4 Evidence and interpretation

The Appraisal Committee (Appendix A) considered evidence from a number of sources (Appendix B).

4.1 Clinical effectiveness

4.1.1 The Assessment Report presents three sets of comparison: PCI without stents versus PCI with BMS; PCI with BMS versus CABG; and BMS versus DES.

4.1.2 Assessment of the relative clinical effectiveness of stents considers the likelihood of restenosis discovered on follow-up and the requirement for repeat intervention (revascularisation). Repeat intervention may occur: (i) because of a requirement of the trial protocol, specifying a repeat angiographic examination at a predetermined interval (so-called 'protocol-driven' reintervention); or (ii) following a recurrence of symptomatic angina in the patient (that is, 'clinically-driven' reintervention). The frequency of protocol-driven reinterventions is higher than that of clinically driven reinterventions, because angiography is usually mandatory at 6 months in the trial protocol whereas in clinical practice it is carried out only after recurrence of symptoms. Accordingly, the absolute differences observed between the treatment and control arms of clinical trials are likely to be higher than would be expected to occur in clinical practice.

PCI with BMS versus PCI (without stents)

4.1.3 Fifty randomised controlled trials (RCTs) were analysed comparing the use of PCI with BMS versus PCI without stents. Because of differences in, and completeness of, the reporting of these trials, the number of trials on which meta-analyses are based is a subset of these 50 trials. In a meta-analysis comprising 12 trials involving 5700 patients with non-specified ischaemic heart disease, where a composite endpoint of revascularisation, MI or death (MACE) was reported, the MACE rate was statistically significantly different at 6 months' follow-up: 23.0% for the PCI without stents group versus 15.4% for the PCI with BMS group, with an odds ratio (OR) of 1.66 (95% confidence interval [CI], 1.45 to 1.90). The difference was smaller after 12 months' follow-up but still statistically significant. Of the above 12 trials, seven, involving 3500 patients, reported data for 12-months' follow-up: the MACE rate was 22.0% for the PCI without stents group versus 18.9% for the PCI with stents group (OR 1.33; 95% CI, 1.12 to 1.58).

4.1.4 Differences in MACE rates were due almost entirely to differences in the rate of restenosis. For the outcomes of acute MI and deaths, for which individual trials were not powered to detect statistically significant differences, meta-analyses showed that while both of these sets of events occurred less frequently in those treated with PCI using BMS than in PCI without stents, in neither case was the result statistically significant.

4.1.5 Overall, the results of the RCTs showed that the use of PCI with BMS has significant advantages over the use of PCI without stents, in terms of lower rates of restenosis at 6 and 12 months.

4.1.6 According to the joint professional submission, the likelihood of restenosis is greater in small vessels, because a given tissue regrowth will have a greater proportionate effect in a vessel of smaller calibre. In eight out of nine studies that looked at vessels of a small calibre (less than 3 mm), restenosis rates were lower in the BMS arm than in the PCI without stent arm, and in two of these studies, the difference was statistically significant. A meta-analysis showed a statistically significant advantage for PCI with BMS. However, restenosis rates were still high in this group.

4.1.7 According to the joint professional submission, the restenosis rate increased by an estimated 8 to 13 percentage points with every 10-mm increase in the length of BMS required.

PCI with BMS versus CABG

4.1.8 There were six RCTs in the meta-analysis. None of the trials involved a DES.

4.1.9 For single-vessel disease, the MACE rate was statistically significantly different at 6 months' follow-up in two trials involving a total of 300 patients: 12.6% for CABG versus 25.8% for PCI with BMS (OR 0.41; 95% CI, 0.22 to 0.74). The higher MACE rate for PCI with BMS reflects the higher rate of restenosis following this procedure.

4.1.10 For multiple-vessel disease, the MACE rate was statistically significantly different at 12 months' follow-up in two trials involving a total of 2300 patients: 12.3% for CABG versus 24.5% for PCI with BMS (OR 0.43; 95% CI, 0.34 to 0.54).

4.1.11 At 36 days' follow-up, the rate of acute MIs was statistically significantly lower following PCI with BMS, but the difference between the two procedures was not statistically significant at 6 and 12 months.

4.1.12 No statistically significant differences were reported for deaths, because the trials were not powered to detect differences in these uncommon events.

