Drug-eluting stents for the treatment of coronary artery disease

4.1.1 A total of 17 randomised controlled trials (RCTs) were identified that compared DESs with BMSs, and data from all 17 were included for at least one outcome in the meta-analysis.

4.1.2 Study outcomes used in the RCTs included rates of mortality, acute MI, target lesion revascularisation (TLR), target vessel revascularisation (TVR), composite events (major adverse coronary event [MACE] and/or target vessel failure [TVF]), angiographic binary restenosis and late luminal loss. Revascularisation was usually prompted by protocol-driven angiographic evidence of restenosis either for all participants or for a selected subgroup of participants. Only one trial (BASKET) explicitly reported that no protocol-driven angiographic follow-up was included. This trial compared both SES (Cypher) and PES (Taxus) DESs with a newer BMS in a three-arm study.

4.1.3 All but three of the 17 RCTs were multicentre trials. Study size ranged from 60 to over 1300 patients. Of the 17 trials, 11 included patients with single lesions. The studies covered a range of vessel diameters from 2.25 mm to 4.00 mm, although the lower range was not reported in some studies. Lesion length also varied, ranging from 10 mm to 33 mm, although again the data were not always reported. All studies permitted the inclusion of people with diabetes, and all but three studies excluded acute or evolving MI. The presence of unprotected left main coronary artery excluded patients from many trials, as did severe calcification or tortuosity, total occlusion, bifurcation, the presence of thrombus in the target vessel, previous PCI within 30 days or PCI other than balloon required as part of the study intervention.

4.1.4 A total of 12 trials described the co-therapies used. Aspirin was prescribed before intervention in 11 of these studies and used after the procedure in all 12. Clopidogrel was used as an antiplatelet therapy in all of the 12 studies; ticlopidine was available for use as an alternative to clopidogrel in five studies. In one trial, tirofiban, used in combination with a DES, was compared with abciximab used with a BMS. The duration of antiplatelet therapy after intervention ranged from 2 months in three trials to 1 year in one study.

4.1.5 Meta-analysis was carried out by the Assessment Group for rates of mortality, acute MI, TLR, TVR, composite event (MACE and/or TVF), angiographic binary restenosis and late luminal loss. Analysis of mortality, acute MI and event rates used pooled results from over 7000 participants. Data in the form of an odds ratio (OR) and 95% confidence interval (95% CI) were analysed using the Mantel–Haenszel method fixed-effect model. For continuous outcomes, weighted mean differences (WMDs) were analysed. Where there was significant heterogeneity, analysis using a random-effects model was also undertaken.

4.1.6 In addition to analyses of the individual studies, pooled estimates (giving the OR and 95% CI) were provided for each 'eluted drug' group (for example, comparing a PES [Taxus] and all BMSs in the paclitaxel studies). Data related to the SES (Cypher) and the sirolimus-analogue stent, ZES (Endeavor) in some instances were pooled and presented as pooled SES results. All eluted drug group results were also pooled to obtain estimates for a meta-analysis of any-type DESs compared with any-type BMSs. The meta-analysis was performed for available data at follow-ups of up to 1 month, 6 to 9 months, 1 year, 2 years and 3 years. The Assessment Group assumed, when making decisions about the appropriateness of combining data, that all BMSs are similar and, likewise, all DESs are similar except in the drug delivered; and that the stent design and the insertion system do not have an impact on clinical outcomes.

4.1.7 For rates of mortality and rates of acute MI, one study found a statistically significant difference in favour of the SES (Cypher) compared with the BMS for MI at 6 to 9 months (OR 0.19, 95% CI 0.04, 0.87). There were no statistically significant differences between the DES and the BMS in the individual studies for all other follow-up periods analysed to 3 years. There were no statistically significant differences between the DES and the BMS for the pooled eluted drug groups (PES [Taxus] and pooled SESs) and for the pooled analyses of any-type DES compared with any-type BMS for any of the follow-up periods.

4.1.8 For event rate (MACE and TVF), the individual studies of PES (Taxus), SES (Cypher) and ZES (Endeavor), and the pooled eluted drug groups analysis showed statistically significant differences in favour of DESs over BMSs. This was also the case for the overall meta-analysis, which favoured any-type DESs over any-type BMSs at all follow-up time points: 6 to 9 months (OR 0.46, 95% CI 0.40 to 0.53), 1 year (OR 0.39, 95% CI 0.33 to 0.47), 2 years (OR 0.43, 95% CI 0.34 to 0.54) and 3 years (OR 0.42, 95% CI 0.32 to 0.55). Statistical heterogeneity was indicated at the 6 to 9 months follow-up; a random-effect analysis for this time point showed only a small effect on the OR (OR 0.44, 95% CI 0.36 to 0.54). The difference between the EES (Xience V) and the BMS was not statistically significant at the only follow-up period (6 to 9 months).

