Clopidogrel and modified-release dipyridamole for the prevention of occlusive vascular events (review of Technology Appraisal No. 90): a systematic review and economic analysis

Myocardial infarction

Myocardial infarction (also known as a heart attack) is the interruption of the blood supply to the heart muscle. This is most commonly caused by occlusion of a coronary artery following the rupture of an atherosclerotic plaque. The resulting restriction in blood supply and oxygen starvation can cause damage to, or the death of, the heart muscle. Typical symptoms of MI include sudden chest pain with sweating or nausea, but MIs can also be symptomless. Women may experience different symptoms to men. Based on the results of changes in electrocardiogram (ECG) readings, MIs are classified into two subtypes: non-ST-segment elevated MI (NSTEMI) or ST-segment elevated MI (STEMI). The distinction has implications for future antiplatelet treatment. After a MI, a patient remains at high risk of a further MI or other occlusive vascular event.

Data from 2006 for the UK demonstrate that, across all ages, there were 146,000 cases of MI: 87,000 in men and 59,000 in women. 1 The incidence of MI varies across regions, between men and women, and increases with age. 1 Higher incidence rates are apparent in northern areas of the UK than in southern areas. In the UK, among men and women aged > 35 years, the prevalence is thought to be over 1.4 million. 1 Approximately 30% of people who experience an acute MI die before they reach hospital. 4 Patients who experience a MI and survive are likely to have a further cardiac event. 5

Ischaemic stroke

There are a number of different types of stroke; however, the majority of cases (approximately 70%) are ischaemic, caused through the blockage of an artery in the brain. 6 This leads to damage to, or death of, the brain cells due to lack of oxygen. The symptoms of stroke can include: numbness, weakness or lack of movement on one side of the body, slurred speech, difficulty finding words or understanding speech, problems with vision, confusion and/or severe headache. 7 A stroke happens suddenly and the effects are experienced straight away. 7 Anyone who suddenly has symptoms that might be caused by a stroke should be assessed as soon as possible using a test such as FAST (Face, Arm, Speech Test) and, on arrival at hospital, the ROSIER (Recognition of Stroke in the Emergency Room) test may be used. 7 A stroke may be classified as disabling or non-disabling.

The British Heart Foundation (BHF) reports that approximately 98,000 people experience a first ischaemic stroke every year in the UK, with little difference in rates between men and women but an increased risk with age. 8 Additionally, they estimate from 2006 data that, in the UK, as many as 1.1 million people have experienced a stroke; this is equivalent to a prevalence rate of 1.6% in the population in England and 2% in Wales. 8 The risk of recurrent stroke is greatest in the first 6 months following the initial event, but a patient may remain at greater risk of stroke than the general population for a number of years. 3 As many as 30% of strokes are thought to be recurrent. 9 Patients who have experienced a stroke are also at risk of further occlusive vascular events, including MI. 10,11

Transient ischaemic attack

A TIA is a disorder caused by temporary disturbance of blood supply to an area of the brain that results in a sudden but brief decrease (< 24 hours, usually < 1 hour) in brain functions and causes stroke-like symptoms. If the neurological deficit lasts more than 24 hours it is described as a stroke. Estimates for the UK indicate that between 46,000 and 65,000 people suffer a TIA each year and prevalence of TIA is projected to be 510,000. 8 In contrast with the trend noted in stroke data, there appear to be higher rates of TIA in women, and, as noted for stroke, incidence and prevalence rates increase rapidly with age. 8 Patients experiencing a TIA are at high risk of suffering a subsequent stroke, with 90-day risks of stroke reported to be as high as 10.5%. 12 In patients enrolled in clinical trials after a TIA or non-disabling ischaemic stroke, the annual risk of important vascular events (death from all vascular causes, non-fatal stroke or non-fatal MI) is reported as being between 4% and 11%; the corresponding estimate for population-based studies is 9% per year. 13

Peripheral arterial disease

Peripheral arterial disease is a condition in which the arteries that carry blood to the arms or legs become narrowed or congested, slowing or stopping the flow of blood. Data related to the prevalence of peripheral arterial disease vary. Detailed data from the USA indicate peripheral arterial disease rates of 4.3% in those < 40 years of age and 14.5% in those > 75 years. 14 Other reports estimate rates at 12–14% in the general population and 20% in those > 75 years of age. 15 The scoping document for this review indicated that, for the UK, approximately 20% of people aged from 55 to 75 years of age have evidence of lower extremity peripheral arterial disease and 5% of these appear to have symptoms, the most common of which is intermittent claudication (pain on walking). As the size of the UK population aged ≥ 55 years is approximately 17 million, this equates to a prevalence of around 170,000 with intermittent claudication. 16 It is thought that worldwide, and in the UK, peripheral arterial disease is underdiagnosed and undertreated. 17,18 Patients with peripheral arterial disease may experience significant pain and poor quality of life (QoL). 19 Over 5 years, about 20% of people with intermittent claudication have a non-fatal cardiovascular event (MI or stroke). 15 People with peripheral arterial disease, including those who are asymptomatic, have a high risk of death from MI and ischaemic stroke, their relative risks (RRs) being two to three times that of age- and gender-matched groups. 19 coronary heart disease is the major cause of death in people with peripheral arterial disease of the legs. 20

Although the diagnosis of peripheral arterial disease can generally be made from clinical history and examination, objective evidence of significant peripheral arterial disease can be made by obtaining an ankle–brachial pressure index. This index is the ratio of the ankle to brachial systolic pressure and may be measured using a sphygmomanometer and handheld Doppler device. 19 Obtaining an ankle–brachial pressure index is non-invasive and relatively easy, but is rarely used in clinical practice. 21

Multivascular disease

As noted earlier, there are a number of patients with cardiovascular disease who have disease in more than one vascular bed (otherwise known as multivascular disease patients). The REACH registry (supported by Sanofi–aventis, Bristol–Myers Squibb and the Waksman Foundation) collected data from approximately 67,888 patients who were recruited from 5473 physician practices in 44 countries worldwide. 17,22 Patients in the registry are described as being > 45 years old with least three atherothrombotic risk factors (e.g. treated diabetes mellitus, diabetic nephropathy, ankle–brachial index < 0.9, asymptomatic carotid stenosis of ≥ 70%) or documented cerebrovascular disease, coronary artery disease or peripheral arterial disease. A survey22 of data from the REduction of Atherothrombosis for Continual Health (REACH) registry identified that 15.9% of patients had symptomatic polyvascular disease, defined as coexistent symptomatic (clinically recognised) arterial disease in two or three territories (coronary, cerebral and/or peripheral). A further analysis indicated that rates of cardiovascular death, MI or stroke at 1 year increase substantially with the number of affected vascular beds. 2 This recognition of the importance of multivascular disease, problems with its definition, and its inherent increased risk of further events is explored in Chapter 3 (see Patients with multivascular disease).

