

3.1.2. Clinical methodological introduction

The Cochrane database was searched from 1995 to 2009 for systematic reviews comparing aspirin with placebo in the management of people with non ST-segment elevation ACS. For a review to be included, it had to be specific to the non ST segment elevation ACS population (it had to contain > 60% unstable angina/NSTEMI). Studies were included if they reported death, MI, bleeding, stroke, re-revascularisation, left ventricular function, and quality of life.

One well-conducted systematic review compared anti–platelet therapy with placebo in a large group of people at high risk of occlusive arterial disease (195 RCTs; N=135,640). The risk of vascular events (defined as nonfatal MI, nonfatal stroke, or death from a vascular cause or death from an unknown cause) in a sub-population of people with UA was compared in those receiving anti–platelet agents (predominantly aspirin) and those receiving placebo ⁵⁰.

3.1.3. Clinical evidence statements

3.1.4. Health economic methodological introduction

One relevant study was identified. This was a cost-effectiveness analysis that evaluated aspirin versus no aspirin use in UA patients ⁵¹.

Fidan et al.⁵¹ reported a cost–effectiveness analysis from a UK NHS perspective. It incorporated the cost of aspirin and life-years gained with treatment to estimate cost effectiveness in terms of cost per life-year gained. Aspirin costs were based on doses from clinical trials and national UK costs (2000). A mortality model (the IMPACT model⁵²) was used to estimate deaths prevented/postponed with aspirin treatment in UA over one year and median survival estimates were then applied to extrapolate this impact in terms of life-years gained. The IMPACT mortality model was based on CHD patient numbers, uptake of treatment, median survival in people with and without CHD developed using data from sources describing England and Wales 2000. The effectiveness of aspirin was based on a meta–analysis by the Antithrombotic Trialists' Collaboration (2002)⁵⁰. Results were presented overall and for ten-year age bands.

The study is judged directly applicable to the UK NHS. The key potential limitation of the study is that it only incorporates the cost of aspirin - other relevant events would have cost implications (such as MIs avoided). In addition, the incorporation of treatment-related costs for the full time horizon is recommended NICE methodology and is not included. Other minor limitations include the unclear reporting of methods regarding the cost calculations – it is unclear if aspirin use is specifically acute use or continued for the whole year – and the lack of incorporation of quality of life (to estimate QALYs).

3.1.5. Health economic evidence statements

Fidan et al.⁵¹ reported an incremental cost–effectiveness ratio (ICER) of £58 per life year gained for aspirin use compared to no aspirin use. ICERs in different ten-year age bands ranged between £42 and £85 per life-year gained. Sensitivity analysis was carried out where ICERs were recalculated using minimum and maximum estimates for cost of aspirin, efficacy of aspirin and life-years gained and ranged between £34 and £114 per life year gained.

The lack of inclusion of costs other than the cost of aspirin could potentially be a serious limitation. It is not possible to judge exactly how their inclusion would impact results although it would probably increase some costs (such as bleed costs) while decreasing other cost (such as MI costs due to a reduction in events with aspirin). Nevertheless, as the estimated ICERs are so low it is judged likely that aspirin would remain cost effective if additional costs were incorporated. Incorporation of quality of life is also judged unlikely to change conclusions about cost effectiveness.

3.1.6. Evidence summary

The meta-analysis involving 197 RCTs with over 135,000 patients randomised to receive an anti–platelet agent versus placebo was accepted as the sole source for review.

The risk of vascular events was considered for various ‘at-risk’ groups (such as those with coronary artery disease, stroke, or peripheral arterial disease) and for sub-populations such as those with MI, UA, stable angina, and those undergoing coronary revascularisation (angioplasty or coronary bypass grafting). Of the trials analysed aspirin was the predominant anti–platelet agent given.

One of the sub-groups analysed was those with UA but because of the more recently changed definition of MI⁴⁹ many of the patients in this previous category will have been those who would currently be classified as having NSTEMI. The GDG were therefore unable to separate those who would currently be regarded as having UA from those with NSTEMI, but in practice this is of little importance because the investigators demonstrated that anti–platelet therapy significantly reduced the number of vascular events in all the relevant coronary disease sub-groups (acute MI, UA, stable angina; range of odds reduction 25 to 46%). The group classified by the previous definition as having UA (n=5031) had a 46% odds reduction of having a vascular event during the follow-up period, which varied between trials (6 days to 18 months).

3.1.7. Evidence to recommendations

The GDG concluded that aspirin therapy reduces the risk of a vascular event and should be offered to all patients with UA or NSTEMI unless contraindicated (such as by active bleeding, current peptic ulceration, or for those considered clinically to be at a high potential risk of the consequences of bleeding, for example, recent neurosurgery or haemorrhagic stroke)⁵³. It should be noted that those at higher risk of bleeding, such as those with renal impairment, may have a higher absolute risk of a vascular event and therefore may have a higher potential absolute benefit from aspirin, which may outweigh even the higher bleeding risk associated with their underlying renal impairment. Individual patient circumstances will dictate the advisability of giving aspirin but in only a small minority would it be anticipated that the risk of prescription will outweigh the benefit.

Use of anti–platelet agents has also been associated with about a twofold increase in the rate of major bleeding, but because the background rate of bleeding was low this increased risk was far outweighed by the longer term benefit of anti–platelet treatment, a finding also supported by others⁵⁴. No additional longer term benefit was found from maintenance doses of aspirin higher than 75-150 mg, though the Trialists recommended a loading dose of 150-300 mg in clinical situations where an immediate antithrombotic effect is required “such as MI….and UA”.

An aspirin loading dose of 300 mg should be given as soon as possible followed by daily maintenance of 75-150 mg. The use of other anti–platelet agents, such as clopidogrel and the GPIs are considered in this guideline but would normally be given on the background of regular aspirin therapy except where aspirin is considered contraindicated.

3.1.8. Recommendations

R6.

Offer aspirin as soon as possible to all patients and continue indefinitely unless contraindicated by bleeding risk or aspirin hypersensitivity.

R7.

Offer patients a single loading dose of 300 mg aspirin as soon as possible unless there is clear evidence that they are allergic to it.

R8.

For patients with aspirin hypersensitivity, clopidogrel monotherapy should be considered as an alternative treatment. (This recommendation is from ‘MI: secondary prevention’, NICE clinical guideline 48.)

3.2.1. Clinical introduction

Clopidogrel was the subject of a NICE TA (TA80) published in July 2004. This made three recommendations:

• the use of clopidogrel with aspirin in the management of NSTEMI considered to be at high or medium risk of MI or death
• the relevance of assessing risk in such patients,
• duration of treatment.

Only the first two recommendations from this TA are pertinent to the scope addressed by this guideline and will be updated in this guidance.

Clopidogrel is an anti–platelet agent and part of the thienopyridine group that block platelets by inhibition of the adenosine diphosphate (ADP) pathway. Clopidogrel has been investigated for its potential to decrease the risk of an adverse cardiovascular outcome in patients with ACS, for reasons which are similar to those described earlier with respect to aspirin therapy. The data reviewed in this chapter refers to clopidogrel hydrogen sulphate; we have not addressed whether other, more recently introduced, clopidogrel salts are equivalent.

Prasugrel^55-57 is an anti–platelet similar to, though with various features different from clopidogrel, but the subject of a separate NICE Appraisal (published 2009) and not considered in this guideline.