4.1.9 For TVR, not all individual studies of PES (Taxus) showed statistical significance compared with BMSs for all time periods up to 3 years. The individual studies for SESs (Cypher) and ZESs (Endeavor) all showed statistical significance over BMSs up to 3 years. The pooled eluted drug groups analysis showed statistically significant differences in favour of a PES (Taxus) over BMSs at follow-up time points up to 2 years: 6 to 9 months (OR 0.54, 95% CI 0.43 to 0.68), 1 year (OR 0.40, 95% CI 0.29 to 0.55) and 2 years (OR 0.45, 95% CI 0.34 to 0.59). At 3 years, the difference was no longer statistically significant, but the data at this time point were derived from a single, relatively small study that may have been underpowered. TVR data for a SES (Cypher) versus a BMS were available for two trials at 6 to 9 months and for single trials at 1 and 3 years. These showed statistically significantly differences in favour of the SES (Cypher) compared with the BMS at: 6 to 9 months (OR 0.33, 95% CI 0.18, 0.62), 1 year (OR 0.34, 95% CI 0.19 to 0.60) and 3 years (OR 0.35, 95% CI 0.25 to 0.49). TVR data for the ZES (Endeavor) at 6 to 9 months, the only time period available, was statistically significant in favour of the ZES over the BMS (OR 0.41, 95% CI 0.27 to 0.63). There were no data for EES (Xience V) for this outcome measure.

4.1.10 Rates of revascularisation (TLR) at 1 year for procedures carried out with a DES within individual trials were less than 5%, and typically in the 10% to 25% range for procedures that used a BMS. For example, in three trials of PES (Taxus), the rates were 0%, 4.7% and 4.2% for the DES compared with 10.0%, 12.9% and 14.7% for the BMS, respectively. Rates at 1 year in three trials of a SES (Cypher) were 4.6%, 0% and 4.9% for the DES compared with 24.9%, 13.6% and 20.0% for the BMS, respectively. For TLR, the pooled eluted drug groups analysis showed statistically significant differences in favour of a PES (Taxus) over BMSs at follow-up periods of up to 2 years: 6 to 9 months (OR 0.37, 95% CI 0.28 to 0.49), 1 year (OR 0.26, 95% CI 0.18 to 0.39) and 2 years (OR 0.28, 95% CI 0.20 to 0.40). At 3 years, the difference was no longer statistically significant, but the data at this time point were derived from a single, relatively small study that may have been underpowered. TLR data for a SES (Cypher) showed it to be statistically significantly more effective than a BMS at all time points up to 3 years: 6 to 9 months (OR 0.21, 95% CI 0.15 to 0.30), 1 year (OR 0.17, 95% CI 0.12 to 0.25), 2 years (OR 0.22, 95% CI 0.15 to 0.30) and 3 years (OR 0.25, 95% CI 0.17 to 0.36). The data for the ZES (Endeavor) at the follow-up period of 6‑9 months showed it to be statistically significantly more effective than the BMS (OR 0.35, 95% CI 0.22, 0.56). Lower rates of TLR (3.8% versus 21.4%) were apparent for the EES (Xience V) group at 6 months (the only follow-up period) but the difference was not statistically significant. For TLR, the meta-analyses showed statistically significant differences in favour of any-type DES over any-type BMS, with improved rates of lesion revascularisation at all follow-up time points up to 3 years: 6 to 9 months (OR 0.30, 95% CI 0.25 to 0.37), 1 year (OR 0.21, 95% CI 0.16 to 0.27), 2 years (OR 0.24, 95% CI 0.19 to 0.31) and 3 years (OR 0.25, 95% CI 0.17 to 0.35).

4.1.11 The TES (Janus) was examined in a non-controlled study as was the PES (Taxus Liberté). A range of formulations of the PES (CoStar) was evaluated in two non-randomised controlled studies, and the Yukon DES was evaluated in a dose-ranging non-randomised controlled study comparing Yukon coated with sirolimus with the same stent carrying no drug. The Dexamet DES was studied in one non-randomised study of Dexamet compared with a BMS and four non-controlled studies (including two registries).

