Drug-eluting stents (DES) have been recognised as a ‘breakthrough technology’ since 2003 due to their unique capacity to reduce clinical restenosis. The difference in restenosis rates, as assessed by angiography, was so large between DES and bare-metal stents (BMS) in the first trials – 0% versus 26% after six months in the first published trial (RAVEL) with the sirolimus-eluting stent (SES),1 and roughly 75% reduction with either SES or palitaxel-eluting stent (PES) in the two pivotal trials SIRIUS2 and TAXUS IV3 – that no doubt can persist in terms of evidence-based medicine, at least under the conditions of selected patients and short-term follow-up used in these first trials. However, even at the time of these first reports, several questions were raised concerning the limitations of these randomised trials:
- Will the beneficial effect in terms of reduction of new revascularisations persist at long-term follow-up, or will a ‘catch-up phenomenon’, as seen after brachytherapy, erase the initial beneficial effect?
- Are there any safety concerns for the use of DES as used in the conditions of these trials? In particular, is there an excessive rate of stent thrombosis, as these first trials were largely underpowered to detect a difference between DES and BMS for such rare but dramatic events?
- When DES are used in less highly selected patients rather than in randomised trials, and when off-label use of DES occurs in ‘real-life conditions’, are the results in terms of efficacy maintained, or are the results in terms of safety (mostly stent thrombosis) worse?
- For patients shown to be at high risk for implantation of a BMS – namely diabetic patients and, particularly, diabetic patients with multiple vessel disease4 – what are the results of DES either in terms of efficacy and safety, as only few such patients had been included in the pivotal trials?
In mid-2007, precise and reassuring answers can be given to the first two questions, but only partial answers can be reported for the second two questions.
First, the beneficial effect of DES in terms of reduction of new revascularisations of the treated lesion (TLR), demonstrated in the first randomised trials, persists for these patients at long-term follow-up (at least four years). There is no catch-up phenomenon and the two revascularisation curves remain parallel between one and four years. Stone’s meta-analysis5 for the four pivotal trials with SES shows a four-year TLR rate of 7.8% after SES implantation versus 23.6% for the control BMS group, and a 3.2-year TLR rate of 10.1% after PES implantation versus 20% for the control BMS group. As a direct, randomised, long-term comparison between SES and PES is still not available, it cannot be stated that SES is superior to PES (inclusion criteria for the two DES and control BMS being slightly different).
The second point regarding safety concerns about the use of DES in the conditions of the randomised trials has been recently resolved with a cautious analysis of the results of these trials, including the long-term results. However, this point has raised considerable debate for many sound reasons, fired by: cases of incomplete re-endothelialisation of the DES when autopsy was performed in some patients dying from stent thrombosis, and sometimes in patients dying from another cause;6 high rates of thrombosis after brachytherapy; reports of cases of dramatic ‘late’ DES thrombosis, often after interruption of the dual antiplatelet therapy;7 findings of higher rates of death, acute myocardial infarction and stent thrombosis seven to 18 months after implantation of a DES in the small, randomised BASKET-LATE study;8 and difficulties in defining stent thrombosis. To resolve this issue, a re-analysis of all the randomised studies has been performed, including a hierarchical analysis of the definitions of stent thrombosis – from ‘certain’ stent thrombosis, as assessed by coronary angiography following an acute coronary syndrome, to the ‘probable’ or ‘possible’ stent thrombosis of a patient presenting with myocardial infarction or sudden death at distance of the implantation (definitions of the Academic Research Consortium, ARC). Whatever the criteria considered, there was no excess of DES thrombosis at four-year follow-up.9,10
The power of these analyses is now sufficient to assert that any difference in stent thrombosis rates between DES and BMS, if present, is slight. Of note, the meta-analysis of these pivotal trials using the initial definition of stent thrombosis used in these trials showed a slight excess of ‘very late’ (>1 year) stent thrombosis with DES (5 SES versus 0 BMS; 8 PES versus 2 BMS). These differences were no longer present using the ARC definitions of stent thrombosis.
Third, what are the results of DES when they are used in conditions other than those of the early randomised trials? In the pivotal randomised trials, patients were more highly selected than in everyday practice for three main reasons: only discrete de novo lesions were included, very skilled teams were selected to participate and the inclusion of all consecutive patients with on-label lesions was not required. In everyday practice, off-label use of these stents is not rare, reaching up to 60% of the patients treated – such as patients with myocardial infarction within less than 72 hours, those with left main stenosis, bifurcation lesions, very long (>30mm) lesions, in-stent restenosis, a stenosed vein graft, etc. Two approaches – randomised studies and registries – are currently underway to define the efficacy and safety of DES in these off-label situations.