The clinical question asked, and upon which the literature searching was undertaken, was:

‘What is the efficacy and safety of clopidogrel in the medical management of patients with UA or NSTEMI compared to other antiplatelets or placebo?’

3.2.2. Clinical methodological introduction

To look at evidence published since the NICE TA80, the literature was searched for systematic reviews or RCTs published from 2003 to 2009. Because of the high number of randomised trials in this area, the GDG only considered RCTs with a sample size of 250 or more. In addition, for a study to be included at least 60% of patients enrolled needed to have a diagnosis of non ST-segment elevation ACS, and the study had to report on at least one of the six key clinical outcomes agreed for this guideline (30 day survival, re-infarction, LV function, re-vascularisation, quality of life, and serious complications).

Overall, studies identified in this area add some evidence to a number of issues:

• Timing of clopidogrel
  The current two approaches are either to initiate treatment early (for example, in A&E, or ‘upstream’) or wait until the time of cardiac catheterisation when the coronary anatomy can be defined and a decision made on whether revascularisation is deemed appropriate. The advantage of starting treatment early is the potential to reduce early ischaemic events, but the disadvantage is the potential for increased bleeding in patients who subsequently require early CABG⁵⁸. The delayed approach, of using clopidogrel only after cardiac catheterisation, would avoid the increased bleeding risk for patients who undergo CABG.
• Loading dose of clopidogrel (300mg versus 600mg)^f
• Benefits of clopidogrel with, or without, glycoprotein IIb/IIIa inhibitors (GPIIb/IIIa), on a background of aspirin therapy

NICE TA80 assessed the double blind CURE RCT (N= 12562; mean follow-up nine months), in which patients with non ST-segment elevation ACS were randomized to clopidogrel (loading dose of 300 mg followed by 75 mg/day) or placebo and both arms received aspirin (75–325 mg/day) ^{59 60}. The primary end-point (cardiovascular death, MI, or stroke) at 30 days was significantly lower in the clopidogrel group. There was also some further benefit which developed later (30 to 365 days). There was no significant excess in life-threatening bleeds in each period.

Since the NICE TA80, two additional subgroup analyses of the CURE study have been published ^61,62. Lewis et al. compared clopidogrel with placebo (on a background of aspirin) in a subgroup of people undergoing PCI (N=2658). Outcomes were assessed in those who received PCI less than 48 hours since randomisation, greater than 48 hours since randomisation, and after hospital discharge. Fox et al. evaluated the benefits and the potential for increased bleeding among the patients who underwent PCI, CABG or medical therapy (no revascularisation) ⁶¹.

Two new RCTs ^63,64 were identified that compared different doses of clopidogrel (300 mg versus 600 mg) on a background of aspirin. In the Cuisset et al. RCT (N=292 non ST-segment elevation ACS; follow-up 30 days), the timing between the loading dose of clopidogrel and PCI was 12 to 24 hours. In the Yong et al double blind RCT (the PRACTICAL Trial) (N=256; follow-up six months) all patients received 300 mg of clopidogrel 12 hours prior to randomisation. At randomisation, patients received either another 300 mg of clopidogrel (the 600 mg group) or matching placebo (300 mg group). Angiography was performed no sooner than two hours after study drug administration. Mean time between randomisation and the first 300 mg dose of clopidogrel was 12 hours. Mean time between study drug administration and angiography was 13.2 hours (SD=14.4 hours) and between study drug administration and PCI was 16.1 hours (SD=10.9 h) ⁶³.

The CREDO study evaluated the effects of long-term treatment (12 months) with clopidogrel (75 mg once daily) in patients undergoing elective PCI (N=2116; 52.8% UA; 13.7% recent MI; 32.8% stable angina and other) ⁶⁵. The CREDO trial was excluded from TA80 on the grounds that the population was undergoing ‘elective’ PCI; however the GDG included CREDO as the population contained a large proportion of people with UA and recent MI such that it reached the 60% UA/NSTEMI inclusion criterion. The optimal timing for the initiation of clopidogrel (300mg) before PCI was evaluated in a post-hoc analysis of the CREDO RCT ⁶⁶. The analysis included 1815 patients who underwent PCI during the index cardiac catheterization procedure, and assessed the effect of the duration of clopidogrel pre-treatment (<15 hours or ≥15 hours before PCI) on the composite outcome of death, MI, or urgent target vessel revascularisation at 28 days. The timing of clopidogrel pre-treatment was not randomised and there was a high (40-50%) concomitant use of GPI.

The TARGET study randomised patients undergoing elective or urgent PCI-stenting to tirofiban or abciximab on a background of aspirin (75 to 325 mg), and heparin (to achieve ACT ≥ 250 seconds). In a post-hoc analysis ⁶⁷ outcomes were assessed according to whether patients received 300mg of clopidogrel before PCI (N=4477) versus immediately after the procedure (N=332). A limitation of this study is that the timing of clopidogrel administration was at the cardiologist's discretion and thus, was not randomised.

It should be noted that differing study designs, dosing and titration regimens and the differing populations included might limit direct comparisons between studies.

3.2.3. Clinical evidence statements

Pre-treatment with clopidogrel in patients receiving PCI, CABG, or medical management⁶¹

see summary

Table 3-1.

Table 3-1

Subgroup analysis of the CURE study by type of revascularisation strategy .

Compared to placebo, clopidogrel significantly reduced the risk of:

• CV death, MI or stroke in people undergoing PCI
• CV death, MI or stroke in people having medical management

There was a non–significant difference between the placebo and clopidogrel arms for:

• CV death, MI or stroke in people undergoing CABG.
• Major bleeding in people undergoing CABG
• Life threatening bleeding in people undergoing CABG.

This study concluded that clopidogrel use was associated with a lower incidence of the composite endpoint compared with placebo. This trend was similar across different subpopulations undergoing CABG or PCI, and those patients treated medically.

Evidence Level 1+

Fox et al. also highlighted that whereas no excess in any bleeding was observed for patients stopping clopidogrel more than five days before surgery, a non–significant excess in major bleeding was seen for those who continued the drug within five days of surgery. However, the study indicates that when using the more stringent TIMI or GUSTO definitions of major bleeding (used in most trials), there was not an increase in major bleeding. These results suggest that the use of clopidogrel within five days before CABG is associated with more mild to moderate bleeding but no excess life-threatening bleeding.

Relationship between pre-treatment with clopidogrel and PCI timing

See Table 3-2 and Table 3-3.

Table 3-2

CURE study – subgroup analysis by timing of PCI .

Table 3-3

CREDO study – Post hoc analysis by timing of pre-treatment with clopidogrel before PCI .

Another subgroup analysis of the CURE RCT ⁶² (N= 2538 undergoing PCI) showed consistent treatment benefit of clopidogrel over the nine-month follow-up period regardless of the timing of PCI after randomisation (PCI < 48 hours, PCI ≥ 48 hours, PCI after discharge) for the composite endpoint of CV death or non-fatal MI. The data suggested that the greatest benefit accrued in those patients undergoing earlier intervention, though differences did not reach significance.