Trends in coronary heart disease and stroke

Coronary heart disease causes over 90,000 deaths a year in the UK: approximately one in five deaths in men and one in six deaths in women. There is geographical variation in prevalence, with greater rates in the northern areas of England than in southern areas and intermediate rates in Wales. There are also social inequalities in mortality from coronary heart disease: higher mortality is noted in people from more deprived areas and those working in manual jobs. 1

Death rates from coronary heart disease have been declining since the late 1970s and death rates from stroke have declined in the last 10 years, although these trends appear to be plateauing, particularly in younger people. It is thought that the decline in rates of coronary heart disease is owing to reductions in risk factors (mainly smoking) and better treatment (including secondary prevention). Although mortality appears to be falling, coronary heart disease-related morbidity is rising. 1

Stroke accounts for around 53,000 deaths each year in the UK (approximately 9% of all deaths). According to the BHF8 it is not possible to know how many deaths each year are attributable to each stroke subtype. However, it reports that age-standardised mortality rates from stroke have decreased markedly in the last four decades, with a 90% reduction in ischaemic stroke mortality. 8 There is geographical variation in death rates from stroke in the UK; the highest rates are in Scotland, followed by northern England, Wales and Northern Ireland. The south of England (particularly London) exhibits the lowest stroke mortality rates. Socioeconomic inequalities in stroke mortality are evident; historically, rates have decreased more quickly in adults from higher social classes and mortality increases with deprivation. 8

The majority of people survive an initial stroke, but often have significant morbidity. 7 Stroke causes a greater range of disabilities than any other condition and has a greater disability impact than other chronic diseases. 23 It is thought that more than 900,000 people in England are living with the effects of stroke, with half of these being dependent on other people for help with everyday activities. 7

Impact of health problem

In 2006–7 there were 428,000 inpatient episodes for coronary heart disease in England and over 175,000 for stroke. 1,8 Data from 2006 underline the high cost of coronary heart disease and stroke to the UK health-care system: each cost around £3.2B. A cost per capita of just over £50 for each condition was observed. 1 Hospital care costs for coronary heart disease accounted for 73% of the total cost, whereas for stroke hospital costs accounted for 94%. 1

Production losses from death and illness and from informal care of people with coronary heart disease and cardiovascular disease are a substantial financial burden. 1 Data from 2006 for the UK demonstrate that production losses owing to mortality and morbidity associated with coronary heart disease cost over £3.9B: 65% owing to death and 35% owing to illness in those of working age. Informal care costs were approximately £1.8B. 1 For stroke, 65% of production losses were owing to illness and costs of informal care were £2.9M, reflecting the debilitating impact of stroke on individuals. 1

Study design

Only randomised controlled trials (RCTs) were included in the assessment of clinical effectiveness. Full economic evaluations were included in the assessment of cost-effectiveness.

The Assessment Group also identified and assessed the quality of existing systematic reviews (SRs) in order to cross-check for the identification of additional studies, as well as to gain an understanding of the issues related to the combining of data in this complex area. A summary and critique of relevant SRs is presented in Appendix 3.

Interventions and comparators

The effectiveness of two antiplatelet agents, used within their licensed indications, was assessed (1) clopidogrel alone and (2) MRD alone or in combination with ASA. Studies that compared clopidogrel alone or MRD (alone or in combination with ASA) with ASA or, where appropriate, with each other were included in the review. Trials in which clopidogrel was used as an adjunct to percutaneous coronary intervention were excluded from the review. Trials in which clopidogrel was combined with ASA were also excluded, as they were not within the remit of the scope. 16

Patient populations

For clopidogrel, patients with a history of MI, ischaemic stroke or established peripheral arterial disease were included. Patients with acute coronary syndromes were not included, and neither were those with atrial fibrillation. For MRD, patients with a history of ischaemic stroke or TIA were included.

Outcomes

Data on any of the following outcomes were included in the assessment of clinical effectiveness: MI, stroke, TIA, death and adverse events including bleeding complications. No data relating to health-related quality of life (HRQoL) or unstable angina were identified. For the assessment of cost-effectiveness, outcomes included incremental cost per life-years gained and incremental cost per quality-adjusted life-year (QALY) gained.

Direct evidence

The results of (1) clinical and (2) economic data extraction and quality assessment are summarised in structured tables and as a narrative description. The decision problem of interest to this review was made up of the following comparisons: (1) clopidogrel versus ASA; (2) clopidogrel versus MRD alone; (3) clopidogrel versus MRD + ASA; (4) MRD + ASA versus ASA; and (5) MRD alone versus ASA.

Indirect evidence

Owing to the differences between trials in terms of interventions and comparators, indirect analysis (using a mixed-treatment comparison methodology) was performed on a variety of outcomes. The methods and results of the mixed-treatment comparisons are reported below (see Methods for indirect synthesis).

Additional analysis by the Assessment Group

Using data provided by the manufacturers of clopidogrel, the Assessment Group undertook subgroup analysis and explored the clinical effectiveness of clopidogrel in patients with multivascular disease. The Assessment Group was also able to explore whether or not key outcome events are distributed evenly across the whole period of trial follow-up or if there are particular time points when patients appear to be at a greater risk.

Quality assessment of included randomised controlled trials

All of the included RCTs were of good quality (see Appendix 2). Robust randomisation procedures were used and baseline comparability between treatment groups was achieved. The use of blinding procedures was reported where appropriate and intention-to-treat (ITT) analyses were conducted for each trial. There was no evidence of selective reporting of outcomes in any of the trials.

Trial characteristics

The key characteristics of the included trials are summarised in Table 6. Of the four trials, three were double blind and one was an open-label study (ESPRIT56). The majority of trials were conducted globally, whereas the participating centres in ESPS-230 were located only in Europe. All trials included patients with ischaemic stroke as a qualifying event and two included patients with a qualifying event of TIA. 30,56 Only CAPRIE26 included patients with MI or peripheral arterial disease. The trial sizes ranged from 2763 to 20,332. Mean length of follow-up ranged between 1.91 and 3.5 years. Three trials were industry funded, while ESPRIT56 was funded from a variety of non-industry sources. Two trials (CAPRIE26 and ESPRIT56) utilised a composite as a primary end point, the components of which differed between the trials. In ESPS-2,30 three discrete primary end points were reported, while PRoFESS57 reported on a single primary end point of recurrent stroke. Across the four trials, ASA dosage ranged from 50 mg per day (ESPS-230 and PRoFESS57) to 30–325 mg per day in ESPRIT56 and 325 mg per day in CAPRIE. 26

Patient characteristics

The key characteristics of patients in the included trials are summarised in Table 7. The mean age of the patients was similar across trials. The percentage of males appears to be greatest in CAPRIE. 26 The PRoFESS57 trial included the greatest proportion of patients with hypertension and diabetes mellitus. None of the trials characterised the patient population in terms of the number of affected vascular beds, so the number of patients per trial with multivascular disease is unknown. However, the history of vascular events for the whole cohort of patients is reported for each trial; these are described in the right-hand column of Table 7. Compared with the other trials, in ESPS-230 there was a higher percentage of patients with peripheral arterial disease in addition to the qualifying event of ischaemic stroke/TIA. With the exception of CAPRIE,26 the modified Rankin Scale60 was used as a measure of patient disability; this scale is widely used as an outcome measure for stroke in clinical trials. The scale ranges from 0 to 6, where ‘0’ indicates no disability and ‘6’ is death. All patients in ESPRIT56 were rated as between 0 and 3, with 43% having no disability.

CAPRIE26

The key outcomes of the CAPRIE26 trial are described in Table 8. For the whole trial population, statistically significant outcomes in favour of clopidogrel were noted for the primary outcome (first occurrence of ischaemic stroke, MI or vascular death). The relative risk reduction (RRR) was 8.7% in favour of clopidogrel [95% confidence interval (CI) 0.3% to 16.5%; p = 0.043]. It has been noted,3 elsewhere, that the point estimate favoured clopidogrel, but this benefit appeared to be very small; the boundaries of the CIs raise the possibility that clopidogrel is not more beneficial than ASA. A statistically significant risk reduction (23.8%) in favour of clopidogrel was reported for the subgroup of patients with peripheral arterial disease (95% CI 8.9% to 36.2%; p = 0.0028); however, the trial was not powered to detect differences between patient subgroups and so the finding should be interpreted with caution. No statistically significant differences between clopidogrel and ASA were noted for the subgroup of patients with ischaemic stroke or MI.

