Since its introduction in 1977, the long-term benefits of percutaneous coronary intervention have been limited by the phenomenon of restenosis, i.e. the recurrence of significant stenosis at the site of intervention. While in restenosis after plain balloon angioplasty roughly two-thirds of the late lumen loss is due to negative vessel wall remodelling, the late lumen loss after stent implantation is virtually exclusively caused by neointimal hyperplasia, which is a proliferation of smooth muscle cells with deposition of abundant extracellular matrix. Recently, a multitude of antiproliferative drugs have been tested, several of which have been found effective in inhibiting this neointimal hyperplasia when locally delivered from a polymer-encapsulated stent. The two compounds that have undergone the most clinical testing are paclitaxel and sirolimus, the latter being a cytostatic macrocyclic lactone with both anti-inflammatory and antiproliferative properties.
After the US Food and Drug Administration (FDA) approved a sirolimuseluting Cypher stent, produced by Cordis, in 2003, and Boston Scientific’s paclitaxel-eluting Taxus stent in 2004, interventional cardiologists quickly embraced the new technology. By the end of 2004, drug-eluting stents (DES) were used in nearly 80% of percutaneous coronary interventions in the US, and within three years several million DES had been implanted worldwide. Although the efficacy of DES in preventing restenosis has been well established over the last few years, recently concerns have been raised about a possible increased incidence of late stent thrombosis – more than six to 12 months after stent implantation – compared with the very low incidence of late stent thrombosis after implantation of bare-metal stents (BMS).
The Suspicion
At the Annual Congress of the American College of Cardiology (ACC) in March 2006, Mathias Pfisterer shocked the audience with his presentation of a new analysis of the Basel Stent Cost Effectiveness Trial (BASKET), a randomised study that compared BMS with DES. This new analysis, known as Late Clinical Events Related to Late Stent Thrombosis After Stopping Clopidogrel (BASKET-LATE), was performed to determine the risk of late stent thrombosis after discontinuation of clopidogrel, and focused on 746 patients from the BASKET study who experienced no cardiac complications within the first six months after stent implantation and were advised to discontinue clopidogrel at that time.1 Principal findings showed that in the year following clopidogrel discontinuation, the primary composite end-point of cardiac death or myocardial infarction (MI) occurred significantly more frequently in the DES group (4.9 versus 1.3%).
Non-fatal MI was also higher in the DES group (4.1 versus 1.3%), as was cardiac death (1.2 versus 0%). There was no difference in restenosis-driven target vessel revascularisation (4.5 versus 6.7%). Late stent thrombosis (the combination of angiographically documented stent thrombosis and thrombotic clinical events) occurred in 2.6% of the DES group and 1.3% of the BMS group. The median time of the late thrombotic event was 116 days after clopidogrel discontinuation, but events occurred throughout the one-year follow-up.
Whereas many trials had demonstrated a reduction in target lesion revascularisation with DES compared with BMS in recent years, none had ever demonstrated an effect on the hard end-points of death or MI. Hence, this presentation of BASKET-LATE raised serious concerns, showing a more than three-fold increase in death or MI with DES in the year after clopidogrel discontinuation.
These figures were considerably higher than expected from previous reports. Ong et al.2 reported the angiographic incidence of late stent thrombosis in an unselected population of 2006 patients who had received a DES. The rate of early stent thrombosis had earlier been reported to be 1%. More than 30 days after DES implantation – during a follow-up of at least one year (mean 1.5 years) – eight cases of late stent thrombosis were angiographically documented in seven patients, three of 1,017 with sirolimus-eluting stents (SES) and five of 989 with paclitaxel-eluting stents (PES), yielding an incidence of angiographically proved late stent thrombosis of at least 0.35%. Obviously, the main limitation of this report is that it is confined to patients who presented with acute symptoms and angiographically proved late stent thrombosis, thus precluding an accurate assessment of the true rate of late stent thrombosis. In a report on the one-year follow-up of the e-Cypher Registry – in which data were collected on 15,157 patients who underwent implantation of at least one SES – rates of acute, subacute and late stent thrombosis of 0.13, 0.56 and 0.19% of patients, respectively, were reported, representing a 12-month actuarial incidence of 0.87%.3 Cumulative rates of major adverse cardiovascular events were 1.36% at 30 days, 3.38% at six months and 5.8% at one year. Although this analysis suggests a higher degree of safety of SES, it should be interpreted with some caution because of the potential under-reporting of adverse events in large multicentre registries.