4.1.12 Outcome data were limited due to the short follow-up periods: 30 days for the PES (Taxus Liberté) study; 4 months for one PES (CoStar) study and 1 year for the other PES (CoStar) study; 6 months for the Dexamet studies and for the TES stent; and up to 1 year for the SES (Yukon) study. Angiographic outcomes, binary restenosis and/or late loss were reported for the PES (CoStar), Dexamet and the SES (Yukon). Because of the variety of DESs considered in these studies, the methodological limits of the available studies, and the varied and limited follow-up, the Assessment Group did not consider pooled analysis to be appropriate.

4.1.13 For the TES (Janus), limited data were reported; at 30 days no events (death, MI or TLR) had occurred. For the PES (Taxus Liberté) the data available at 30 days were marked as commercial in confidence. For the PES (CoStar), the only data available were for one of the two arms at 1 year for one trial and interim data from two of the four arms of the other ongoing study.

4.1.14 Data for the SES (Yukon) were reported at 1 month and 1 year. No deaths occurred in the first month. Rates of acute MI up to 1 month were 1.8% in the SES group and 1.3% in the BMS group. At 1 year, the composite of death or non-fatal MI was 2.7% for the Yukon SES and 3.9% for the BMS. No statistically significant differences were detected. At 1 year, TLR was reported in 12.6% of the SES group and 21.5% of the BMS group, and the difference was statistically significant in favour of the SES (OR 0.53, 95% CI 0.34 to 0.81).

4.1.15 The non-randomised trial that compared Dexamet (DES) with a BMS reported no deaths among the 100 participants receiving either stent and only one incidence of acute MI, which was in the BMS group, up to a mean of 8 months' follow-up. Revascularisations for this time period were 2% TLR in the DES group and 10% TLR (12% TVR) in the BMS group. Composite rates of MACEs, consisting entirely of revascularisations, were 2% for the DES and 12% for the BMS. Neither of these comparisons showed statistically significant differences.

4.1.16 Eight RCTs comparing different DES types were identified by the Assessment Group. Six RCTs compared a SES (Cypher) with a PES (Taxus) (including one trial that was also assessed in the DES versus BMS clinical section because it had a BMS arm as well as two DES arms), one studied SES (Cypher) in comparison with the newer SES (Cypher Select) and one compared the Yukon, as a SES, with a PES (Taxus).

4.1.17 Six trials were conducted in only one or two centres in European countries, and two were multicentre and multinational. Study sizes ranged from 200 to 1350 patients. Two trials were distinct in that they did not incorporate planned angiographic assessment of trial participants. Only two trials presented randomisation details and none of the studies presented adequate information on whether the RCTs were conducted under 'blind' conditions. One study did not present an intention-to-treat analysis, and for two studies it was unclear whether events were reported according to the original randomised allocations.

4.1.18 A meta-analysis was conducted according to which pairing of DES types was compared within trials (most commonly this was the SES [Cypher] versus the PES [Taxus]). No total pooled effect estimate was calculated across multiple groupings of DES versus DES trials.

4.1.19 There were no statistically significant differences in rates of mortality or acute MI for any of the pairings of DES types.

4.1.20 For TLR, one individual study showed a statistically significantly better rate of TLR, at 6 to 9 months, with the SES (Cypher) compared with the PES (Taxus) (OR 0.56, 95% CI 0.33 to 0.93). Only one RCT had data available beyond 9 months; in this study, rates of TLR at 1 year were 5.7% for the SES (Cypher) compared with 9.0% for the PES (Taxus); the difference was not statistically significant. The Assessment Group's pooled analysis of TLR up to 9 months was statistically significant in favour of the SES (Cypher) over the PES (Taxus) (OR 0.70, 95% CI 0.51 to 0.97).

4.1.21 A statistically significant reduction in TVR with the SES (Cypher) compared with the PES (Taxus) was determined from a meta-analysis of two trials at 6 to 9 months (OR 0.59, 95% CI 0.39 to 0.89). A reduction in the composite event rate (MACE) at 6 to 9 months was also statistically significant with the SES (Cypher) compared with the PES (Taxus) (OR 0.75, 95% CI 0.59 to 0.96).