ESPS-230

Table 9 shows the key outcomes of ESPS-2. 3,30 For the first primary outcome of stroke, statistically significant differences in favour of MRD + ASA were observed for two comparisons: MRD + ASA vs ASA [relative risk (RR) 0.76; 95% CI 0.63 to 0.93] and MRD + ASA vs MRD alone (RR 0.75; 95% CI 0.61 to 0.91). No difference was observed for the MRD-versus-ASA comparison. No other primary outcome (all-cause death, stroke and/or all-cause death) showed statistically significant differences between any two treatment arms.

Of the secondary outcomes, stroke/TIA, other vascular event, ischaemic events and vascular events, statistically significant differences were recorded in favour of MRD + ASA when compared with ASA (RR 0.80, 95% CI 0.70 to 0.92; RR 0.55, 95% CI 0.33 to 0.94; RR 0.77, 95% CI 0.65 to 0.92; RR 0.78, 95% CI 0.67 to 0.91, respectively).

Of the secondary outcomes of TIA, stroke/TIA, ischaemic events and vascular events, statistically significant differences in favour of MRD + ASA compared with MRD alone were noted (RR 0.80, 95% CI 0.66 to 0.97; RR 0.78, 95% CI 0.69 to 0.90; RR 0.76, 95% CI 0.64 to 0.90; RR 0.76, 95% CI 0.65 to 0.89, respectively).

ESPRIT56

The key outcomes of the ESPRIT56 trial are described in Table 10. For the primary outcome of first occurrence of death from all vascular causes, non-fatal stroke, non-fatal MI or major bleeding complication, the risk of event occurrence was statistically significantly lower in the MRD + ASA arm than in the ASA arm [hazard ratio (HR) 0.80; 95% CI 0.66 to 0.98].

For the secondary outcome of death from all vascular causes and non-fatal stroke, the rate of event occurrence was also statistically significantly lower in the MRD + ASA arm than in the ASA arm (HR 0.78; 95% CI 0.62 to 0.97). This was also true for the outcome of all vascular events (HR 0.78; 95% CI 0.63 to 0.97).

There were no statistically significant differences reported for any other outcome.

PRoFESS57

The key outcomes from the PRoFESS57 trial are described in Table 11. Although the rate of recurrent stroke of any type was very similar in the MRD + ASA and clopidogrel groups [9% vs 8.8%, HR 1.01 (95% CI 0.92 to 1.11)], the null hypothesis (that MRD + ASA is inferior to clopidogrel) could not be rejected as the predefined non-inferiority margin was –1.075.

For the secondary outcomes, the only statistically significant difference was in favour of MRD + ASA for the outcome of new or worsening congestive heart failure [HR 0.78 (95% CI 0.62 to 0.96)].

Adverse events

Adverse events reported for each trial are described in Table 12. In ESPS-230 and CAPRIE,26 bleeding events in the trials were reported as secondary outcomes rather than as adverse events. The reporting of adverse events differed between trials. In CAPRIE,26 adverse events were recorded as ‘patients ever reporting,’ in ESPS-230 as ‘number of patients reporting at least one adverse event during the study’. In PRoFESS,57 only selected adverse events leading to treatment discontinuation are presented in the published paper. Adverse events other than those related to bleeding were not reported for ESPRIT56 (see Table 10).

In CAPRIE,26 patients in the clopidogrel arm were reported as experiencing significantly higher rates of rash and diarrhoea than in the ASA arm. In the ASA arm, patients reported significantly more incidences of indigestion/nausea/vomiting and abnormal liver function. The number of patients experiencing gastrointestinal haemorrhage was greater in the ASA arm than in the clopidogrel arm, a result reported to be statistically significant. The rates of trial discontinuation because of adverse events were similar in both arms of the trial.

In ESPS-2,30 there was a significant difference between each arm in the occurrence of headaches. These appear to be greater in the arms where MRD was a feature of the treatment regimen. It is recorded in the published paper30 that bleeding episodes were significantly more frequent and more often moderate or severe/fatal in treatment arms that included ASA. Any site bleeding was reported by 8.2% of patients in the ASA arm and by 8.7% in the MRD + ASA arm, but by 4.7% and 4.5% in the MRD alone and placebo groups, respectively. The rates of trial discontinuation because of adverse events differed significantly, with higher rates reported in the two MRD arms than in the ASA or placebo arms.

Of the other reported adverse events in ESPS-2,30 gastrointestinal events, vomiting, diarrhoea and headache were reported as being significantly different between treatment groups, but where the differences lie are unclear. 30

In PRoFESS,57 the rates of trial discontinuation were statistically significantly different between trial arms in favour of clopidogrel. Notably, there was an increased risk of a major haemorrhagic event for MRD + ASA compared with clopidogrel (HR 1.15; 95% CI 1.00 to 1.32) as well as intracranial haemorrhage (HR 1.42; 95% CI 1.11 to 1.83). Headache appears to be reported by many more patients in the MRD + ASA arm – an unsurprising outcome, as MRD acts as a vasodilator.

First ischaemic stroke

Three trials (CAPRIE,26 ESPRIT56 and PRoFESS57) provided direct head-to-head data on the ‘first ischaemic stroke’. Therefore, it was possible to combine these trials through the mixed-treatment comparison approach to calculate the relative efficacy of clopidogrel versus ASA, MRD + ASA versus ASA and MRD + ASA versus clopidogrel.

Table 14 shows head-to-head trial data and the relative estimates calculated using the mixed-treatment comparison analysis. The results show no major differences between the mixed-treatment comparison results and head-to-head estimates from the included trials. Results from the mixed-treatment comparison showed that no single estimated RRs were found to demonstrate a statistically significant difference between any pair of interventions. The observed RR for clopidogrel and MRD + ASA appeared to reflect a lower risk of the ‘first ischaemic stroke’ compared with ASA. A RR of 0.968 was observed for MRD + ASA compared with clopidogrel. However, differences were not significant. There is no evidence to suggest that any intervention is superior to another in terms of prevention of ‘first ischaemic stroke’.

Any recurrent stroke

Two trials (ESPS-230 and PRoFESS57) provided direct head-to-head data on recurrent stroke outcome. Therefore, it was possible to combine these trials through the mixed-treatment comparison approach to calculate the relative efficacy of MRD + ASA versus ASA, MRD alone versus ASA, MRD + ASA versus clopidogrel, and MRD alone versus MRD + ASA. We were also able to estimate the indirect estimates from the mixed-treatment comparison for clopidogrel versus ASA and MRD versus clopidogrel. Table 15 presents head-to-head trial data and results from the mixed-treatment comparison analysis. No major differences in the mixed-treatment comparison results and head-to-head estimates from the included trials were observed. Results from the mixed-treatment comparison showed that clopidogrel and MRD + ASA were associated with fewer recurrent strokes relative to ASA. An increased risk of recurrent stroke was observed for MRD alone compared with clopidogrel or MRD + ASA. There was no difference between MRD alone compared with ASA or between MRD + ASA and clopidogrel, in terms of reducing recurrent stroke.