BMS versus DES

4.1.13 There were 12 RCTs comparing BMS with DES. Of these, seven involved paclitaxel, four sirolimus, one everolimus and one actinomycin stents. The first two sets of trials (paclitaxel and sirolimus) are considered separately below, and the last two trials have not been considered here because they involved products that have not been granted CE marking. No RCT for the third DES with a CE mark (eluting dexamethasone) has yet been reported.

4.1.14 Paclitaxel-eluting stents Based on four trials with a paclitaxel DES (Taxus and non-CE-marked stents), involving almost 1000 patients, the MACE rate for PCIs using a paclitaxel DES was not statistically significantly lower at 36 days or at 1 year, but it was statistically significantly lower at 6 months: 7.4% for DES versus 15.4% for BMS (OR 0.48; 95% CI, 0.31 to 0.73). From a random effects model (which takes account of heterogeneity of results between trials), the 6-month data for the MACE rate were not statistically significant (OR 0.58; 95% CI, 0.24 to 1.43). Most of the MACE events refer to restenosis. However, the two trials of the Taxus DES stent (which has a CE mark), involving 583 patients, yielded a statistically significantly lower MACE rate at 6 months: 7.2% for DES versus 18.4% for BMS (OR 0.35; 95% CI, 0.21 to 0.59); and at 12 months: 9.7% for DES versus 20.5% for BMS (OR 0.41; 95% CI, 0.25 to 0.67). Paclitaxel DES have not been demonstrated to show an advantage over BMS in either mortality or prevention of MI. However, in a series of trials of the Taxus stent (the TAXUS trials), the MI rate for PCIs using a paclitaxel DES was statistically significantly lower at 6 months: 1.7% for DES versus 5.9% for BMS (OR 0.35; 95% CI, 0.12 to 0.99). The statistical significance of this result was not maintained at 12 months: 2.8% for DES versus 5.8% for BMS (OR 0.56; 95% CI, 0.23 to 1.37). Multivariate analysis of data from the TAXUS trials shows that once the effect of small-calibre arteries and long lesions has been allowed for, the difference in performance between DES and BMS for people with diabetes is not statistically significantly different from that of people without diabetes.

4.1.15 Sirolimus-eluting stents The MACE rate for PCIs using a sirolimus DES (Cypher and non-CE-marked stents) was not statistically significantly lower at 36 days, but it was lower at 9 months and at 1 year; at 9 months the rate was 7.4% for DES versus 18.9% for BMS (OR 0.34; 95% CI, 0.23 to 0.47), and at 1 year it was 7.8% for DES versus 21.8% for BMS (OR 0.30; 95% CI, 0.22 to 0.43). Most of the MACE events refer to restenosis. The trials of the Cypher sirolimus DES (which has a CE mark) showed a statistically significantly lower MACE rate compared with trials of BMS at 9 months (OR 0.32; 95% CI, 0.16 to 0.45), 12 months (OR 0.31; 95% CI, 0.22 to 0.43), and 24 months (OR 0.46; 95% CI, 0.22 to 0.97). Sirolimus-eluting stents in general have not been shown to have either a mortality or acute MI advantage over BMS in trials, and neither do the DES within the subset of Cypher stent trials. According to the joint professional submission, in larger arteries, PCIs using a sirolimus DES have shown very low rates of restenosis, approaching zero. In small-calibre arteries, PCIs using a sirolimus DES have shown lower rates of restenosis than PCIs using a BMS (for example, 7% versus 20% restenosis at 9 months in the SIRIUS trial for vessels of mean calibre 2.3 mm). In patients with diabetes and those with longer lesions, rates of restenosis following PCIs using a sirolimus DES have been higher than those of the 'average' patient, but still much lower than following PCIs using the BMS control. Subsequent post hoc subgroup analysis from one of the manufacturers from a trial involving patients who received a DES was considered by the Committee. The analysis compared the restenosis rate for people with diabetes with that for people without diabetes. It showed that the restenosis rate for those with diabetes as a whole was higher than for the non-diabetes group, but the difference was not statistically significant. In addition, the analysis did not control for artery calibre or length. According to the joint professional submission, for every 10 mm increase in the length of the stent, the difference in restenosis rate between a Cypher DES and Cypher BMS increased by between 1 and 1.6 percentage points.