Furthermore, stent thrombosis was considered likely when the patient had a target vessel-related MI or died of a coronary event possibly caused by stent thrombosis, but only within 30 days of the index procedure. Also of note is that at one year after the index procedure, 43% of patients were still on dual antiplatelet therapy. Finally, this report highlights the fact that stent thrombosis is not a benign condition: of the 126 patients with adjudicated stent thrombosis, 53 (42.1%) died, 55 (43.7%) suffered an MI and 63 (50%) underwent a target lesion revascularisation (TLR).
The interventional team at The Washington Hospital reported 38 patients (1.27%) who presented with angiographic evidence of stent thrombosis in a total cohort of 2,974 consecutive patients treated with DES.4 The stent thrombosis occurred acutely in five patients, subacutely (<30 days) in 25 patients and late (>30 days) in eight patients. However, the authors acknowledge that they did not include in the analysis seven patients with Q-wave MI and 23 patients who died. Therefore, it is possible that the 38 thromboses within the time period specified is an underestimate. In contrast, the longer-term follow-up of randomised clinical trials with DES, available in early 2006, was rather reassuring. In a meta-analysis of eight trials in 3,817 patients with coronary artery disease who were randomised to PES or BMS, Bavry et al.5 found no increased hazard for thrombosis up to 12 months (risk ratio (RR) 1.06, 95% confidence interval (CI) 0.55–2.04; p=0.86). In another meta-analysis pooling 10 randomised studies comparing DES with BMS comprising 5,030 patients, Moreno et al.6 found a similar incidence of stent thrombosis for DES (0.58%) and BMS (0.54%) (odds ratio (OR) 1.05, 95% CI 0.51–2.15; p=1.000).
The presentation of BASKET-LATE had shocked the cardiological community, and had even found its way to the lay press. As such, it formed the impetus for numerous meta-analyses of trials comparing DES with BMS, with a focus on stent thrombosis and hard clinical end-points such as death and MI. One of the main flaws of previous reports had been the high degree of inaccuracy in assessing the incidence of stent thrombosis. Either only angiographically proved stent thrombosis was taken into account – thus undoubtedly vastly underestimating the true incidence of the phenomenon – or very different definitions had been used in various analyses. Therefore, the Academic Research Consortium (ARC) met in 2006, consisting of academic investigators, regulators and industry representatives, to develop a new set of definitions of stent thrombosis, classifying it as acute (<24 hours), subacute (1–30 days), late (31 days to 1 year) or very late (>1 year) regarding the timing of occurrence, and as possible, probable or definite regarding the degree of certainty about stent thrombosis as the cause of a clinical event. Although this new set of definitions has been criticised by some, it has become the new standard for reporting on stent thrombosis.
However, all of these new analyses, while describing the problem in more detail, were unable to resolve the uncertainty, as some indicated an increased mortality with the use of DES, others showed an increase in the incidence of stent thrombosis that was not translated into increased mortality or incidence of MI and yet other study results were simply reassuring. In a meta-analysis of 17 trials including 8,221 patients treated with PES and SES, Nordmann et al.7 calculated that SES but not PES may lead to increased non-cardiac mortality.