4.1.22 In December 2006, following publication of data on longer-term risks associated with DES (thrombosis, MI and mortality), the FDA convened a public meeting of its Circulatory System Devices Advisory Panel to review and analyse the available data and to provide recommendations for appropriate actions to address this issue. In January 2007, the Circulatory System Devices Advisory Panel made recommendations to the FDA. The Panel stated, 'When the DES, which are indicated for use in the USA (SES [Cypher]) and (PES [Taxus]), are used in accordance with their approved indications both are associated with a small increase in stent thrombosis compared with BMS at 1 year after stent implantation; the increased risk of stent thrombosis was not associated with an increased risk of death or MI compared with BMS; and the concerns about thrombosis do not outweigh the benefits of DES compared with BMS when DES are implanted within the limits of their approved indications for use.' The FDA stated that a longer duration of antiplatelet therapy may be beneficial, and this has led to the recommendation in the PCI guideline of the American College of Cardiologists/American Heart Association (endorsed by the Society for Cardiovascular Angiography Interventions) that in patients receiving DESs the duration of clopidogrel use should be increased to 12 months. The BCIS has also recommended 12 months' use of clopidogrel in patients receiving a DES.

4.1.23 The Circulatory System Devices Advisory Panel also considered use of DESs in patients with more complex coronary lesions than those studied to support initial marketing approval ('off-label' use). The Panel agreed that use of DESs 'off-label' is associated with an increased risk of stent thrombosis, MI or death compared with 'on-label' use, and until more data are available DES labels (IFUs) should state that when DESs are used off-label, patient outcomes may not be the same as the results observed in the clinical trials conducted to support marketing approval. The FDA has since defined off-label use to mean the use of a medical product for treatments other than those for which the product was initially approved; or use not explicitly included in product labelling (intended use and IFU). The UK Medicines and Healthcare Products Regulatory Agency (MHRA) supports this definition of off-label use for the DESs that have been approved for use in Europe.

4.1.24 Each DES included in this appraisal has an IFU document that lists the indications for which it can be used. The sizes of vessels (diameter and length) to be treated are stated in the majority of the IFUs, as are the specific contraindications. The FDA considers that although patients with diabetes were included in the pivotal trials, the number of patients was insufficient for either the SES (Cypher) or the PES (Taxus) to earn a specific labelled indication for people with diabetes. The UK MHRA supports the view of the FDA with regard to individuals with diabetes and only one of the DESs, the PES (Taxus), included in this appraisal has recently been specifically indicated for people with diabetes.

4.1.25 There were no statistically significant differences detected in death or MI between the pooled subgroups and pooled any-type DES groups. The pooled DES analysis indicated that revascularisation rates were reduced by approximately three quarters compared with BMSs, consistent across most studies of the PES (Taxus) and SESs (Cypher; Endeavor at 6 to 9 months). The benefits of DESs over BMSs for TLR were seen at 1 year, and this significant difference was maintained for up to 3 years. For the TVR outcome there were statistically significant differences in favour of any-type DES over BMS for most of the time points assessed.

4.2.1 A total of 10 full economic evaluations were included in the assessment report, all of which compared an SES with a BMS, although four evaluations also included a PES. One of the evaluations was conducted in the UK; the rest were conducted in the USA, Canada or the rest of Europe. Seven evaluations used a 1-year time horizon, one used 2 years, one used 6 months and one used a patient's lifetime. Of the 10 evaluations, nine estimated that the cost of DESs incurred a price premium/difference (the difference in cost between a given BMS and the drug-eluting equivalent), which ranged from £233 to £1225. Four of the evaluations reported health outcomes in terms of quality-adjusted life years (QALYs). Three evaluations provided incremental costs per QALY for a general population, and these costs ranged from US$27,450 to Can$96,523 (approximately US$93,000). The fourth evaluation did not include a general population because subgroups were found to be too dissimilar for comparison. Two evaluations reported the incremental cost-effectiveness ratio (ICER) per repeat revascularisation avoided; one estimated it to be US$1650 over 1 year and the other estimated it to be approximately US$7000 over 2 years. The majority of evaluations concluded that DESs are more cost effective than BMSs for patients with types of arteries that have a higher risk of restenosis, although there was great disparity between evaluations, with a variety of outcomes and a range of ICERs being reported.

4.2.2 Only one economic study, carried out alongside the BASKET RCT, reflected clinical practice because no protocol-driven angiographic follow-up was included. This study's results suggested that, at a threshold of €7800 per MACE avoided, DESs could potentially be cost effective in the following subgroups of patients: those older than 65 years; those with more than one segment treated; those with triple-vessel disease; those with a stent length of more than 20 mm; and those with small stent diameters.