Evidence Level: 1+

The CREDO trial⁶⁵ was undertaken to investigate two principal objectives; first, to evaluate the benefit of long-term (12-month) treatment with clopidogrel after PCI, and second, to determine the benefit of initiating clopidogrel with a pre-procedure 300mg loading dose. Patients were randomly assigned to receive a 300-mg clopidogrel loading dose (n=1053) or placebo (n=1063) three to 24 hours before PCI. Thereafter, all patients:

• received clopidogrel, 75 mg/d, through day 28. From day 29 to 12 months,
• patients in the loading-dose group received clopidogrel, 75 mg/d, and those in the control group received placebo. Both groups received aspirin throughout the study. At one year, long-term clopidogrel therapy was associated with a 26.9% relative reduction in the combined risk of death, MI, or stroke (95% CI 3.9% to 44.4%; p=.02; absolute reduction, 3%). Clopidogrel loading pre-PCI overall did not significantly reduce the combined risk of death, MI, or urgent target vessel revascularization at 28 days (reduction 18.5%; 95% CI, −14.2% to 41.8%; p=.23). However, in a pre-specified subgroup analysis, patients who received clopidogrel loading at least six hours before PCI did show a relative risk reduction of 38.6% (95% CI, −1.6% to 62.9%; P=.051) compared with no reduction with treatment less than six hours before PCI. Risk of major bleeding at one year increased, but not significantly (8.8% with clopidogrel vs 6.7% with placebo; p=.07).

The optimal timing for a 300mg loading dose of clopidogrel before PCI was further evaluated in a post hoc analysis of the CREDO study ⁶⁶. All patients received 75 mg of clopidogrel at the time of PCI but some were randomized to receive also a loading dose of clopidogrel (300 mg) 3 to 24 hours before PCI. The incidence of the 28-day combined endpoint of death, MI, or urgent target vessel revascularization, was similar in those patients who simply received 75mg of clopidogrel at the time of PCI and those who received a clopidogrel loading dose less than 15 hours before PCI. The benefit of clopidogrel loading was confined to those patients pre-treated more than 15 hours before the PCI procedure (RR reduction 58.8% [p= 0.028] versus placebo).

Evidence Level: 2+

Loading doses of clopidogrel in patients undergoing PCI (300mg versus 600mg)^g

Two RCTs ^{64 63} addressed the issue of clopidogrel loading dose (600 mg versus 300 mg) prior to PCI or angiography. One (Cuisset, 2006 103 /id) randomized 292 patient with NSTEMI/UA to receive either 300mg or 600mg of clopidogrel at least 12 hours before undergoing PCI and excluded the use of GPIs. The other (Yong, 2009 4178 /id} randomized 256 patients with UA/NSTEMI to receive either 300mg or 600mg of clopidogrel prior to undergoing coronary angiography. 140 patients then underwent PCI and 68.6% of these received a GPI. See Table 3-4 for a summary of results.

Table 3-4

Clopidogrel loading dose (300mg versus 600mg).

One RCT ⁶⁴ showed a significant reduction in recurrent ischaemic events in the 600 mg clopidogrel group compared with the 300 mg group with no patient experiencing post-procedural major bleeding or requiring transfusions.

Evidence Level: 1+

By contrast, the PRACTICAL trial ⁶³ showed a non–significant difference between the 600 mg and 300 mg clopidogrel groups for:

• Post-PCI myonecrosis
• Death at six months
• MI at six months
• Stroke at six months
• Death / nonfatal MI / nonfatal stroke / hospitalizations for recurrent ischemia at six months
• TIMI major haemorrhage at one month
• TIMI minor haemorrhage at one month

Evidence Level: 1+

3.2.4. Health economic methodological introduction

3.2.5. Health economic evidence statements

Karnon et al. ⁶⁹ reported a cost per QALY gained for clopidogrel (for one year) in combination with aspirin compared to aspirin alone in patients with UA/NSTEMI of £7365 that was robust to various sensitivity analyses.

Heeg et al. ⁷⁰ reported a cost per life year gained of £771 in patients with UA/NSTEMI although there were some methodological concerns regarding the paper which may account for the more favourable result. Karnon et al. reported a cost of £6991 per life year gained in the same population.

Lamy et al. ^71,71 reported a cost per event avoided of £10,366 in patients with UA/NSTEMI (one year analysis based on RCT resource use). Karnon et al. reported a similar cost of £10,599 per event avoided in the same population (lifetime modelling analysis). Lamy et al. reported that at 30 days clopidogrel in combination with aspirin dominated aspirin (that is it reduced costs and improved outcomes).

Heeg et al. ^70,70 found that in patients with UA/NSTEMI undergoing PCI and in patients undergoing PCI in general clopidogrel in combination with aspirin was found to dominate aspirin alone (it reduced costs and improved outcomes).

The new economic evidence identified in this literature review supports the recommendation made in TA80 for use of clopidogrel in combination with aspirin in patients with UA/NSTEMI.

The NICE TA80 model, and the manufacturer's model submitted during the development of the TA, were both based on the TA047 glycoprotein IIb/IIIa inhibitor model. Duration of treatment was the main area of uncertainty but long-term treatment is outside the scope of this guideline. The Karnon study ⁶⁹ assessed the uncertainty around cost effectiveness and found that 77% of simulations were under £20K / QALY and therefore affordable to the NHS. Lamy's 2004 post-hoc stratification of CURE data by TIMI risk ⁷¹ showed no change in cost-effectiveness conclusions, though the CURE study recruited patients which the investigators categorised as being medium and high risk patients.

The cost effectiveness of a 600mg loading dose compared to a 300mg loading dose has not been assessed. The additional cost is £5.04⁸⁶.

The group noted that clopidogrel will come off patent in 2010/11 and the effect this has on costs may need to be considered (though the likely reduction in cost would increase cost-efficacy).

3.2.6. Evidence summary

The purpose of reviewing the use of clopidogrel in this guideline was to take account of research published since TA80 and determine whether the previous recommendations should be revised, and particularly to address:

• which people with UA/NSTEMI should be offered clopidogrel
• optimal time of administration
• optimal dosage
• its use peri-operatively in patients undergoing CABG
• its use when possible PCI is planned
• risks associated when combined with other therapies
• whether the previous assessment of cost-effectiveness still applies

3.2.7. Evidence to recommendations

In the previous technology appraisal ‘moderate-to-high risk’ was determined by “clinical signs and symptoms, accompanied by one or both of the following:

• the results of clinical investigations, such as new ECG changes (other than persistent ST elevation) indicating ongoing myocardial ischaemia, particularly dynamic or unstable patterns
• the presence of raised blood levels of markers of cardiac cell damage such as troponin”

Such clinical determinants of risk were still felt applicable, although the use of single risk components (such as troponin) predict risk poorly, particularly when used in a binary fashion (troponin elevated, or not)¹¹. This guideline has addressed the issue of risk in more detail and offers a more comprehensive analysis of factors that clinicians may more accurately use to categorise individual patients into their broad categories of risk, and the use of risk scoring systems (see section 2).

The CURE trial also used a risk scoring system (TIMI 0 to 7; lowest-highest risk) to assess the effect of clopidogrel with increasing levels of baseline risk of an adverse outcome. Our interpretation of the data suggests that most patients enrolled in CURE were at low-medium risk of an adverse cardiovascular outcome, in the context of an unselected population of people with non ST-segment elevation ACS. High risk patients were not enrolled, which is at variance with previous interpretations of the trial's risk profile. See Figure 2-1 below.