4.2 Cost effectiveness

4.2.1 The most recent evidence of cost effectiveness comes from models supplied by four manufacturers and one from the Assessment Group, including an addendum. The manufacturer models show that PCIs with a BMS are cost effective compared with PCIs without stents and also compared with CABG. However, these models are relatively short-run, ranging from 6 months to 2 years. The manufacturers of the Cypher and of the Taxus stents each provided a model which examined the cost effectiveness of their own DES compared with the corresponding BMS. Each of these models showed that the DES is cost effective compared with the corresponding BMS. The Assessment Group's model showed that the Cypher and Taxus stents, as a group, are cost effective compared with BMS.

PCI with BMS versus PCI without stents

4.2.2 For patients with moderate or severe angina, PCI has been shown to be a cost-effective alternative to conventional medical treatment. Since the previous appraisal (by the Institute in 2000) of PCI with BMS versus PCI without stents, several further studies have demonstrated the cost effectiveness of stents in a number of patient populations and clinical settings, including elective stenting and stenting immediately following an acute MI.

PCI with BMS versus CABG

4.2.3 Comparative data on PCI using BMS versus CABG are available for only 3 years of follow-up. There are no data beyond 3 years and little data from years 2 to 3. The best available data are up to 2 years. Long-term models are needed to determine cost effectiveness because most patients who have stents fitted live longer than 5 years, and it is impossible to give a proper answer to the question of cost effectiveness by taking a short-term perspective.

4.2.4 The Assessment Group's model extrapolates the results to 5 years based on the currently available 3-year data. However, the extrapolation is very sensitive to the functional form chosen for the survival curve of patients who have undergone either stenting or CABG.

4.2.5 The benefits/disadvantages of PCIs using BMS compared with CABG, in terms of quality-of-life differences, derive from stents being a less invasive procedure on one hand but having higher rates of restenosis on the other. Neither of these two effects, in terms of quality-adjusted life years (QALYs), has been estimated to be very great, which means that if there was any appreciable difference in mortality between the two therapies, this factor would determine which of the therapies had the greater benefits. However, none of the meta-analyses from the trials shows any mortality benefit from PCI with either BMS or DES compared with CABG in the first 2 years. Hence, all the measurable benefits from using stents rather than CABG derive from an increase in the quality of life. Since stenting is considerably cheaper than CABG, under the 2-year models, it is therefore more cost effective, and indeed, dominates CABG.

4.2.6 The Assessment Group's model, however, estimated a survival benefit for CABG over PCI using BMS of the order of 0.05 QALYs per patient, after the model was extrapolated to 5 years. This benefit would be enough to make CABG the preferred technology (in terms of both clinical and cost effectiveness) for patients who were candidates for both stents and CABG. The clinician consultees to the appraisal process, however, vigorously challenged this, stating that previous studies had not reached this conclusion.

BMS versus DES

4.2.7 The quality-of-life component of the QALY differences between BMS and DES is small, because it relates only to the extent of the differences in restenosis rates. No differences in mortality have been demonstrated. Thus, the greatest benefits of DES over BMS will occur for categories of patients for whom the absolute differences in restenosis rates are greatest.

4.2.8 The addendum to the Assessment Report showed that, for single-vessel disease, PCI using a DES was estimated to be cost-saving compared with PCI using a BMS at 12 months for patients with diabetes and long lesions; the estimated cost per QALY for patients without diabetes and long lesions was £15,000 and, for all patients with narrow vessels, it was £16,000. These estimates were derived from patient-level data derived from the TAXUS II trial. For the total population of patients with single-vessel disease, the cost per QALY was £94,000. This estimate was derived from registry data.

4.2.9 These estimates are sensitive to five factors:

  • the percentage point reduction in the risk of revascularisation

  • the price differential between BMS and DES

  • the proportion of repeat interventions needing CABG

  • the disutility caused by recurrent symptoms

  • the average waiting time for repeat intervention.

4.2.10 There are no RCT data for two-vessel disease. The estimated incremental cost per QALY for two-vessel disease gained from PCI using a DES compared with PCI using a BMS for all non-diabetic patients is £195,000. This estimate is derived from registry data.

4.3 Consideration of the evidence

4.3.1 The Committee reviewed the evidence available on the clinical and cost effectiveness of PCIs using a BMS and DES, having considered evidence on the nature of the condition and the value placed by users on the benefits of BMS and DES from clinical experts and those who represent patients with angina. It was also mindful of the need to ensure that its advice took account of the efficient use of NHS resources.

PCI with BMS versus PCI without stents

4.3.2 The Committee considered that no new evidence had been found since the previous appraisal to change its view that where PCI is being undertaken, the use of stents is likely to be both clinically and cost effective.