4.2.3 Three models were submitted by DES manufacturers.

4.2.4 The decision analytic model from Boston Scientific compared the PES (Taxus) with the equivalent BMS, for a general population and for subgroups. The incremental costs per QALY at 1 year were given as £29,587 for the overall population and £1020 for patients with diabetes. For patients with small vessels and long lesions, the PES (Taxus) was dominant (both more effective and less costly than the BMS). For the PES (Taxus) at 2 years, the incremental cost per QALY for the overall population was given as £13,394, and it was dominant for patients with small vessels and those with diabetes. The model was highly sensitive to variations in the duration of clopidogrel therapy and the average number of stents used. In the manufacturer's sensitivity analyses, when the number of stents used per procedure was increased from 1.4 to 1.7, in line with the Assessment Group's model, the estimated cost per QALY at 1 year for the overall population increased to £56,731; however, the subgroup estimates were only marginally increased. If the duration of clopidogrel therapy after DES implantation was increased from 6 to 12 months, the cost per QALY at 12 months increased to £71,634 for the total population and to over £30,000 for the subgroup with diabetes.

4.2.5 The decision analytic model from Cordis compared the SES (Cypher) with the equivalent BMS for a 'no-risk-factor' population and for subgroups. The model was split into a two-way analysis of the BMS versus the SES (Cypher) and a three-way analysis of the BMS versus the PES (Taxus) versus the SES (Cypher). In extending the three-way analysis to 2 years, an indirect comparison was undertaken that made an assumption that the BMSs in both trials (Boston Scientific and Cordis BMS) are equivalent. The cost data for the technologies (the BMS and Taxus) were considered by the Assessment Group to be overestimated and when the Assessment Group re-ran the model increasing the SES (Cypher) price premium over the equivalent BMS from £433 to £695, the incremental cost per QALY increased from £29,259 to £69,613 for the 'no risk factor' subgroup; from £10,178 to £39,508 for the small-vessels subgroup; from £16,460 to £49,345 for the long-lesions subgroup; and from £9702 to £38,446 for the group with diabetes.

4.2.6 The Markov model presented by Medtronic compared the ZES (Endeavor [sirolimus analogue ABT-578]) with Medtronic's comparable BMS, for a total population. The submission measured costs and benefits at 5 years, extrapolating the 9-month trial data. In one scenario, the two arms were assumed to be equivalent in terms of risk of repeat revascularisations after 1 year. The incremental cost per QALY was estimated at £11,220. No subgroup analyses were undertaken. The Assessment Group found the model to be highly sensitive to baseline TVR rates and the number of index stents used. If baseline TVR rates were reduced below 12% (for both the BMS and the ZES), then the ICER exceeded £30,000. If the average number of stents used for the index procedure was increased to 1.4, the ICER increased to £39,174. The Assessment Group noted that the two factors to which the model was sensitive were taken from a single positive trial.

4.2.7 The submission from Kiwimed compared an SES (Yukon) with the Kiwimed BMS for a total population. The effectiveness data were taken from the SES (Cypher) trials, so an untested assumption was made that the SES (Yukon) has equivalent effectiveness to the SES (Cypher). Extrapolation from 2 to 5 years was undertaken using an assumption that patients remained in the same health state that they were in at the end of 1 year. Kiwimed's submission stated that the model results indicated that the SES (Yukon) was dominant compared with the BMS in the total population. In a sensitivity analysis varying the cost of the stent and the probability of restenosis, the ICERs for the SES (Yukon) were always under £30,000 per QALY.

4.2.8 The Assessment Group's model used the framework from the original appraisal with some minor modifications as follows: the time horizon was restricted to 1 year, so no discounting was necessary; particular risk groups were examined; in addition to the modelling of any-type DES compared with BMS, some head to head comparisons were conducted (SES (Cypher) compared with the PES (Taxus).

4.2.9 A difference between the effectiveness of DES and BMS was only seen with regard to revascularisation (TLR and TVR) and event rate (MACE and TVF). For these endpoints, the clinical trials show evidence of differences between DESs and BMSs at 1 year. The clinical trials show evidence in favour of DESs over all follow-up periods up to 3 years with a trend towards the greatest benefit occurring within the first year.

4.2.10 In the Assessment Group's model the most important factors in determining the incremental cost were the additional cost per DES implanted (price premium/difference) and the number of stents implanted per patient. The most important factors in determining benefit in the model were the absolute risk of revascularisation for patients treated with a BMS and the risk reduction attributable to the use of a DES.