Any bleeding

Three RCTs (ESPS-2,30 ESPRIT56 and PRoFESS57) provided direct head-to-head data on any bleeding. It was possible to combine these trials through the mixed-treatment comparison approach to calculate the relative efficacy of MRD + ASA versus ASA, MRD alone versus ASA, MRD + ASA versus clopidogrel, and MRD alone versus MRD + ASA. We also calculated the indirect estimates for clopidogrel versus ASA and MRD alone versus clopidogrel, as no head-to-head data were available. The category of ‘any bleeding’ includes both minor and major bleeding. Minor events included haematuria, haematemesis, epistaxis, intraocular bleeding, purpura, and gynaecological, internal and intracranial bleeding. Major bleeding included severe or fatal bleeding, life-threatening bleeding, intracranial bleeding, major haemorrhage and major gastrointestinal tract haemorrhage. Table 19 shows head-to-head trial data and the estimates calculated using the mixed-treatment comparison analysis. There were no major differences in the mixed-treatment comparison results and head-to-head estimates from the included trials. Results from the mixed-treatment comparison showed that MRD alone was associated with significantly fewer bleeding events than all comparators; the MRD vs clopidogrel estimates are based on indirect comparisons and are not supported by head-to-head trial data. There was no evidence to suggest any differences between ‘clopidogrel versus ASA’ and ‘MRD + ASA versus ASA’ for any bleeding.

Major bleeding

Two RCTs (ESPRIT56 and PRoFESS57) provided direct head-to-head data on major bleeding. It was possible to combine these trials through the mixed-treatment comparison approach to calculate the relative efficacy of MRD + ASA versus ASA and MRD + ASA versus clopidogrel. We also estimated the indirect estimates for clopidogrel versus ASA as no head-to-head data were available. The category of ‘major bleeding’ included severe or fatal bleeding, life-threatening bleeding, intracranial bleeding, major haemorrhage and major gastrointestinal tract haemorrhage. Table 20 shows head-to-head trial data and the estimates calculated using the mixed-treatment comparison analysis. There were no major variations in the mixed-treatment comparison results and head-to-head estimates from the included trials. Results from the mixed-treatment comparison showed that clopidogrel was associated with significantly fewer bleeding events than ASA; these estimates are based on indirect comparisons and are not supported by head-to-head trial data. No statistically significant differences among MRD + ASA, clopidogrel and ASA in major bleeding events were observed.

Characteristics of economic evaluations

Five69,71,72,74,76 of the 1169–79 studies included were described as cost-effectiveness analyses and six70,73,75,77–79 as cost–utility analyses. The cost-effectiveness analyses have used a range of health outcomes including life saved, events avoided, life-years lived, time spent free of stroke recurrence or disability and life expectancy. All of the cost–utility analyses have used QALYs as the main measure of health outcome. As presented in Table 27, seven studies69,71,75–79 compared clopidogrel versus ASA; Karnon et al. 73 compared clopidogrel for the first 2 years followed by ASA indefinitely versus ASA; Chen et al. 72 compared clopidogrel + low-dose ASA versus ASA; Beard et al. 70 compared MRD + ASA versus MRD single agent, low-dose ASA, clopidogrel or no treatment; and Matchar et al. 74 compared placebo versus ASA, ASA + MRD or clopidogrel.

The study populations in the included studies were made up of patients with a history of cardiovascular disease (MI, ischaemic stroke, TIA or peripheral arterial disease); this matches the populations described in the key clinical trials used to derive efficacy data. Only one study78 explicitly considered patients with multivascular disease. The mean age varied according to the trial source used, ranging from 60 to 70 years. Only four studies70,73,77,78 described a UK population. Most of the studies adopted a lifetime perspective; however, four69,71,72,77 adopted a short-term perspective (e.g. duration of the clinical study follow-up).

Economic models

Only one of the included studies was not based on an economic model; Chen et al. 72 performed an economic evaluation using data from the CHARISMA59 trial without any survival projection beyond 28 months. Matchar et al. 74 used an individual sampling model based on a model previously developed for the secondary prevention of stroke. Berger et al. 71 adapted the model developed by Annemans et al. 69 and Beard et al. 70 based their model on the model developed by Cambers et al. 81 All relevant assumptions and extra information describing the models are summarised in Table 28.

Cost data and cost sources

All of the studies stated the currency used; five of them also included the currency year, which ranged from 2002 to 2007. Four studies used euros, three used pounds sterling and four used US dollars. The majority of the studies discussed cost items and provided useful definitions of costs. Drugs costs have been taken from a variety of different sources including local cost lists;69 published literature;71 the BNF;70,73 and the WEB of pharmacy wholesale suppliers. 74,75 Costs of acute events, including hospitalisations and acute care, have been taken from the trial-based papers,71,73 Medicare diagnosis-related groups data,74,75 NHS Trust Financial Return data70 and the published literature. 71,78 Only three papers76,77,79 do not state the sources of the cost data used. All papers but one74 have mentioned a discount rate for costs, as Table 29 shows.

Efficacy data and data sources

Only Palmer et al. 77 and Stevenson et al. 78 present data related to efficacy; the rest of the studies only point out that efficacy data are taken from a specific trial. Table 30 describes the information from the main trials used in each of the economic evaluations.

Health outcome data and data sources

Six of the economic evaluations used QALYs as the main measure of health outcome; other outcomes include life-years saved and life expectancy.

Only Matchar et al. 74 have not discounted health outcomes. In the study by Delea et al. 76 it is not clear if discounting has been applied to both costs and benefits. In the study by Palmer et al. ,77 discounting was used, but the discount rate is not explicitly stated. Health outcome information from the included studies is summarised in Table 30.

Cost-effectiveness ratios

The results of the cost-effectiveness analyses are described in Table 31. In summary, Annemans et al. 69 and Berger et al. 71 conclude that, for the overall population (MI, ischaemic stroke and peripheral arterial disease), clopidogrel is cost-effective compared with ASA, with an incremental cost-effectiveness ratio (ICER) of €13,390 per QALY and €14,380 per life-years saved (scenario 1) or €18,790 per life-years saved (scenario 2). Chen et al. 72 and Delea et al. 76 show an ICER of US$36,343 per life-years saved and a range of US$40,204 to US$49,107 per life-years saved, respectively, concluding that clopidogrel is cost-effective compared with ASA.

Schleinitz et al. ,75 Palmer et al. 77 and van Hout et al. 79 conclude that clopidogrel is cost-effective when compared with ASA (see Table 31), although Schleinitz et al. 75 also conclude that the current evidence does not support increased efficacy of clopidogrel in MI patients. Stevenson et al. 78 estimated that the mean cost per QALY for clopidogrel compared with ASA was £5443 in patients with a previous history of MI who then sustained an ischaemic stroke or a peripheral arterial disease event.

The evaluation by Beard et al. 70 concludes that MRD + ASA is a cost-effective option with an ICER below €5000 per QALY when compared with ASA or MRD alone, and it dominates when compared with clopidogrel or no treatment.

The study by Karnon et al. 73 concludes that the comparison of clopidogrel followed by ASA versus ASA yields an ICER of £21,489 per QALY.

Matchar et al. 74 show that placebo versus ASA and placebo versus MRD + ASA have similarly low ICERs; however, placebo versus clopidogrel yields a high ICER with a low probability of being cost-effective.

The majority of the trials have performed univariate and probabilistic sensitivity analysis. In general, the univariate sensitivity analyses show consistency around the ICER. All univariate sensitivity analyses are summarised in Appendix 8. Beard et al. 70 state that their model is sensitive to the long-term costs of very disabled patients. Matchar et al. 74 conclude that, although the simulations in their model can support the results shown, these are not sufficiently robust.

Summary of evidence and discussion

In general, the results of the literature review of cost-effectiveness evidence show that from a health service perspective, the use of clopidogrel in patients with previous peripheral arterial disease, ischaemic stroke or MI is a cost-effective option compared with ASA in the secondary prevention of occlusive vascular events. However, it is noted that Schleinitz et al. 75 conclude that current evidence does not support increased efficacy of clopidogrel in the MI patient group; this is the only evaluation that includes subgroup analysis to estimate ICERs by patients’ previous event. This is also the only study not funded by a pharmaceutical manufacturer (four papers76–79 did not provide details of industry affiliation).