Randomised studies designed for each of these situations are ongoing and some have already been published. Most of these studies have concluded in favour of a beneficial effect of DES against BMS in terms of efficacy (reduction of the medium-term TLR rate), but are underpowered for safety analysis. Sometimes, conflicting results are seen.11,12 These studies have also sometimes led to an extension of the indications for the use of DES (a previous ‘off-label’ use becomes ‘on-label’, such as the treatment of in-BMS restenosis with the implantation of a new DES).13
Registries can be useful, provided they have followed the rules of good clinical practice: consecutive included patients, thus avoiding a selection bias; high (>98%) follow-up rate, to avoid under-reporting of events; and an independent critical event committee. Even so, registries are inadequate for a comparison of BMS and DES results.14 These registries have shown that the stent thrombosis rate is higher than in the randomised trials,15 even for on-label indications of the stent. Furthermore, off-label patients have the worst outcome. In the EVASTENT study, the off-label use of the SES was a significant univariate predictor of stent thrombosis (4.7% versus 2.1%), but not a predictor of a higher TLR rate.16 To conclude this section, if we can reasonably speculate that the efficacy of DES is maintained in most off-label situations, safety concerns must still be resolved before an extension of the indications to each of the DES is allowed.
Fourth, what are the results of DES in diabetic patients with multiple-vessel disease? The pivotal randomised studies of SES or PES versus BMS have demonstrated the same reduction rate in future revascularisations with DES in diabetic as in non-diabetic patients, leading to a greater number of revascularisations being avoided, as the restenosis rate was greater in diabetic patients after implantation of a BMS. However, in these studies the number of diabetic patients was relatively small – between 16 and 25% of the population included. Accordingly, the DIABETES study has included a limited number of diabetic patients.17 The findings have permitted the regulatory authorities to consider diabetic patients, either insulin-requiring or under oral treatment, to become an on-label use for DES, and that DES implantation could be an alternative to bypass surgery in such patients. However, these studies were underpowered to assess the safety of DES.
Two recent meta-analyses of these studies have added to the confusion: one9 demonstrated an excess of four-year mortality after implantation of an SES; the other18 did not find any difference compared with BMS. We must keep in mind the poor results of BMS versus bypass surgery in diabetic patients with multiple-vessel disease.4 The EVASTENT study has been supported by the national health authorities in France to assess the efficacy and safety of SES in diabetic patients in everyday practice. To allow some comparison with non-diabetic patients, and also to compare single-vessel and multiple-vessel disease patients, the design was as follows: for each diabetic patient included (stratified as single-vessel disease or multiple-vessel disease), a non-diabetic patient was included, leading to four separate groups of patients (matched-cohort registry), and complete revascularisation with DES was required. About 1,800 patients (half with diabetes) have been included in 50 centres. On-label use of SES was recommended, but 20% of inclusions were off-label. Careful on-site monitoring allowed a 99% follow-up of patients. An independent critical event committee has allocated all efficacy (TLR, TVR, etc.) and safety parameters (death, acute myocardial infarction, stent thrombosis assessed as ‘certain’ or ‘possible’ according to the ARC definitions). Complete interim one-year results will be published soon. The efficacy results were briefly in accord with the expected value from the previous randomised trials, although they were higher in diabetic patients (TLR 5.8% in diabetics versus 3.3% in non-diabetic patients).
Conversely, the safety results were worst in diabetic patients and, in particular, in diabetic patients with multiple-vessel disease: the one-year stent thrombosis rate was, as expected, less than 1% in non-diabetic single-vessel disease patients, but reached 2.3% in diabetic single-vessel disease patients and 4.3% in diabetic patients with multiple-vessel disease. Most of the late stent thrombosis occurred between three and nine months, when bi-therapy with aspirin plus clopidogrel was interrupted in some patients. The three-year follow-up results are ongoing.
To conclude, after four years of use DES have confirmed their efficacy in reducing clinical restenosis, even in most subgroups of high-risk patients, such as diabetic patients. They have proven to be safe when on-label use has been followed. However, in off-label use, safety concerns may persist. Furthermore, in diabetic patients with multiple-vessel disease a randomised comparison with bypass grafting is awaited (FREEDOM, STIRTAX) to choose the best therapeutic option for our patients.