4.2.11 The acquisition cost of a given stent may vary in different settings because of negotiated procurement discounts. The Assessment Group in their economic evaluation used the prices from a market survey of NHS purchasers. The survey was conducted by the NHS Purchasing and Supply Agency in May/June 2005 to identify the prices in contracts covering the period 2004/05 for both DESs and BMSs. The combined data from 12 purchasing bodies covering 20 hospital trusts provided consistent estimates of average unit prices and of the differences in price between DESs and BMSs. Results were provided for the two main suppliers of DESs: Boston Scientific (Taxus) and Cordis Corporation (Cypher). The effective sale price per Taxus PES (excluding VAT) was £815. Because there was only one recorded instance of a significant local volume discount agreement for Cypher in the survey, the average sample price for the Cypher SES (excluding VAT) was £937. The estimated average price for a BMS in the survey (excluding VAT) was £278, so the price differences are £537 and £659 per DES for Taxus and Cypher, respectively.

4.2.12 Information received during the consultation period for the appraisal consultation document suggested that the prices for DESs had decreased since the 2005 survey. Therefore updated prices were sought from the NHS Purchasing and Supply Agency. From the sample 2007/08 data received from the NHS Purchasing and Supply Agency for NHS organisations in England for the stents included in this appraisal, the mean price of DES was £529 and the mean price of BMS was £131. The price difference between DESs compared with BMSs ranged from £203 to £615 across a number of Health Authorities in England.

4.2.13 To calculate the PCI procedure costs it was necessary to subtract the included costs of stents (DES and BMS) from the published PCI costs, and then to add back the model estimates of the number of stents, the type of stent and the cost per stent. In the final analyses, the Assessment Group assumed a wastage rate of 1%.

4.2.14 The Assessment Group base-case analysis used results from two observational studies of stented patients treated at the Liverpool Cardiothoracic Centre, to convert the efficacy of any-type DES to effectiveness estimates for repeat revascularisations and lesions treated in repeat revascularisations. The Assessment Group found that 51% of patients who underwent a second PCI required repeat treatment to previously treated lesions only. An additional 17% of patients received repeat treatment to a target lesion at the same time as treatment to a previously untreated lesion in the same vessel; these are the patients in whom DESs can be expected to produce benefit. Applying these proportions to the relative risk reduction from the trials of 74.6% (for TLR obtained from the meta-analysis of any-type DES trials) yielded an expected risk reduction in all revascularisations at 12 months of between 38% (95% CI 32 to 44%) and 50% (95% CI 44 to 57%). The Assessment Group also considered the likely benefit that any-type DES may offer in reducing the number of lesions treated in repeat revascularisations. When applied to the TLR relative risk reductions from the meta-analysis, this suggested that the reduction in the number of lesions treated in subsequent interventions was between 37% (95% CI 31 to 42%) and 53% (95% CI 47 to 59%) based on TLR (the Assessment Group counted lesions treated but excluded cases undergoing CABG rather than PCI). The Assessment Group noted that the BASKET trial, which did not include protocol-driven angiography, reported a risk reduction for DESs of 41% at 6 months. The Assessment Group therefore used 41% for the risk reduction associated with DESs in its base-case analyses.

4.2.15 The Assessment Group stated that the ICERs of any-type DESs compared with any-type BMSs are very dependent on the estimates of relative risk reduction in revascularisations and on the absolute rate of revascularisation in the types of patients in whom they are used. The absolute rates of revascularisation were derived from the Liverpool Cardiothoracic Centre audit data, and the potential of benefit was reassessed on the basis of the audit data concerning those patients in whom the repeat procedure required treatment of new lesions. This resulted in an absolute rate of revascularisation for all patients of 7.43%. The Assessment Group also carried out sensitivity analyses, varying rates of revascularisation for all patients up to 13%, based on trial data.

4.2.16 Using the Liverpool Cardiothoracic Centre audit data, the Assessment Group developed separate risk models for elective and non-elective patients using patient and lesion characteristics known at the time of the index intervention. Risk factors for the patients were identified in the assessment report as being calcification, angulation greater than 45 degrees, restenotic lesion, triple-vessel disease, vessel diameter of less than 2 mm and prior CABG. In the final analyses, the absolute rates of repeat revascularisations for the conventional risk factors (long lesions, small vessels, diabetes and all combinations of these) were provided using the Liverpool audit data. The Assessment Group stated that, because none of these three factors in the multivariate model achieved conventional significance using the Liverpool audit data, the individual relative risks have wide confidence intervals and should be considered only as illustrative. The mean 12-month repeat revascularisation rate for all patients with small vessels was 15.25% (95% CI 9.38% to 22.24%), with a rate difference p = 0.02; for those with long lesions it was 10.23% (95% CI 8.10% to 12.57%), p = 0.09; and for those with diabetes it was 10.61% (95% CI 7.52% to 14.14%), p = 0.14.