Figure 2-1

Six-month mortality (y-axis) and GRACE score (x-axis) data from the GRACE Registry. Six month mortality in CURE for placebo (red) and clopidogrel (blue) groups shown by TIMI risk stratum on the ‘GRACE curve’ (dark blue). TIMI risk score (more...)

The GDG concluded that clopidogrel was likely to be of benefit to those at risk levels 1b and above (six-month mortality >1.5%) by our classification, but that any benefit for those in the lowest risk cohort (1a; six-month mortality 0-1.5%) was likely to be very small and may be outweighed by any additional bleeding caused. In this lowest risk group of people admitted to hospital with UA/NSTEMI in England & Wales the decision regarding whether or not to prescribe clopidogrel may be left to individual physician discretion and based on an assessment of its potential benefit (particularly reducing ischemic events) against bleeding risk.

The group felt that evidence had now accumulated clearly supporting a loading dose of 300mg of clopidogrel for most people admitted with UA/NSTEMI. Those who are at lowest risk (predicted six-month mortality 0-1.5%; cohort 1a) have least to gain and the decision to prescribe clopidogrel for these patients should be made on an individual basis, depending on circumstances. If a very early (<24 hours) invasive intervention is planned, a higher loading dose should be considered, especially if a patient is undergoing intervention within six hours. With a standard loading dose of 300 mg, it is likely that some patients will not yet have obtained the full anti-platelet effect of clopidogrel prior to the PCI procedure. The group considered that a higher loading dose for patients in whom a very early (within 24 hours) invasive strategy is planned was reasonable, on the basis that there was no evidence of any increased risk (acknowledging that this is not the same as saying that there is evidence of no increased risk) and the additional cost was modest. As the group were not able to formally review all the evidence for a 600-mg loading dose they were not able to recommend this at the time of publication. The group also stressed that clopidogrel should not be given without a confirmed diagnosis of ACS, because of its potential to increase bleeding risk.

In the circumstance where a cardiac arrest occurs before medical attendance, or where there is no clear clinical indicator of prior ischaemia then decisions about medical therapy should await assessment in hospital (clinical review, ECG, risk assessment, troponin etc.). It would not be appropriate to recommend clopidogrel to all patients who have had a cardiac arrest because clearly other conditions than an ACS may have precipitated the arrest.

After publication of TA80, NICE had clarified the recommendation “up to 12 months” to mean “for 12 months”. This guideline is now in a position to formalise this change, along with a reference to the secondary prevention of MI guideline⁴ and advising clinical review prior to stopping treatment (because of concerns about prescriptions automatically being stopped at 12 months, sometimes inappropriately, through primary care prescribing software reminders. Some patients, for instance those who have had drug eluting stents as part of complex PCI procedures, or those who have had late stent thrombosis, may be advised to remain on clopidogrel and aspirin indefinitely.

3.2.8. Recommendations

R9.

As soon as the risk of adverse cardiovascular events has been assessed, offer a loading dose of 300 mg clopidogrel in addition to aspirin to patients with a predicted 6-month mortality of more than 1.5% and no contraindications (for example, an excessive bleeding risk)^h.

R10.

Offer a 300-mg loading dose of clopidogrel to all patients with no contraindications who may undergo PCI within 24 hours of admission to hospitalⁱ.

R11.

It is recommended that treatment with clopidogrel in combination with low-dose aspirin should be continued for 12 months after the most recent acute episode of non-ST-segment-elevation ACS. Thereafter, standard care, including treatment with low-dose aspirin alone, is recommended. (This recommendation has been incorporated from TA80).

R12.

Consider discontinuing clopidogrel treatment 5 days before CABG in patients who have a low risk of adverse cardiovascular events.

R13.

For patients at intermediate or higher risk of adverse cardiovascular events, discuss the continuation of clopidogrel before CABG with the cardiac surgeon and base the decision on the balance of ischaemic and bleeding risk.

3.3.1. Clinical introduction

This section is intended to update the NICE TA on glycoprotein inhibitors (GPIs) (TA47) published in 2002.

Aspirin was the first anti–platelet therapy to be shown to improve outcome in acute coronary syndromes, and has been followed by other oral antiplatelet agents such as the thienopyridine clopidogrel, and also the intravenously administered GPIs, such as abciximab, eptifibatide or tirofiban. With increasingly aggressive platelet inhibition, and concomitant anticoagulant/antithrombotic therapy, the risk of bleeding has increased. TA47 made recommendations regarding the use of the GPIs in the treatment of ACS, and highlighted the importance of assessment of underlying patient risk because the overall benefit of these agents (the balance of benefit against risk of an adverse event) is greatest in those at highest underlying risk of recurrent myocardial ischaemia or infarction.

GPIIb/IIIa antibodies and receptor antagonists inhibit the final common pathway of platelet aggregation (crossbridging of platelets by fibrinogen binding to the GPIIb/IIIa receptor). Of these, abciximab is a large monoclonal antibody directed against the receptor, whereas tirofiban and eptifibatide are non-antibody receptor (often referred to collectively as “small molecule”) inhibitors.

The clinical question asked, and upon which the literature was searched, was:

‘What is the safety and efficacy of adding a GPI (tirofiban, eptifibatide and abcixmab) to aspirin and heparin therapy as adjunct therapy to patients with UA/NSTEMI undergoing PCI compared to the combination of aspirin and LMWH?’

3.3.2. Health economic methodological introduction

TA47 reviewed the economic literature published since the previous appraisal of GPIIb/IIIas, TA12. Economic models from the eptifibatide sponsors (Schering Plough), tirofiban sponsors (MSD), and abciximab sponsors (Eli Lilly) were also reviewed and the Assessment Group undertook their own analysis.

The systematic literature review from TA47 identified the following studies:

• Medical management of UA/NSTEMI. Seven studies were identified in the TA12 review; no new studies were found as part of TA47. Of these seven, none were UK based and only one study was considered of interest. This was a US study by Mark et al. that was the only prospective economic analysis undertaken alongside a RCT (PURSUIT, eptifibatide) and was of value only as a source of comparison with the Schering Plough analysis.
• Alongside PCI. Seventeen studies were identified in the TA12 review; six new studies were found as part of TA47. While the majority found GPIs to be cost effective in the context of patients undergoing PCI, the studies were not from a UK perspective and most were judged to have serious limitations as inputs to decision making in the UK; these included the use of effectiveness data, disease-specific endpoints (such as CV events avoided), and lack of consideration of down-stream consequences of short-term outcomes from trials.

The model submitted by the eptifibatide sponsors (Schering Plough) for TA12 evaluated the cost effectiveness of eptifibatide in the medical management of UA/NSTEMI. It uses a Western European (n=3697) and UK (n=429) subgroup of the PURSUIT RCT as its main data source for outcomes and resource use (up to six months). Lifetime outcomes are modelled based on these data. The UK analysis found eptifibatide to be dominant (i.e. cost saving and more effective), but this may be considered unreliable due to the small patient group. The Western European analysis, which might be considered more reliable found the incremental cost effectiveness ratio (ICER) to be £8179-£11,079 per life year gained (depending on discount rate used for outcomes). A key limitation is that costs are not extrapolated past six months which would feasibly impact the results.