The combination of MRD + ASA seems to be cost-effective compared with any other treatment (vs ASA, vs clopidogrel, vs no treatment) in patients with previous ischaemic stroke or TIA in the secondary prevention of occlusive vascular events. There is only one evaluation70 that includes this combination (MRD + ASA) and therefore the evidence base is limited.

Although model structures are similar, the length of the cycles differs from one study to another and the assumptions regarding the transition probabilities (e.g. Annemans et al. 69 – life expectancy assumptions) are not always reliable. Data in the models are from a broad variety of sources, which makes it difficult to pool the results and make definitive conclusions.

All evaluations except three71,72,78 were published prior to 2006; this means that more recent trials and papers have not been used to inform the economic evaluations (e.g. clinical data from PRoFESS,57 REACH17 or MATCH58 are not described in the papers). The relevance of this cost-effectiveness review to decision-making is therefore limited as the economic evaluations are not based on the most up-to-date clinical data.

Overview of submitted manufacturer’s submission

A Markov model was designed to assess the cost-effectiveness of MRD + ASA versus ASA alone, clopidogrel and no treatment for the secondary prevention of occlusive vascular events in patients who have experienced:

• an ischaemic stroke and are tolerant of ASA

• a TIA and are tolerant of ASA.

The model is based on the model developed by the Technology Appraisal Group to inform the previous guidance. 3

The structure of the manufacturer’s model is shown in Figure 4.

FIGURE 4.

Schematic structure of the Boehringer Ingelheim’s model.

The model estimates costs from the perspective of the UK NHS, and health outcomes in terms of life-years and QALYs in a simulated cohort of 1000 patients initially aged 45–80 years using a time horizon of 2.5–50 years and a cycle length of 6 months.

Costs and benefits have been discounted at a rate of 3.5% per annum.

The model presents five health states:

1. no recurrent stroke

2. recurrent ischaemic stroke

3. haemorrhagic stroke

4. vascular death

5. non-vascular death.

Patients enter into the model in the ‘no recurrent stroke’ health state, from where they may move to any other state or remain in the same state. From the ‘recurrent ischaemic stroke’ state patients may move to ‘haemorrhagic stroke’, ‘vascular death’ or ‘non-vascular death’ or remain in the ‘recurrent ischaemic stroke’ state. In the ‘haemorrhagic stroke’ state, patients will either remain in this state or die. Once patients enter the ‘haemorrhagic stroke’ health state, any additional recurrent haemorrhagic stroke events are not recognised in the model. The manufacturer states that this restriction is introduced to avoid the situation where an additional event (e.g. new ischaemic stroke) leads to a patient’s utility state improving. If multiple events occur in a single cycle, one event is given priority in allocating patients to a health state in the following order of descending priority: death, haemorrhagic stroke, ischaemic stroke. The model also includes two tunnel health states: ‘other haemorrhagic events’ and ‘new or worsening congestive heart failure’.

Summary of clinical effectiveness data

Transition probabilities during the first 4 years are derived from different trials for each of the arms:

• MRD + ASA and clopidogrel – PRoFESS57 trial

• ASA alone – combination of ESPRIT56 trial and ESPS-230 trial

• no treatment – ESPS-230 trial.

Beyond the first 4 years, the transition probabilities are assumed to remain constant at the values of the last monthly cycle of the fourth-year period for the following transitions:

• new recurrent ischaemic stroke from the ‘no recurrent stroke’ state

• haemorrhagic stroke from the ‘no recurrent stroke’ state

• haemorrhagic stroke from the ‘new recurrent ischaemic stroke’ state.

The manufacturer used published data from the Oxfordshire Community Stroke Project90 and the Lothian Stroke Registry91 to estimate the overall death rate among stroke patients compared with the general population. A multiplier of 1.5 was used to generate an overall expected age-related death rate beyond the trial period from the Office for National Statistics death rate data for the general population. The vascular and non-vascular death rates beyond the 4 years of the trial were assumed to sum to this rate.

The manufacturer has assumed that those patients who have experienced a TIA had a rate of previous ischaemic stroke events equal to 80% of those who had experienced a previous ischaemic stroke. This assumption is made on the basis of the previous multiple technology assessment (MTA)3 in which the Assessment Group made the same assumption.

Summary of costs and resource use

Summary of submitted results

The base-case analysis includes second-line treatment with ASA for those patients discontinuing first-line treatment in clopidogrel and MRD + ASA groups. A summary of the results is shown in Table 36 for ischaemic stroke patients and in Table 37 for TIA patients.

Summary of sensitivity analysis

Probabilistic sensitivity analysis

After generating 500 iterations, the results for the probabilistic sensitivity analyses were as follows:

• ischaemic stroke patients MRD + ASA versus clopidogrel: MRD + ASA has more than 90% probability of being cost-effective at a threshold of £30,000 per QALY.

• TIA patients MRD + ASA versus clopidogrel: MRD + ASA has more than 90% probability of being cost-effective at a threshold of £30,000 per QALY.

Critique of Boehringer Ingelheim’s economic model by the Assessment Group

The submitted model considers a wide range of treatment alternatives and describes a wide range of resources to populate the model. The model is mainly based on the PRoFESS57 trial, although some data have been taken from ESPS-230 and ESPRIT56 to obtain probability transitions in the ischaemic stroke group. The transition probabilities during the first 4 years for the MRD + ASA and clopidogrel arms are derived from the above-mentioned trials, and beyond that point they have used the same transition probability as used for the last 6-monthly cycle. This is an unreliable basis for long-term projection, as close to the end of the trial patient numbers and the number of events are much reduced. As a consequence, estimated incidence rates are very volatile and should not be relied on to drive the major part of the model calculations.

Death rates amongst patients who have had strokes have been derived from two main papers;90,91 when these papers were checked, the figures quoted in appendix 9 of the manufacturer’s submission do not clearly match with those in the published papers. In relation to the TIA incidence rates, the manufacturer has assumed that patients who experienced TIAs had a rate of ischaemic stroke events equal to 80% of those who had experienced a previous ischaemic stroke; there is no evidence to support this assumption and it has not been tested in the one-way univariate sensitivity analysis.

The design of the model also includes tunnel health states to model adverse events. The tunnel health states are not depicted in the manufacturer’s submission and are poorly addressed in the excel model. The manufacturer’s submission is sometimes hard to follow because of several mistakes in the appendices notation (e.g. manufacturer’s submission, p. 27, section 3.2.1) and within the Excel model (e.g. Overview spreadsheet E35 cell in the excel model says 10 years’ time horizon instead of 50 years). The figure describing the model (manufacturer’s submission, p. 25) has two arrows from ‘no recurrent stroke’ health state to ‘non-vascular death’, which is not consistent with the structure described.

The parameter distributions of costs used in the probabilistic sensitivity analysis are not commonly used distributions and their use is not justified by the manufacturer.

The manufacturer states that ‘MRD + ASA long term first line is cost-effective against clopidogrel … Based on these ICERs at a threshold of £20,000 per QALY, it remains cost-effective until clopidogrel drops by 45% of brand price for ischaemic stroke patients or 51% for TIA patients’ (manufacturer’s submission, pp. 41–2). The Assessment Group notes that the generic price of clopidogrel as listed in the Electronic drug tariff 33 March 2010 is £10.90 (30 × 75-mg tablets); this constitutes a 69% reduction in price [branded Plavix (£36.35) was used in the model] and means that compared with MRD + ASA, clopidogrel is cheaper and more effective for both ischaemic stroke and TIA populations.