Using data from the Swedish Coronary Angiograpy and Angioplasty Registry, Lagerqvist et al.8 evaluated 6,033 patients treated with DES and 13,738 patients treated with BMS. They found that DES were associated with an increased rate of death compared with BMS. This trend appeared after six months, when the risk of death was 0.5% higher, and a composite of death or myocardial infarction was 0.5–1% higher per year. In a pooled analysis of data from four trials comparing SES with BMS, no significant differences were found in the rates of death, MI or stent thrombosis between the two groups at four years.9 A subgroup analysis suggested that mortality was higher among patients with diabetes receiving an SES than among those receiving a BMS. In another pooled analysis of double-blind trials, Stone et al.10 reported four-year rates of stent thrombosis of 1.2% in the SES group versus 0.6% in the BMS group (p=0.20) and 1.3% in the PES group versus 0.9% in the BMS group (p=0.30). There were no significant differences in the cumulative rates of death or MI at four years. In an analysis using data readjudicated according to the criteria set by the ARC, no increased risk of stent thrombosis with either SES or PES was found compared with BMS.11 Finally, in another meta-analysis of trials comparing SES with BMS, Kastrati et al.12 found no significant effect of the use of SES on overall long-term survival and survival free of MI compared with BMS. In addition, they confirmed a sustained reduction in the need for re-intervention after the use of SES, with a risk of stent thrombosis at least as great as that seen with BMS.
So, even after reading all these recent studies, the clinician is left puzzled. What is the exact incidence of stent thrombosis? Is late stent thrombosis after DES implantation a real problem or only a perceived problem? In as far as it exists, the general consensus seems to be that the two DES used most frequently carry a somewhat higher risk of late stent thrombosis that often presents as MI or sudden death. Therefore, the question is: what causes this increased hazard of stent thrombosis after DES implantation? While several clinical factors have been correlated with stent thrombosis – such as cessation of clopidogrel therapy, renal failure, bifurcation lesions and in-stent restenosis – the problem had actually been predicted by Renu Virmani, a pathologist, repeatedly drawing the attention of the cardiological community to delayed endothelial healing and localised hypersensitivity reactions after DES implantation.13
The Evidence
The idea that delayed endothelial healing plays an important role in the pathophysiology of stent thrombosis after DES implantation is essentially derived from three methods of observation: coronary angioscopy, examination of retrieved tissue and post mortem pathological examination. Coronary angioscopy was developed in the early 1980s in order to obtain direct visual imaging of the interior of coronary arteries. The equipment used comprises a catheter containing optical fibres, some of which are used for illumination and the vast majority for imaging. At its distal tip, this catheter has an occlusion balloon that is made of a very compliant material. After inflation of this balloon, a crystalloid solution is flushed distally to clear the field of view for the angioscope, thus allowing direct visualisation of the coronary vessel wall. Using this technique, Takano et al.14 studied the angioscopic differences – in neointimal coverage and in thrombus disappearance – between SES and BMS six months after implantation. In short, they studied 46 patients, of whom 36 had stable angina and 10 presented with an acute coronary syndrome, treated with 33 SES and 33 BMS immediately and six months after stent implantation using coronary angioscopy. They graded the degree of neointimal coverage of the stent struts (0 = absent neointima; 1 = visible struts through thin neointima; 2 = invisible struts) and evaluated the presence of thrombi. Their main finding was that the neointimal coverage six months after implantation of SES was significantly lower than after BMS implantation (edge: 1.4±0.7 versus 1.9±0.2; body: 1.0±0.5 versus 1.8±0.5; overlapping segment: 0.6±0.7 versus 1.8±0.5; p<0.0001, p<0.0001 and p=0.0069, respectively). In fact, 27 of the 153 segments (18%) of the SES were judged to have a complete absence of neointimal coverage, with exposure of the struts on angioscopy. In addition, they found a marked difference in the presence of thrombus. Thrombus was observed at baseline in seven lesions of seven SES patients, and in seven lesions of seven BMS patients. In the SES group, six of seven thrombi found at baseline remained present at the six-month follow-up, and one thrombus was newly recognised at follow-up. In the BMS group, only two of seven thrombi found at baseline remained at six-month follow-up, and there was no new thrombus formation during this time period. These findings result in an incidence of persistence of thrombus of 86% in the SES group compared with 29% in the BMS group (p=0.031). Hence, this study suggests a delayed neointimal stent coverage and slower thrombus disappearance in SES compared with BMS.