The model submitted by the tirofiban sponsors (MSD) for TA12 evaluated the cost effectiveness of tirofiban in the medical management of UA/NSTEMI. It uses effectiveness data from the PRISM-PLUS RCT. The primary analysis reports a cost per event avoided (all cause mortality, new MI, refractory ischemia or readmission for UA/NSTEMI) of £8,760 and £9995 using 7-day and 180-day outcomes respectively and the additional cost of tirofiban. A secondary analysis estimates that 22% of additional drug cost is offset by savings due to reduction in events.

The model submitted by the abciximab sponsors (Eli Lilly) for TA12 evaluated the cost effectiveness of abciximab alongside PCI in a UK setting. Baseline event rates and effectiveness of abciximab were based on the EPIC, EPILOG and EPISTENT RCTs. Impact on life years was evaluated by assuming that patients surviving at one year would live a further fifteen. Costs were not extrapolated past one year. The ICER was found to be £3554, £6247 and £12,421 per QALY gained with EPIC, EPILOG and EPISTENT respectively.

The assessment group judged the published and sponsor-driven cost-effectiveness analyses to have significant limitations with regard to UK decision-making. In particular the fact that effectiveness trials used in analyses were undertaken largely or wholly outside of the UK; given the different practice patterns in the UK (e.g. lower rates of PCI), the baseline risks, and possibly the relative risks associated with GPIs, may be different. This may translate to differences in cost-effectiveness. Also many used condition specific endpoints that inhibit interpretation of results in the decision-making context The Schering Plough analysis was considered the most relevant to UK decision-making.

The model developed by the Assessment Group examined four GPI treatment strategies:

• a GPI used immediately as part of initial management
• a GPI used after making a decision to carry out angiography with a view to PCI
• a GPI used as adjunct to PCI started up to an hour before the procedure
• no use of a GPI

The analysis showed that GPI use immediately as part of initial management was the most cost effective strategy with an ICER of £5738 per QALY gained compared to no GPI use. This conclusion was robust to various sensitivity analyses. Restricting strategy 1 to high risk patients only reduced the cost per QALY gained to £3966 and appeared more cost effective than treating all ACS patients. The additional benefits in all patients compared to high risk only was at a cost of £91,000 per QALY gained.

3.3.3. Clinical evidence statements

GPIs in conservative & invasive strategies

An individual patient data meta-analysis ⁸⁸ of six trials (PRISM, PRISM-PLUS, PARAGON-A, PARAGON-B, PURSUIT, and, GUSTO-IV ACS) compared GPIs with placebo or control therapy in 31,402 non ST-segment elevation ACS patients who were not routinely scheduled for early revascularisation (refer to summary). Most of the trials in this meta-analysis were undertaken in the pre-stent era. Also, most patients did not receive a thienopyridine anti–platelet agent (in GUSTO-IV ACS, the most recent of the GPI trials in the Boersma analysis, only 2% received a thienopyridine).

Table 3-5

Summary table of Boersma et al meta-analysis (six RCTs).

Compared to the control group, the GPI group had a significantly reduced chance of:

• death or MI at 30 days (primary outcome)

Evidence Level 1+

There was a non-significant difference between the control and GPI groups for:

• death at 30 days
• nonfatal MI at 30 days
• revascularisation (CABG or PCI) at 30 days
• intracranial haemorrhage at 30 days.

Evidence Level 1+

Compared to the control group, the GPI group had a significantly increased chance of:

• major bleeding at 30 days.

Evidence Level 1+

A highly significant interaction with respect to cardiac events was seen between gender and allocated treatment. In men, GPIs were associated with a 19% reduction in the odds of 30-day death or MI compared with placebo or control. By contrast, in women, there was a 15% increase. A further stratification by troponin concentration showed no evidence of a gender difference in treatment response, and a non–significant trend to a risk reduction was seen in men and women with raised troponin (see Table 3-6, Table 3-7 and Table 3-8).

Table 3-6

Meta-analysis by Boersma et al (interaction by gender). All patients.

Table 3-7

Meta-analysis by Boersma et al (interaction by gender). Patients with normal baseline cardiac troponin T or I <0·1 ug/L

Table 3-8

Meta-analysis by Boersma et al (interaction by gender). Patients with elevated baseline cardiac troponin T or I ≥0·1 ug/L

Further sub-groups analysis from this meta-analysis reported data on the effect of GPIs in the time period preceding a PCI (medical treatment):

• The authors reported that among patients who received PCI within 5 days (N=4378), the GPI group experienced significantly fewer MIs before the PCI occurred compared with the control group (OR, 0.70 [95% CI, 0.55 to 0.89]).
• For the subgroup of patients who did not undergo an early PCI (N=27024), there was a non-significant difference between the control and GPI group for death or MI at 30 days (OR, 0.95 [95% CI 0.87 to 1.02]).

These subgroup analyses should be interpreted with caution as the specific sub groups had not been randomised to control or GPI a priori. Pieper et al. have highlighted the pitfalls of inappropriate sub-group analyses undertaken in GPI trials and the potential for differing conclusions to be drawn depending on the analytical approach¹⁰⁸.

A second meta-analysis ⁹⁸ of published data included the same six RCTs pooled by Boersma et al, and analysed the effect of GPIs in 29,570 patients initially managed medically, and then treated with PCI. In this meta-analysis patients were defined according to the procedure received. In PRISMPLUS, the study arm not including heparin (n=345) was discontinued before completion of the trials and was excluded from this analysis. In PURSUIT, the protocol mandated the discontinuation of the lower-dose arm of eptifibatide (N=1487) after documentation of an acceptable safety profile of the higher dose in the interim analysis; thus the lower dose arm was not included in the Roffi et al meta-analysis. Therefore, the Roffi et al. meta-analysis had a total of 29,570 patients compared with the 31,402 included in the Boersma et al meta-analysis.

The findings of the Roffi meta-analysis suggested a gradient of benefit conferred by GPIs depending upon the revascularisation strategy used. Accordingly, patients undergoing PCI while on GPIs derived a significant benefit, while patients undergoing revascularisation after drug discontinuation demonstrated a moderate event reduction that did not reach statistical significance, and only a marginal benefit (non significant) was observed among patients managed medically (see Table 3-9).

Table 3-9

Summary of meta–analysis by Roffi et al.

Evidence Level 1+

Triple anti–platelet therapy

The ISAR-REACT 2 ^91,94, and ELISA-2 ⁹⁷ RCTs assessed the addition of a GPI to aspirin, clopidogrel (or ticlopidine) and heparin (i.e. triple antiplatelet therapy) in people with non ST-segment elevation ACS.

These studies differed in several respects such as the GPI evaluated, the baseline risk of population in which they were conducted, the follow-up period and the loading dose of clopidogrel used (see Table 3-10).

Table 3-10

Summary of triple antiplatelet therapy studies.

In ELISA-2 and ISAR-REACT 2, compared with people receiving dual antiplatelet therapy (aspirin + clopidogrel) together with heparin, people randomised to triple antiplatelet therapy (aspirin + clopidogrel + a GPI) with background heparin had a significantly reduced risk of:

• Death, MI, or urgent target vessel revascularisation at 30 days
• Death, MI, or target vessel revascularisation at 1 year

Evidence Level: 1+

There was a non–significant difference between the groups for major bleeding.