Overview of submitted manufacturer’s submission

A Markov model is designed to assess the cost-effectiveness of clopidogrel, MRD + ASA, ASA and MRD alone for the secondary prevention of occlusive vascular events: MI, ischaemic stroke and vascular death. Cost-effectiveness estimates are calculated for four different patient populations:

• patients who have previously suffered a MI

• patients who have previously suffered an ischaemic stroke

• patients who were diagnosed with peripheral arterial disease

• patients with multivascular disease which is described as ischaemic disease in more than one vascular bed.

The same model structure is used throughout, but the baseline risks of vascular events differ for each population. The four treatments under consideration are compared against each other only in the ischaemic stroke population; whereas, in the MI, peripheral arterial disease and multivascular disease populations only clopidogrel is compared with ASA.

The model estimates costs from the perspective of the UK NHS and health outcomes in terms of life-years and QALYs. A cohort of 1000 patients with the qualifying diagnosis (MI, stroke peripheral arterial disease or multivascular disease) and aged 65 years progresses through the model over a time horizon of 35 years. The starting age of 65 years was chosen as the average age in the PRoFESS57 trial was 66.1 years, in CAPRIE26 62.5 years and in REACH17 68.6 years. The cycle length is 3 months and only one event can occur in each cycle. The model structure is depicted in Figure 5. Costs and benefits have been discounted at a rate of 3.5% per annum.

FIGURE 5.

Diagram of the Markov model. ACS, acute coronary syndrome; ASP, aspirin; CLOP, clopidogrel; PAD, peripheral arterial disease; TAG, technology assessment guidance.

The model utilises six health states (see Figure 5):

1. Initial state This is the starting condition for all patients, and is considered to be a ‘stable’ state.

2. Death Separately recorded for deaths of non-vascular and vascular origin.

3. History of MI The condition of patients following a non-fatal MI.

4. History of stroke The condition of patients following a non-fatal ischaemic stroke.

5. History of MI and stroke The condition of patients who have suffered both a non-fatal MI and a non-fatal stroke.

6. TA80 state This intermediate state relates to the TA80 guidance,45 which recommends that treatment with clopidogrel+ASA should be continued for up to 12 months (four cycles in the model) after the most recent acute episode of NSTEMI. In the model, after four cycles, patients go back to antiplatelet monotherapy.

All adverse events are included in the cost and QALY calculations, but are not recorded separately as distinct health states or events in the model.

Each patient population (MI, stroke, peripheral arterial disease and multivascular disease) progresses through the model subject to its specific risk profile and parameters depending on previous history. The presence of previous vascular events thus influences the risk of future health states. Patients in the model can either remain stable or experience a MI or a stroke or death (from vascular or non-vascular causes). Deaths within 30 days of a new MI or stroke are defined as vascular deaths and such patients will progress directly to death.

Summary of effectiveness data

The baseline risk of events related to ASA has been taken from the REACH17 registry and from a network meta-analysis of six studies: ESPS-2,30 ESPRIT,56 CAPRIE,26 MATCH,58 CHARISMA59 and PRoFESS. 57 The REACH17 registry recruited a large international cohort of patients (n = 68,236) with either established atherosclerotic arterial disease or at least three risk factors for atherothrombosis, and considered the outcomes of cardiovascular death, non-fatal MI and non-fatal stroke. The event rates were different for year 1 (REACH registry17), year 2 (unpublished – academic in confidence) and year 3 (published online101). The model assumes the 3-year data to be applicable for all subsequent years (years 3–35).

The manufacturer has constructed a matrix to allocate the correct risk of events to patients as they change health states through the model, such that state- and population-specific event rates and probabilities are assigned. This trace matrix is reproduced in Table 43.

The REACH17 event risks are assumed to be applicable to a population treated with ASA, as 67% of registry patients received ASA monotherapy. Aspirin was chosen to be the treatment of reference to which the three other comparators are modelled. The relative treatment effects of the other three treatments (MRD, MRD + ASA and clopidogrel) versus ASA have been estimated based on direct estimates from clinical trials or indirect estimates from the network meta-analysis of the six studies mentioned above. The network meta-analysis was conducted for the end points – stroke, MI, vascular death, non-vascular death, and major and minor bleeding events.

The base case in the model considers all ASA arms in the network meta-analysis studies to have equal efficacy. Non-vascular death rates have been derived from life tables. The non-vascular mortality rate is estimated by removing deaths owing to the diseases of circulatory system from age-specific deaths from all causes.

The following assumptions were used by the manufacturer in the model:

• Non-vascular death was assumed to be the difference between ‘all-cause mortality’ and ‘death from vascular causes’.

• When fatal and non-fatal vascular events were not reported separately then the total of fatal and non-fatal events was used as an approximation for non-fatal events in the dataset.

• In the absence of any evidence on non-vascular death having a dose–response relationship with ASA (in contrast with the vascular events and adverse events), it was assumed that the risk of non-vascular death was equal for all ASA doses.

• As the ESPRIT56 trial did not impose a specific ASA dose, but left the decision on dosing to the local investigators, the ASA arm of this trial was assumed to be a weighted average of the low, medium and high ASA dose arms, with weights equal to the proportion of patients observed on the different doses: 46%, 48% and 5%, respectively.

• The Antithrombotic Trialists’ Collaboration (ATTC) data66 describing the efficacy of ASA versus no treatment reported only on the composite end point of ‘serious vascular events’ but not on the separate components. Therefore, the assumption was made that the relative efficacy of ASA versus no treatment was equal for all these separate end points: MI, stroke and vascular death.

The model presents six different effectiveness analyses derived from the above sources:

1. network meta-analysis with the six studies above and ASA doses pooled (base case)

2. network meta-analysis splitting up the ASA comparator into three separate comparators: low-, medium- and high-dose ASA

3. head-to-head analysis based solely on the PRoFESS57 trial

4. head-to-head analysis based solely on the CAPRIE26 trial

5. head-to-head analysis based on post hoc analysis on multivascular disease patients from CAPRIE26 trial.

To estimate the efficacy of clopidogrel + ASA in the TA8045 state versus ASA, data from the CURE27 trial have been used.

Summary of adverse events data

Baseline risk of adverse events relating to ASA has been derived from three papers: one meta-analysis66 and two RCTs. 26,30

The risk of a major bleeding event is taken from a meta-analysis of RCTs of antiplatelet therapy. 66

The risk of minor bleeding event is derived from the ESPRIT56 trial. The risk of dyspepsia is taken from the ESPS-230 trial comparing ASA with MRD and a combination of MRD + ASA for the secondary prevention of stroke.

Summary of costs and resource use

Summary of results

Summary of sensitivity analysis

The manufacturer has reported a deterministic scenario analysis using the different efficacy analyses included in the model. In the stroke population, clopidogrel is dominated by MRD + ASA in all of the possible efficacy analyses, and with or without treatment effect for non-vascular death. Clopidogrel is shown to be cost-effective when compared with ASA using CAPRIE26 data only in both treatment-effect scenarios for non-vascular death (Table 51).

The ICERs for the other populations (MI, peripheral arterial disease and multivascular disease) also change slightly with the assumption concerning the treatment effect for non-vascular death in each of the efficacy analyses, resulting in clopidogrel appearing cost-effective with an ICER of < £30,000 per QALY. The best results for clopidogrel are in multivascular disease patients using data from the post hoc CAPRIE26 trial efficacy analysis.