4.2.17 The Assessment Group used patient survey data from the Health Outcomes Data Repository (HoDAR) database for its utility values. The difference in HoDAR health-related quality of life scores between patients with severe angina and those recovered from revascularisation (0.158) was similar to the ARTS trial result (0.16), which was used in the previous appraisal. The assumptions made by the Assessment Group in the final analyses for disutilities associated with CABG versus PCI in the 6-week period following the procedure were that for a 2-week post-operative period, patients undergoing CABG experience a very low quality of life (0.0), and for the next 2 weeks the mean utility score recovers in a linear fashion achieving full benefit (0.660) by 4 weeks after the operation. Patients undergoing PCI were assumed to recover full benefit linearly over a 2-week period following the procedure. The Assessment Group concluded additionally that there was no evidence to suggest an effect on the rate of acute MI or mortality with CABG compared with PCI plus stenting.

4.2.18 When BMSs and DESs were compared, the meta-analysis showed a trend towards increased numbers of non-fatal acute MIs with BMSs. The Assessment Group concluded in their analyses that based on the reviewed evidence, the maximum likely effect of this is equivalent to, per patient, an overall cost saving of about £13 and a utility gain of about 0.00055 when DESs are used.

4.2.19 The clinical evidence from the meta-analyses in the assessment report suggested that the SES (Cypher) reduces repeat revascularisations compared with the PES (Taxus). Because most of the clinical trials were limited to 6 to 9 months' duration, the Assessment Group carried out the economic evaluation assuming clinical equivalence and distinguished between stents only on the basis of price.

4.2.20 In the additional analyses requested by the Committee, the Assessment Group undertook sensitivity analyses by varying a number of the parameters in the original model. The results of the sensitivity analyses were presented as a number of tables containing the ICERs for a range of price premiums, for a range of absolute risks of revascularisation of BMSs, for the total population of stented patients and the risk-factor groups (small vessel, long lesion, diabetes and all combinations of these). Tables were presented for different numbers of stents (mean number, 1, 2 or 3) per patient. The Assessment Group also undertook new sensitivity analyses that took account of an additional 9 months' use of clopidogrel in patients receiving DESs, in accordance with the recent BCIS recommendations. These additional costs were applied only to 56% of the patient population because it was suggested by Consultees that 44% of patients would have acute coronary syndrome and therefore already be receiving 12 months' treatment with clopidogrel in accordance with 'Clopidogrel in the treatment of non-ST-segment-elevation acute coronary syndrome' (NICE technology appraisal guidance 80).

4.2.21 The Assessment Group, at the request of the Committee, also provided estimates based on the analyses using a relative risk reduction with DESs of 55% as the base case and a sensitivity analysis of 65%. These relative risk reductions with DESs were used for a rate of revascularisation, using BMSs, of 11%, which was obtained from combining results for all patients (equivalent to 10% for elective and 13% for non-elective patients). The corresponding risk of revascularisation using BMSs for the risk groups and the mean number of stents were also calculated from the combined datasets for the elective and non-elective patients.

4.2.22 Following consultation, the Assessment Group provided two additional analyses. The first additional analysis updated the 2003/04 resource costs with reference costs from 2005/06. The waiting times for PCI and CABG were updated from 20 to 20.5 weeks for PCI and from 13 to 20.9 weeks for CABG. The number of patients with acute coronary syndrome was changed from 44% to 48.5%, and consequently the number of patients receiving an extra 9 months' treatment with clopidogrel was reduced from 56% to 51.5%.

4.2.23 In addition to the changes to resource costs, waiting times and number of patients with acute coronary syndrome, the second additional analysis included new parameters based on alternative suggestions from the BCIS. The Assessment Group modified BCIS's suggestion regarding the presentational case-mix. The following parameters were used in the model: the percentages of elective and non-elective patients were changed from 68% to 57% for elective patients and from 32% to 43% for non-elective patients; the absolute risk of revascularisation of BMS for all patients was changed from 11% to 13% by combining 11.5% of elective and 15% on non-elective patients from the Liverpool Cardiothoracic Centre audit dataset; the relative risks for revascularisation in high-risk groups were set to 1.75 for small vessels, 1.35 for long lesions and 1.52 for diabetes; and the relative risk reductions from using DES were set to 60% for all patients; 69% for patients with small vessels; 70% for patients with long lesions and 61% for patients with diabetes.