Evidence Level: 1+

In the ISAR REACT 2 trial ⁹¹, there was non signficant difference in the incidence of death/MI/UTVR at 30 days between the abciximab group (4.6%) and the placebo group (4.6%) in people who had normal troponin concentrations ≤ 0.03 μg/L [N=973; RR, 0.99; (95% CI, 0.56 to 1.76); p= 0.98),]. In patients with an elevated troponin level (N=1049; troponin > 0.03 μg/L), death/MI/UTVR at 30 days was significantly lower in the abciximab group (13.1%) compared with the placebo group (18.3%) [RR 0.71 (95% CI, 0.54 to 0.95; p=.002) (p=0.07 for interaction).

Evidence Level: 1+

Timing issues

Prospective randomised trial data comparing GPI administration upstream versus in the catheterisation laboratory are limited. Only three RCTs (ISAR COOL, ACUITY TIMING, and EARLY ACS) ^{95,99 100} addressed this area.

In the ACUITY-TIMING RCT ⁹⁹ deferred selective vs. routine upstream administration of GPIs was evaluated. Patients assigned to routine upstream GPI received either eptifibatide or tirofiban started at a median time of 35 minutes after randomisation and infused for a median of 4.0 hours before PCI. In contrast, patients randomised to deferred selective GPI use were assigned treatment with either eptifibatide or abciximab started just prior to PCI, approximately 3.9 hours later than GPIs were begun in the upstream use group. The GPI infusion continued during angioplasty and for 12 to 18 hours thereafter. For patients assigned to deferred selective GPI use, the investigator chose whether eptifibatide or abciximab was administered only to patients undergoing angioplasty, begun 5 to 10 minutes prior to first balloon inflation, and continued for 12 hours (abciximab) or 12 to18 hours (eptifibatide) thereafter. It should be noted that people randomized to upstream or deferred GPI had also been randomized to either heparin or bivalirudin, and thus there is a mixture of antithrombin use in the upstream and deferred GPI arms.

In the EARLY ACS trial ¹⁰⁰ people with non ST-segment elevation ACS undergoing an early invasive strategy (N=9492) were randomised to either early upstream eptifibatide or to matching placebo. After coronary angiography, but before PCI, investigators could request a ‘PCI-study drug kit’ for patients who could benefit from eptifibatide on the basis of angiographic evidence. The first bolus of the “PCI-study drug kit” contained eptifibatide for patients who had previously had placebo and placebo for people who previously had eptifibatide. An open label infusion of eptifibatide was started and continued for at least 18 to 24 hours after PCI. During PCI if a thrombotic complication occurred after the catheter guide wire had crossed the lesion, a “bailout drug kit” that contained a bolus therapy opposite to the initial study group drug was given. The median time from randomisation to study drug initiation was 0.5 hours in both groups. The median time from randomisation to angiography was 21.4 hours and to PCI was 22 hours. See Table 3-11.

Table 3-11

Summary of primary outcomes in people randomised to upstream or deferred GPIs.

Two meta-analyses were performed pooling the outcomes of the ACUITY TIMING and EARLY ACS trials. In the first meta-analysis the entire trial populations of the two RCTs were pooled. This means that for ACUITY TIMING, the upstream and the deferred GPI arms are a mixture of heparin and bivalirudin. Whilst pooling studies increases statistical power, the bivalirudin contamination in the ACUITY TIMING trial is a limitation of this meta-analysis. (see Figure 2-2, Figure 2-3, Figure 2-4, Figure 2-5, Figure 2-6, Figure 2-7, and Figure 2-8).

Figure 2-2

Death, MI, or unplanned revascularization at 30 days in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-3

Death or MI at 30 days in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-4

Death at 30 days in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-5

MI at 30 days in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-6

Unplanned revascularization in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-7

TIMI Major Bleed in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-8

TIMI Minor Bleed in the entire trial populations (bivalirudin contaminating the ACUITY TIMING arms).

The meta-analysis was re-run with unpublished data from the ACUITY TIMING RCT ¹⁰⁹, in which the upstream and deferred GPI arms were from patients only randomised to heparin (no bivalirudin contamination). This provides a more comparable pharmacological background between the two RCTs, although with fewer patients, the statistical power is decreased. Table 3-12 summarised the two meta-analyses pooling the EARLY ACS and ACUITY TIMING RCTs. (See Figures 2-2, through to 2-15).

Table 3-12

summarises the two meta-analyses pooling the ACUITY TIMING and EARLY ACS RCTs.

Figure 2-9

Death, MI, or unplanned revascularization at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-10

Death or MI at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-11

Death at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-12

MI at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-13

Unplanned revascularisation at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-14

TIMI major bleed at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

Figure 2-15

TIMI minor bleed at 30 days (no bivalirudin contaminating the ACUITY TIMING arms).

When the meta-analyses were run without bivalirudin contaminating the ACUITY TIMING arms and compared with deferred GPI use, upstream GPI use) significantly:

• Decreased the risk of MI at 30 days
• Decreased the risk of unplanned revascularisation at 30 days
• Increased the risk of TIMI major bleed
• Increased the risk of TIMI minor bleed

Evidence Level: 1+

There was no significant difference between upstream and deferred GPI use for:

• the composite outcome of death, MI, or unplanned revascularisation at 30 days;
• death at 30 days
• death or MI at 30 days

Evidence Level: 1+

3.3.4. Health economic evidence statements

Bakhai et al.¹⁰² reported an incremental cost-effectiveness ratio of £13,388 per event averted for tirofiban plus standard therapy compared to standard therapy alone in the initial medical management of UA/NSTEMI. Without the estimation of QALYs it is difficult to interpret the results.

Brown et al.¹⁰² reported an incremental cost-effectiveness ratio of £8436 per life year gained for eptifibatide plus standard therapy compared to standard therapy alone in the initial medical management of UA/NSTEMI. Note that costs were not extrapolated past six months. The 30-day analysis produced an ICER of £22,760 per event avoided. While reporting slightly different results, this analysis is judged to be consistent with the Schering Plough cos-effectiveness analysis evaluated as part of TA47 and as such does not give cause to change the recommendations made.

The new evidence does not contradict the existing TA model and recommendations.

Health economic modelling

Cost-effectiveness modelling was undertaken for this guideline to look at the use of GPIs taking into account contemporary management In particular it addressed the use of GPIs in combination with clopidogrel, bivalirudin was included as a possible alternative to heparin plus a GPI and fondaparinux as an alternative to heparin was incorporated.

For the full analysis methods and detailed results and discussion see the report in Appendix B and Appendix C. A summary is provided below.

Methods

A cost–utility analysis was undertaken with costs and quality-adjusted life-years (QALYs) considered over patients' lifetimes from a UK NHS perspective. The analysis is relevant to patients undergoing an early invasive management approach – that is coronary angiography with revascularisation if indicated – because trial results utilised for GPIs and bivalirudin used in the analysis were only relevant to a population undergoing angiography. This is discussed in more detail in the full report in Appendix C.

This compared the following treatment strategies in the acute management of UA/NSTEMI (heparin baseline):

• Aspirin +clopidogrel +heparin (LMWH or UFH)
• Aspirin +clopidogrel +heparin + GPI during PCI only
• Aspirin +clopidogrel +heparin + GPI upstream of angiography
• Aspirin +clopidogrel +bivalirudin upstream of angiography
• Aspirin +clopidogrel +heparin +bivalirudin during PCI only.

In addition the analysis was run as above but with fondaparinux substituted for heparin in the first three arms (fondaparinux baseline). Fondaparinux was not incorporated in the bivalirudin arms in this analysis as there is no experience with these agents combined and so it was not judged appropriate.