In summary, the cost-effectiveness of treatments for the secondary prevention of occlusive vascular events is sensitive to a range of different scenarios. Removing the treatment effect on non-vascular deaths is found to improve the cost-effectiveness estimates of clopidogrel. Cost-effectiveness is also found to be sensitive to the efficacy estimates: taking account of different ASA doses worsens the cost-effectiveness estimates, whereas using only a head-to-head analysis based on the CAPRIE26 trial improves them. The estimates in the stroke population are least sensitive to a head-to-head analysis using the PRoFESS57 trial.

A probabilistic sensitivity analysis was developed by the manufacturer using a Monte Carlo simulation undertaking 3000 iterations. At a threshold of £30,000 per QALY, the treatment option with the highest probability of being cost-effective in MI, peripheral arterial disease and multivascular disease populations is clopidogrel and in stroke it is MRD + ASA, as Table 52 shows.

In stroke patients, the average incremental net benefit of clopidogrel when compared with ASA is –£6 with an associated 95% CI of –£6320 to £7279.

The probabilistic sensitivity analysis in MI patients reports an incremental net benefit of £1187 (CI –£7692 to £10,260). The cost-effectiveness acceptability curve (CEAC) shows that for a threshold of £30,000 per QALY clopidogrel is cost-effective in 60% of the iterations.

For patients with peripheral arterial disease, the probabilistic sensitivity analysis estimates an average incremental net benefit of clopidogrel versus ASA of £1475 (CI –£6106 to £9476). The CEAC suggests that there is a 64% probability that, at a threshold of £30,000 per QALY, clopidogrel would be considered a cost-effective treatment for the prevention of occlusive vascular events.

For patients with multivascular disease, the average incremental net benefit of clopidogrel versus ASA is £1748 (CI –£5475 to £9179) and the CEAC suggests that there is a 68% probability of clopidogrel being cost-effective at a threshold of £30,000 per QALY.

Critique of Sanofi–aventis/Bristol–Myers Squibb’s economic model

The manufacturer of clopidogrel has presented ‘new’ evidence of the clinical effectiveness and cost-effectiveness of clopidogrel on a set of four re-allocated patient populations (stroke, MI, peripheral arterial disease and multivascular disease); this means that none of the effectiveness results used in their modelling of cost-effectiveness are directly derived from publications from the CAPRIE26 trial. The review group accepts that this new categorisation is more appropriate and results in better-defined and less-heterogeneous patient groups. However, the details that would be required to construct and populate a long-term disease model based on CAPRIE26 are not available beyond the summary statistics presented in the manufacturer’s submission.

The Assessment Group notes that the generic price of clopidogrel as listed in the Electronic drug tariff33 March 2010 is £10.90 (30 × 75-mg tablets); this constitutes a 69% reduction in price [branded Plavix (£36.35) was used in the model]. Using this new price in the model improves the cost-effectiveness of clopidogrel.

The manufacturer’s model is depicted in Figure 5 and includes one health state called ‘TA80 acute coronary syndrome’, which represents treatment after a MI following the TA8045 guidelines in the treatment of patients with NSTEMI. This document refers only to NSTEMI patients, yet the manufacturer’s submission does not differentiate between STEMI and NSTEMI patients so the model does not reflect clearly the recommended treatment of patients following a MI.

The baseline event rates in the ASA arm are taken from the REACH17 registry, whose population is a mixed population of patients with history of MI, stroke or peripheral arterial disease and patients with risk factors of cardiovascular disease. The original scope issued by NICE does not mention risk factors, only history of previous events. Also, these baseline event rates have been applied to patients in the ASA group; however, only 67% of the population of the REACH17 registry have received ASA monotherapy.

The model assumes different transition probabilities every year until year 3. Beyond this point the last-cycle transition probabilities are used for the remainder of the time horizon from years 3 to 35. This is an unreliable basis for long-term projection, as close to the end of the trial patient numbers and the number of events are much reduced. As a consequence, estimated incidence rates are very volatile and should not be relied on to drive the major part of the model calculations.

Calculations used to derive utilities are adequately described in the manufacturer’s submission, but sometimes differences between adverse events utilities are not clearly explained (e.g. decrement utility after major bleed and minor bleed: there is a substantial difference between them which is not discussed). Also, utility values are calculated using an assumption of perfect health for patients before the event 1 (‘utility’ spreadsheet in the model) and this is inappropriate.

In the model, half-cycle correction and discount rate methodologies have been applied incorrectly; this affects the final results of the model and overestimates the number of QALYs generated.

Approach to modelling occlusive vascular events

Modelling disease-related health and the economic effects of chronic lifetime conditions presents additional and different challenges to those encountered when dealing with conditions of an acute or time-limited nature. In particular, over a lifetime, patients are subject to multiple interacting competing risks of fatal and non-fatal events, and the accumulation of complex and dynamic health histories with a resulting dynamic pattern of prognostic risks. To overcome these challenges the Assessment Group has chosen to develop a new model of occlusive vascular events involving individual patient sampling. Instead of considering patients in aggregated groups with average characteristics, we generate a series of individual patients whose combined characteristics are representative of the specified population. The advantage of this approach is that individual patient histories can be generated according to a number of known competing risks, so that interactions are automatically accounted for.

Obtaining these advantages often involves significant technical costs in terms of complex programming and long processing times, which involve the use of very large numbers of random numbers in order to achieve stable results. To reduce these difficulties the Assessment Group has designed the model structure to operate within a Microsoft Excel workbook with limited additional coding and incorporating several ‘variance reduction’ techniques.

Patient populations

Four mutually exclusive patient populations are modelled using the following definitions:

• MI only This population is defined as patients suffering a recent acute MI, who may have a prior history of ischaemic heart disease, but have no prior history of ischaemic stroke, TIA or peripheral arterial disease.

• ischaemic stroke/TIA only This population is defined as patients suffering a recent ischaemic stroke or TIA, who may have a prior history of ischaemic cerebrovascular events, but have no prior history of ischaemic heart disease (including MI) or peripheral arterial disease.

• Peripheral arterial disease only This population is defined as patients suffering a recent episode of peripheral arterial disease, but who have no prior history of ischaemic stroke or TIA, or ischaemic heart disease (including MI).

• Multivascular disease This population is defined as patients suffering a recent episode of acute MI, ischaemic stroke or TIA, or peripheral arterial disease, and who have a prior history involving at least one other type of vascular disease.

In order to characterise each of these populations in terms of age and gender, an analysis of data from the Health Survey for England 1996108 has been carried out, using data on self-reported chronic health conditions to identify samples corresponding to the four modelled populations (Table 53). (Note: the Health Survey for England 1996 was commissioned by the Department of Health and carried out by the Joint Surveys Unit of Social and Community Planning Research and the Department of Epidemiology and Public Health at University College London, who bear no responsibility for the analysis or interpretation of its data presented in this report.)

Treatment strategies

It is clear from the available evidence56,57 that a significant proportion of patients do not persist with the medication initially prescribed, either because of unacceptable adverse events associated with the drug or for other personal or lifestyle reasons. When discontinuation occurs, it is necessary to prescribe an appropriate alternative treatment if one is available; as a consequence, the effect of treatment on future risks will be modified. It is therefore necessary to assess the effectiveness and cost-effectiveness of preventive medicines within the framework of lifetime treatment strategies. Tables 54 and 55 set out the treatment strategies that may be compared using the economic model for each patient population.

Model design

The logic flow for generating a full patient history for each sampled patient is shown in Figure 6 for the first two key events. As event times are estimated as continuous variables, it is not possible for a conflict to arise with two events occurring simultaneously. Subsequent events repeat the same pattern. Each patient continues to accumulate additional events until a fatal event is encountered.

FIGURE 6.

Patient sampling model flow chart for a sequence of key events within a single patient history.