BMS versus DES

4.3.3 The Committee noted that there were no head-to-head trials of the sirolimus-eluting Cypher stent and the paclitaxel-eluting Taxus stent, and the clinical experts advised that there was no evidence that would allow them to favour one of these drug-eluting agents over the other.

4.3.4 The Committee considered that, for single-vessel disease, restenosis rates were in general low using a BMS in the majority of patients requiring PCI, and that, therefore, the routine use of a DES was not justified. However, this was not the case for patients presenting with either small-calibre arteries (< 3 mm) or long lesions (> 15 mm); in these patients, the risk of restenosis using a BMS was considerably higher, and the absolute reduction in restenosis rates would justify the use of a DES.

4.3.5 The Committee considered the risk factors predicting the likelihood of higher rates of restenosis after the use of a BMS. It was persuaded that the main determinants of risk were the target vessel calibre and the complexity of the arterial lesion, in particular the length of the stenosis. It recognised that the combination of small-vessel disease and long lesions was particularly prevalent in patients with CAD who also had diabetes. Whilst, in general, patients with diabetes have higher restenosis rates than those without diabetes following PCI with a BMS, the Review Group's analysis indicated that these higher rates arise predominantly from the fact that a much higher proportion of patients with diabetes needing PCI have disease of small-calibre arteries and long lesions than is true for the general population of patients requiring PCI.

4.3.6 The Committee discussed how the RCTs comparing BMS with DES relate to current clinical practice. In particular, in the trials, the decision to reintervene following an initial PCI with stent procedure was often made on the basis of protocol-driven angiographic examinations at certain fixed times (for example at 6 months), rather than in response to the recurrence of clinical symptoms. It is likely that the trials would encourage reintervention that might not be required in clinical practice, where routine re-angiography is not usual. Thus, the Committee was aware that the difference in restenosis rates between BMS and DES identified in the trials could overestimate the extent of the difference that would actually be seen in clinical practice. The addendum to the Assessment Report attempted to correct for this potential overestimate of the benefit of DES versus BMS. The Committee decided that, whether or not the correction factor was applied, the guidance in Section 1 would not be materially affected.

4.3.7 The Committee considered PCI with a DES for more than one target vessel in a person with symptomatic coronary disease. It was aware that the evidence from the RCTs relates to the use of DES in single-vessel disease. However, the experts indicated that treatment of more than one vessel in an individual patient during PCI might be required. This is because, despite additional investigations, it is frequently difficult to determine which of several vessels identified at angiography is the most likely cause of the patient's symptoms. The Committee considered that the risk of a need for future intervention following an initial PCI is likely to be dependent on the degree of stenosis of any of the affected vessels. The appropriateness of a DES or BMS for each diseased artery in turn would therefore depend on considering the artery's characteristics in isolation from those of other diseased arteries. It was therefore persuaded that planned treatment of more than one vessel in a single patient should be based on the requirements laid out in the guidance for a single vessel.

4.3.8 The Committee discussed the use of DES with regard to coronary artery vein grafts and for more complex situations such as bifurcation lesions, but noted that there was no robust evidence in this area at present.

4.3.9 The Committee noted statements from some manufacturers that restenosis rates using a BMS of recent design were low compared with those of other BMS, and were comparable with those of the CE-marked DES. The cost-effectiveness calculations were not based on evidence comparing these stents with other BMS or with DES in head-to-head trials. The Committee considered that this evidence was not sufficient to affect its recommendations for guidance in Section 1. It would, however, wish to include further evidence on new developments in BMS design as part of the next review of the use of coronary stents.

PCI with BMS versus CABG

4.3.10 Having reviewed the Assessment Group's model and the submissions from manufacturers, together with the views of cardiologist consultees, the Committee concluded that the guidance offered in 2000 should be maintained. While it was clear that models with outcomes up to 2 years favoured stents in terms of cost effectiveness, the conclusions to be drawn from longer-term models depended critically on whether a survival advantage accrues to CABG. The Committee concluded that no convincing case had been made on this matter. Its considerations ranged over what may happen to patients requiring one or other of these procedures in different age ranges, and whether the conclusions about the most appropriate procedure would be the same for younger patients (who are more likely to need a repeat procedure) as for older ones. In none of the cases considered was there sufficient evidence of effect to be able to reach any conclusion.