Cost effectiveness was analysed by six risk subgroups, as summarised in Table 3-13 below. The creation and interpretation of these risk groups is discussed in more detail in the Risk chapter of the guideline (section 2) and the report of the analysis of MINAP data for the cost effectiveness analysis (Appendix B).

Table 3-13

Risk groups.

The general approach taken was to obtain contemporary UK estimates of events for the aspirin, clopidogrel and heparin arm of the model from recent MINAP (the national audit of ACS management) data. These were stratified by acute management strategy: PCI, CABG, angiography only. Where inputs were not available from the analysis of MINAP data, figures were sourced from the literature or discussion with the GDG. One-year death, MI and post-acute revascularisation, and in-hospital bleeding were incorporated. The effects of different treatment combinations are then modelled by applying relative risks from randomised controlled trials identified by the systematic review of the clinical literature for the guideline – one-year relative risks were used where available except for bleeding. Relative risks were applied to the appropriate part of the population; for example, only PCI patients, if only relevant to these patients.

Lifetime QALYs were estimated based on one-year status: dead, alive having had a new MI, alive without new MI. At one-year patients were attributed a number of life-years based on this status. Those alive at one year with new MI were attributed a lower estimate than those alive without new MI. Life-years were adjusted by a quality of life weight for people with ACS to estimate QALYs. As the rates of death and MI will vary with treatment strategy, so will the QALYs.

Lifetime costs were estimated taking into account initial drug treatment costs, the cost of MI, bleed and post-acute revascularisation events up to one year and average disease-related costs incurred if alive post one-year.

Treatment effects were based on studies identified in the clinical review. Only studies with at least 50% clopidogrel use were used. Relative treatment effects were based on the following studies:

• ISAR-REACT 2^91,94: GPI versus no GPI in a PCI UA/NSTEMI population
• ACUITY timing (heparin only background, clopidogrel pre-angio/pre-PCI subgroup)^99,109: upstream GPI versus PCI GPI in an early angiography UA/NSTEMI population
• ACUITY (clopidogrel pre-angio/pre-PCI subgroup)^109-111: bivalirudin vs LMWH/UFH + GPI in an early angiography UA/NSTEMI population
• REPLACE-2 ACS subgroup¹¹²: bivalirudin during PCI vs heparin + GPI during PCI in a PCI ACS population
• OASIS-5¹¹³: fondaparinux vs enoxaparin in a UA/NSTEMI population

The Early ACS trial also compares upstream GPI vs PCI GPI use in an early angiography UA/NSTEMI population¹⁰⁰. It was published late in the guideline development process and only reports 30-day outcomes, whereas the model had been developed with one-year baseline event rates and effectiveness data. Sensitivity analyses examined the possible impact of this study.

Two analyses were run:

1. Trial aligned analysis (costing based on trial vial usage where pre-angiography treatment period median 4hrs/mean 10hrs; ACUITY management split)
   • Costing based on trial vial usage; ACUITY management split
   • The ACUITY trial which includes 3 of the 5 comparators had a median treatment period pre-angiography of 4hrs (mean 10hrs)
   • This analysis is most aligned with the available trial data
2. Adjusted analysis (costing based on 72hr pre-angiography treatment period; MINAP management split)
   • Costing based on a simulation assuming 72hr pre-angio treatment duration and a 1hr PCI treatment duration; MINAP management split
   • This analysis makes some adjustments to costing and management split that may be more typical for the UK
   • Note that this analysis potentially biases against upstream treatments as costs are increased but efficacy remains the same and so should be interpreted carefully with this in mind.

The model was built probabilistically in order to take account of the uncertainty around input parameter point estimates. Probability distributions in the analysis were based on error estimates from data sources, for example confidence intervals around relative risk estimates. Various one-way and scenario sensitivity analyses, where one or more inputs were varied, were undertaken to test the robustness of model assumptions and data sources.

Results

Fondaparinux baseline analysis

The analysis incorporating a fondaparinux baseline (that is fondaparinux replaces heparin in the aspirin+clopidogrel+heparin, aspirin+clopidogrel+heparin+GPI during PCI, aspirin+clopidogrel+heparin+GPIupstream arms of the model), was considered most relevant to clinical decision making in the majority of cases. Fondaparinux has been found to be cost-effective compared to heparin as shown in the published literature¹¹⁴. Fondaparinux is cheaper than enoxaparin and is associated with clinical benefits. In the model Aspirin+clopidogrel+fondaparinux dominated Aspirin+clopidogrel+heparin in all of our analyses (although this comparison was a secondary objective of the analysis).

In the trial aligned analysis (when trial vial usage was used for costings and the ACUITY management split employed) routine addition of upstream GPIs seems to be most cost-effective for patients in risk groups 2 and 3, with selective PCI GPI use the most cost-effective in risk group 4. This is based on these options having the highest mean INB at a £20,000 per QALY threshold. In the adjusted analysis (with treatment costs estimated using a 72hr pre-angiography treatment duration and the MINAP management split employed) selective use of GPIs at PCI was found to be most cost-effective strategy; however, this analysis was considered likely to bias against upstream use of GPIs as treatment costs are increased but efficacy is not adjusted.

There was considerable uncertainty in the results. This is evidenced by differences between the deterministic optimal strategy and probabilistic optimal strategy especially in Groups 1a and 4. Also, there is a wide spread of the probability of cost-effectiveness across different strategies. In places the optimal strategy as based on mean INB is not the one with the highest probability of being cost-effective as based on the highest proportion of simulations. In addition there is uncertainty regarding applicability as the trial aligned analysis may not represent typical treatment durations in the UK; whereas the longer term analysis is limited by the lack of effectiveness data. It was also noted that from a clincal perspective, the longer the wait for angiography the more likely a patient would need a GPI prior to angiography and deferring use until PCI is undertaken may not be a clinically acceptable option.

Interpretation of results is complicated by the uncertainty in the analysis. Additional clinical considerations should be employed in interpretation and it was considered that it may be reasonable to recommend more than one option to reflect this uncertainty. Risk group 1 is considered least likely to benefit from additional treatment over and above aspirin+clopidogrel+fondaparinux. Dependent on appropriate clinical interpretation, due to the uncertainty it was considered that either GPI use upstream of angiography or selective GPI use in PCI might be considered likely to be cost-effective in higher risk groups. This is due to the fact that different options were found to be most cost-effective in the trial aligned and adjusted analysis but limitations in the analysis mean that a definitive conclusion is not possible based on these model results alone.

Note that the fondaparinux baseline analysis is dependent on the assumption that the relative effect of GPIs will not be impacted by whether heparin or fondaparinux is used as the baseline antithrombin – there were no studies that assessed GPIs against no GPIs in a population using fondaparinux. The OASIS-5 trial addresses this issue somewhat by examining 30-day outcomes for fondaparinux versus enoxaparin in subgroups of patients receiving clopidogrel and GPIs¹¹⁵. This analysis suggested that the benefits of fondaparinux are maintained in patients receiving clopidogrel or GPIs.