Modelling transient ischaemic attack

The Technology Assessment Report3 that led to the development of the current guidance24 on secondary prevention of occlusive vascular events included some consideration of patients suffering from TIA despite the absence of separate trial information for the effectiveness of either treatment for this patient group. A simple assumption was made that TIA patients were at risk of future events at a reduced (80%) rate compared with ischaemic stroke patients. This failed to take into account two published papers presenting results from the Oxfordshire Community Stroke Project, showing the risk of stroke following a first-ever stroke90 or following a TIA. 116

More recently a Canadian population study117 provided similar findings for TIA patients. Table 60 does not suggest that there is strong evidence to make a distinction between TIA patients and those surviving an ischaemic stroke. On this basis it has been assumed that TIA patients may be subsumed within the stroke model population, as long-term risks appear to be similar.

Technology Appraisal No. 8045 guidance and the myocardial infarction population

On the basis of evidence from the CURE27 trial, NICE guidance document TA8045 recommends that patients surviving a NSTEMI event should receive clopidogrel and low-dose ASA as medication for the prevention of further MI events for a period of 12 months, followed by low-dose ASA alone thereafter. There is no current guidance for surviving STEMI patients beyond the immediate post-MI period.

The only clinical trial evidence submitted for the current appraisal relating to the MI-only patient population is from a subgroup of the CAPRIE26 trial population, which involves a mix of STEMI and NSTEMI patients. No analyses are provided in the CAPRIE26 clinical study report distinguishing between STEMI and NSTEMI patients.

Similar concerns apply to the multivascular disease population, as a proportion of these patients may have MI as the qualifying event. No information is available on the composition of the multivascular disease group in CAPRIE26 by qualifying event so it is difficult to determine how any meaningful subdivisions could be applied.

As reviewing the existing TA80 guidance45 and CG48 guidelines7 is not within the scope of this appraisal, it is necessary to assume that recommendations for post-MI preventive treatment of both NSTEMI and STEMI patients remain valid. However, it would be inappropriate to begin modelling MI-only patients while still subject to these short-term provisions (12 months for NSTEMI and 4 weeks for STEMI patients). We therefore assume that all MI-only patients have survived to the end of the specified period without suffering a further MI, or any other occlusive vascular event (which would require them to be reclassified as multivascular disease patients), prior to embarking on the chosen long-term preventive treatment strategy. This avoids the necessity of identifying MI patients as either STEMI or NSTEMI from the outset.

Technology Appraisal No. 8045 and subsequent myocardial infarction events in all populations

In all four populations defined above there is a risk of future MI events, some of which will be non-fatal. Therefore, the TA80 guidance45 requires that the affected patients (i.e. those suffering an NSTEMI event) should be switched to clopidogrel + ASA for 12 months. For modelling it becomes necessary to estimate the probability of NSTEMI versus STEMI to assign the correct post-event short-term treatment, although none of the available trials provides information on the type of MI suffered. The GRACE (Global Registry of Acute Coronary Events)118 study of acute coronary syndrome patients is used to estimate the proportions of STEMI/NSTEMI in the population as 53.8%/46.2% (MIs excluding unstable angina). To accommodate the effects of TA80 guidance45 in the model, a simplification has been applied, which involves a reduction to the short-term post-MI risk that was estimated from the CAPRIE26 data to reflect the benefits observed in CURE,27 and a corresponding short-term increase in treatment costs for the 12 months post MI, both averaged by the STEMI/NSTEMI proportions in the GRACE118 study.

In addition, the follow-on treatment after 12 months (ASA alone or ‘standard care’) needs to be interpreted in the context of the model treatment strategies. Where an ‘MI-only’ patient suffers subsequent MI events, but no other type of occlusive event, treatment may resume at the stage of the treatment strategy prior to the latest MI(s) requiring short-term follow-up. If an ‘MI-only’ patient suffers a different kind of occlusive event, they attract the higher risks associated with multivascular disease patients for the remainder of their life. In the same way a ‘stroke-only’ or ‘peripheral arterial disease-only’ patient suffering a MI will also be subject to the higher multivascular disease risks once the short-term follow-up care is complete. Equally, an ‘MI-only’ or ‘peripheral arterial disease-only’ patient suffering an ischaemic stroke may receive up to 2 years’ MRD + ASA treatment as required by TA90,24 and, subsequently, resume the long-term care strategy subject to the increased multivascular disease event risks.

Technology Appraisal No. 9024 and subsequent ischaemic stroke events in all populations

The National Institute for Health and Clinical Excellence TA90 guidance24 recommends the use of MRD + ASA for up to 2 years following a non-fatal ischaemic stroke event. The Assessment Group model has been adapted to reflect this feature, which may be rendered active or inactive at the user’s discretion. The adaptation involves introducing a pseudoevent at the end of the TA9024 recommended treatment period, before the patient resumes at his or her prior stage in the assigned treatment strategy. This is an effective mechanism for coping with the added complexity of TA90 guidance. 24

However, it does result in some potential loss of integrity in the matching of random number sequences between comparator model runs (a mechanism used for ‘variance reduction’ in the model); in principle this might introduce some element of bias into the results, but it would occur in only the later stages of a patient’s career when many patients have already died and appears more likely to underestimate incremental differences than to overestimate them. A simple test of this effect is to compare model results with and without this feature activated, as the model results obtained when the TA9024 feature is inactive are not subject to any potential bias. To date the Assessment Group has not detected any evidence of any bias affecting the decision analysis results.

A note of caution is necessary here against attempting to use a comparison of model results with and without the TA9024 feature turned on as a means of reconsidering the validity of TA90 guidance. 24

As currently constructed, the model would not be valid for this purpose and would require important modifications to achieve such an objective. As this is not within the scope of the current appraisal, no effort has been made to pursue this possibility.

Deterministic analysis

Tables 61 and 62 summarise the main economic results obtained with the Assessment Group model for the ischaemic stroke patient population. Figures 7–10 illustrate these findings in the form of a cost-effectiveness plane plot with cost-effectiveness frontier. The primary analysis (first block in Tables 61 and 62 and Figure 7) reveals that only two strategies lie on the boundary, but neither of these involves initial use of clopidogrel. In all scenarios, the most cost-effective strategy begins with MRD + ASA, followed by ASA and finally clopidogrel.

FIGURE 7.

The cost-effectiveness plane and frontier showing available treatment strategies for ‘ischaemic stroke-only’ patients (using MRD + ASA as per the TA9024 guidance). A = ASA; B = ASA to MRD + ASA; C = MRD + ASA to ASA to clopidogrel; and D = MRD + ASA to clopidogrel to ASA.

FIGURE 8.

The cost-effectiveness plane and frontier showing available treatment strategies for ‘ischaemic stroke-only’ patients (without applying the TA9024 guidance). A = ASA; B = ASA to MRD + ASA; C = MRD + ASA to ASA to clopidogrel; and D = MRD + ASA to clopidogrel to ASA.

FIGURE 9.

The cost-effectiveness plane and frontier showing available treatment strategies for ‘ischaemic stroke-only’ patients (using MRD + ASA as per the TA9024 guidance and the generic clopidogrel price). A = ASA; B = ASA to clopidogrel to MRD + ASA; C = MRD + ASA to ASA to clopidogrel; D = MRD + ASA to clopidogrel to ASA; and E = clopidogrel to MRD + ASA to ASA.

FIGURE 10.

The cost-effectiveness plane and frontier showing available treatment strategies for ‘ischaemic stroke-only’ patients (without applying the TA9024 guidance and using the generic clopidogrel price). A = ASA; B + ASA to clopidogrel to MRD + ASA; C + clopidogrel to ASA to MRD + ASA; and D = clopidogrel to MRD + ASA to ASA.