4.2.24 Based on the Liverpool Cardiothoracic Centre audit data, the Assessment Group's original base-case scenario as described in 4.2.20 assumes the overall repeat revascularisation rate for the total population of stented patients in the UK at 12 months after PCI with BMSs is 7.43%. Using 7.43% for all patients, the absolute rates of repeat revascularisation for the risk factors become: 7.8% for long lesions; 9.0% for diabetes; and 9.9% for small vessels. Using the overall mean number of stents implanted per patient from the Liverpool Cardiothoracic Centre audit data (1.615) and assuming a price difference of £600 (approximate average of the price difference of the PES Taxus and the SES Cypher, from the survey data) the resulting incremental costs per QALY for each of the groups of elective patients are approximately: £407,000 for all patients; £380,000 for long lesions; £340,000 for diabetes; and £306,000 for small vessels. Using the overall mean number of stents implanted per patient for non-elective patients (1.467) the resulting incremental costs per QALY for each of the groups, at a price difference of £600, are: £282,000 for all patients; £250,000 for long lesions; £353,000 for diabetes; and £94,000 for small vessels.

4.2.25 Based on the Liverpool Cardiothoracic Centre audit data, the Assessment Group's re-analysis of the base-case scenario assumes an 11% overall revascularisation rate for the total population of stented patients in the UK at 12 months after PCI with BMSs. Using 11% for all patients, the absolute rates of repeat revascularisation for the risk factors become: 11.7% for long lesions; 11.6% for diabetes; and 19% for small vessels. The relative risk reduction with DESs is assumed to be 55%. Using the overall mean number of stents implanted per patient from the Liverpool audit data, both elective and non-elective (1.571) and assuming a price difference of £600 (approximate average of the price difference of the PES [Taxus] and the SES [Cypher], from the survey data) the resulting incremental costs per QALY for each of the groups of patients are approximately: £213,000 for all patients; £183,000 for long lesions; £180,000 for diabetes; and £146,000 for small vessels. Assuming a price difference of £300 the resulting incremental costs per QALY for each of the groups of patients are approximately: £101,000 for all patients; £85,000 for long lesions; £84,000 for diabetes; and £59,000 for small vessels.

4.2.26 For the Assessment Group's sensitivity analysis for the combined elective and non-elective data, using a relative risk reduction with DESs of 65% and a price difference of £600, the resulting incremental costs per QALY for each of the groups of patients are approximately: £174,000 for all patients; £148,000 for long lesions; £146,000 for diabetes; and £116,000 for small vessels. Assuming a price difference of £300 the resulting incremental costs per QALY for each of the groups of patients are approximately: £78,000 for all patients; £65,000 for long lesions; £64,000 for diabetes; and £41,000 for small vessels.

4.2.27 The Assessment Group's additional analysis as described in 4.2.22, assuming a price difference of £600 for the base case (55% relative risk reduction of DES), results in incremental costs per QALY for each of the groups of patients of approximately: £171,000 for all patients; £158,000 for long lesions; £156,000 for diabetes; and £124,000 for small vessels. Assuming a price difference of £300 the resulting incremental costs per QALY for each of the groups of patients are approximately: £74,000 for all patients; £66,000 for long lesions; £65,000 for diabetes; and £41,000 for small vessels. For the updated Assessment Group's sensitivity analysis (65% for the relative risk reduction of DES) the resulting incremental costs per QALY, for a price difference of £600, for each of the groups of patients are approximately: £137,000 for all patients; £126,000 for long lesions; £124,000 for diabetes; and £95,000 for small vessels. Assuming a price difference of £300 the resulting incremental costs per QALY for each of the groups of patients are approximately: £54,000 for all patients; £47,000 for long lesions; £46,000 for diabetes; and £25,000 for small vessels.

4.2.28 The Assessment Group's second additional analysis, as described in 4.2.23, resulted in incremental costs per QALY for each of the groups of patients, assuming a price difference of £600, of approximately: £111,000 for all patients; £51,000 for long lesions; £53,000 for diabetes; and £59,000 for small vessels. Assuming a price difference of £300 the resulting incremental costs per QALY for each of the groups of patients are approximately: £38,000 for all patients; £3,000 for long lesions; £4,000 for diabetes; and £4,000 for small vessels.