Heparin baseline analysis

If fondaparinux is not an appropriate option, then the analysis with a heparin baseline is most appropriate to review. In this analysis, risk group one is least likely to benefit from additional treatment over and above aspirin+clopidogrel+heparin. Heparin use with selective bivalirudin during PCI seems to be most cost-effective in risk groups 2-4. This is based on the mean INB from the heparin baseline analyses in both the trial aligned analysis (reflective of a short time to angiography) and the adjusted analysis (with treatment costs estimated using a 72hr pre-angiography treatment duration and the MINAP management split employed). Bivalirudin use pre-angiography was associated with more QALYs than the selective bivalirudin use but also additional costs and based on the mean INB this use was not cost effective at a £20,000 per QALY threshold.

As in the fondaparinux baseline analysis there was considerable uncertainty in the heparin-baseline analysis. In the trial aligned analysis (reflective of a short time to angiography) bivalirudin PCI was considered the most cost-effective treatment based on mean INB, bivalirudin use upstream of angiography, and upstream GPI use generally also had a high level of simulations where they were optimal. As risk increased the likelihood of bivalirudin initiated upstream of angiography being cost effective increased. It was also raised that there will sometime be a clinical need to give additional treatment upstream of angiography, for example if the patient is actively unstable. Interpretation of results is complicated by the uncertainty in the analysis. Additional clinical rationale should be employed in interpretation and it was considered that it may be reasonable to recommend more than one option to reflect this uncertainty. Dependent on appropriate clinical interpretation, due to the uncertainty it was considered that use of the following might be considered likely to be cost effective: bivalirudin used selectively during PCI; upstream bivalirudin; heparin plus upstream GPIs.

In the adjusted analysis (where costing was based on a 72hr pre-angiography treatment duration) PCI bivalirudin was also most cost effective, as would be expected as the upstream treatments will have higher costs in the model but the effectiveness was not adjusted. In addition, this analysis was considered the least clinically relevant because if patients were not going for angiography relatively quickly they would be most likely to be considered suitable for fondaparinux.

3.3.5. Evidence summary

There have been a number of publications investigating the use of GPIs since the last Technology Appraisal (2002). Trial designs, timing of treatments, patient populations and the use of adjunctive therapies and invasive strategies have differed between studies making comparisons difficult.

3.3.6. Evidence to recommendations

The GDG noted that:

• Whilst GPIs have been shown to reduce the risk of subsequent cardiac ischaemic events, this effect is most apparent when ischaemic risk is high (as judged by formal risk scoring, presence of raised troponins etc.), or if the duration of risk (delays to angiography and revascularisation) is prolonged (suggested by Boersma and Roffi meta-analyses).
• Much of the evidence relating to the use of GPIs preceded the widespread use of clopidogrel in addition to aspirin and an anti-thrombin.
• In the Boersma meta analysis GPIs reduced the 30 day relative odds of the combined endpoint death/MI by 15% (absolute benefit 1.7%) in troponin positive patients, whereas no odds reduction was seen in those who were troponin negative. However, it has also been demonstrated in the GRACE Registry that the presence of an elevated troponin alone does not reliably identify high risk patients, as judged by mortality outcome¹¹.
• Risk assessed by mortality outcome may not adequately reflect risk of a further ischaemic event For instance, using the online GRACE risk calculator and a theoretical patient profile¹¹⁷ it is possible to have a six-month predicted mortality of 4% (which lies in our risk cohort 2a [intermediate], as defined elsewhere in this guideline – see risk assessment chapter), but have a combined risk of death/MI at 6 months as high as 25%. Thus, caution needs to be shown when identifying the levels of risk at which GPIs should or should not be given. This note of caution with regards extrapolation of population data to individual patient decision making has also been highlighted earlier in the section on health economics.
• There may be a gender effect, with females appearing to benefit less from the use of GPIs than males (Boersma et al. ⁸⁸), although the subgroup of women with elevated serum troponin do appear to benefit.

No studies that assessed GPIs against no GPIs in a population using fondaparinux. The OASIS-5 trial addresses this issue somewhat by examining 30-day outcomes for fondaparinux versus enoxaparin in subgroups of patients receiving clopidogrel and GPIs¹¹⁵. This analysis suggested that the benefits of fondaparinux are maintained when used with GPIs, and the GDG felt it therefore unlikely that fondaparinux would result in a worse outcome than the combination of GPIs and other anticoagulants used in the trials.

Trials have tended to enrol people of low-intermediate, rather than high risk of an adverse outcome. Using methodology described earlier (reference to risk chapter) we plotted the six-month mortalities for ISAR-REACT-2, onto a GRACE graph (6-month predicted mortality by GRACE score – see Figure 2-15 below). The prior risk stratification of people with UA/NSTEMI (England & Wales) into risk cohorts 1a, 1b, 2a, 2b, 3 & 4, allowed us to attempt to position the results from this trial to an unselected population in England & Wales. These plots suggest ISAR-REACT-2 mainly enrolled people at low to intermediate levels of risk (risk cohorts 1 & 2) relative to the spectrum of risk in the unselected population of people with UA or NSTEMI.

Figure 2-15

6-month mortality (y-axis) and GRACE score (x-axis) data from the GRACE Registry. Six month mortality in ISAR REACT 2 for placebo (red) and abciximab (blue) groups plotted on the ‘GRACE curve’ (dark blue). Bars are 95%CI. Vertical grey (more...)

The use of GPIs has decreased in the UK since clopidogrel has become so widely used. There has also been a relative lack of evidence relating to the degree to which additional bleeding is associated with adding a GPI to background anticoagulant and antiplatelet therapy, because many trials did not mandate the use of clopidogrel in addition to aspirin. Assuming background triple therapy the GDG felt that the evidence was generally less convincingly in support of the routine use of GPIs in the medical (conservative) management of patients with NSTEMI and UA than was the case when TA47 was published. This was because, with the increased use of early angiography and revascularisation, patients managed conservatively increasingly fall into two categories; those at very low risk of a further ischaemic event, and those at very high risk of a bleeding complication. The evidence does support the use of upstream GPIs in patients at intermediate or high risk who are scheduled to undergo an early invasive strategy, albeit at the expense of some increase in bleeding risk.

GPIs were initially licensed based on clinical trials using unfractionated heparin as the anticoagulant choice. As a result the summaries of product characteristics for GPis state that they are ‘indicated as an adjunct to aspirin and heparin’ in the case of abciximab or ‘intended for use with aspirin and unfractionated heparin’ for eptifibatide and tirofiban. In addition, licensing will often state there is limited or no experience with low molecular weight heparins or fondaparinux.

Low molecular weight heparins such as enoxaparin and more recently the synthetic pentasaccharide fondaparinux are licensed for the treatment of UA and NSTEMI. The clinical trials involved the combination with glycoprotein inhibitors as well as aspirin and clopidogrel. Whilst the licensing authorities do not recommend the combination, it has become established practice to prescribe and administer LMWH or fondaparinux in (within their licensed indication) in combination with glycoprotein inhibitors.

3.3.7. Recommendations

R14.

Consider intravenous eptifibatide or tirofiban^m as part of the early management for patients who have an intermediate or higher risk of adverse cardiovascular events (predicted 6-month mortality above 3%), and who are scheduled to undergo angiography within 96 hours of hospital admission.

R15.

Consider abciximab as an adjunct to PCI for patients at intermediate or higher risk of adverse cardiovascular events who are not already receiving a GPI.

R16.

Balance the potential reduction in a patient's ischaemic risk with any increased risk of bleeding, when determining whether a GPI should be offered.