Very similar findings with angioscopy were reported by Kotani et al.,15 who studied 15 SES- and 22 BMS-treated lesions in 25 patients at three to six months after stent implantation. Thrombi were identified in eight stented segments and tended to be more common with SES (p=0.14), but were not seen on angiography. Three of the 15 SES (20%) showed complete absence of neointimal coverage, and in fact only two SES (13.3%) had complete coverage. In contrast, all 22 BMS showed complete neointimal coverage. Thrombi were more common in stents with incomplete neointimal coverage (p=0.09). A Dutch study group, using directional atherectomy, retrieved tissue from in-stent restenotic lesions in 10 SES and nine PES, six BMS and nine de novo atherosclerotic lesions, and processed it for histology and immunocytochemistry.16 They found that fibrinoid in in-stent restenotic tissue was present only in DES (p=0.004) as late as two years following DES placement, indicating a persistent incomplete healing response. The amount of fibrinoid, given as a percentage of total tissue in each atherectomy specimen, was greater in PES than in SES (17 versus 5%; p=0.026).
Finally, the most robust documentation comes from a pathologist from a registry of 40 autopsies of DES (68 stents), in which 23 DES cases of >30 days duration were compared with 25 matched autopsies of BMS implantation.17 Of 23 patients with DES >30 days old, 14 had evidence of late stent thrombosis. SES and PES showed greater delayed healing characterised by persistent fibrin deposition (fibrin score 2.3±1.1 versus 0.9±0.8; p=0.0001) and poorer endothelialisation (55.8±26.5%) compared with BMS (89.8±20.9%; p=0.0001). Moreover, DES with late thrombosis showed more delayed healing compared with patent DES. In five of 14 patients suffering late stent thrombosis, antiplatelet therapy had been withdrawn. Additional procedural and pathological risk factors for late stent thrombosis were local hypersensitivity reaction, ostial and/or bifurcation stenting, malapposition, restenosis and strut penetration into a necrotic core. In another paper, the same group showed that the most powerful histological predictor of stent thrombosis is endothelial coverage.18 They calculated that the odds ratio for thrombus in a stent with a ratio of uncovered to total stent struts per section >30% is 9.0 (95% CI 3.5–22).
The Verdict
In conclusion, there is clear evidence that the available and most studied DES have clear shortcomings regarding re-endothelialisation and thrombus protection compared with BMS. These data also call into question the optimal duration of dual antiplatelet therapy, especially when overlapping stents are implanted. Many interventionalists have, be it more or less intuitively, adopted the habit of prescribing clopidogrel for a more extended period after implantation of overlapping DES. This approach seems to be supported by the present knowledge, although it remains obscure how long this delayed neointimal coverage with resulting increased risk of stent thrombosis persists. If the recommended duration of therapy with clopidogrel were extended to several years, this would not only markedly increase the total costs of DES implantation, but also put the patients at an increased risk of bleeding or very late stent thrombosis if clopidogrel needs to be stopped for whatever reason, e.g. non-cardiac surgery.
It seems unimagineable that the path of DES will be deserted. Therefore, we will need novel stent/polymer designs with erodable polymers or polymers that do not cause inflammation or hypersensitivity – or no polymer whatsoever – with continued research for the optimal drug to be delivered, at an optimal dose and with optimised elution profiles. This research will not be easy, especially since no animal model perfectly mimics the clinical human situation of atherosclerotic coronary arteries. Animal studies using healthy pigs have shown persistence of fibrin at 28 days; however, endothelialisation was similar between DES and BMS. Also, it is well known that in pig coronary arteries endothelialisation after BMS implantation may be >80% after seven days, and that healing takes longer in humans compared with normal animal coronary arteries. Moreover, rates of endothelialisation after stent implantation seem to vary among currently used animal models. Although experience from porcine coronary implants suggests complete endothelialisation 28 days after stent deployment, data of rabbit iliac implants indicate a clear delay of endothelialisation in SES and PES.
In the meantime, a targeted use of DES seems warranted. In a non-diabetic patient with a discrete lesion in a large vessel, only a very small absolute clinical benefit of using a DES can be expected, and that minor benefit relating to restenosis may well be completely offset by an increased risk of stent thrombosis due to delayed neointimal healing. For the time being, let us reserve the currently available DES for the patients for whom the greatest benefit can be expected, i.e. those with longer lesions in smaller vessels, and patients who are stented in sites where a restenosis would bring along an especially high risk, such as the unprotected left main stem.