Circulation. 2007;115:1440-1455
Published online before print March 7, 2007,
doi: 10.1161/CIRCULATIONAHA.106.666800
(Circulation. 2007;115:1440-1455.)
© 2007 American Heart Association, Inc.
Controversies in Cardiovascular Medicine |
Stent Thrombosis Late After Implantation of First-Generation Drug-Eluting Stents
A Cause for Concern
Edoardo Camenzind, MD;
P. Gabriel Steg, MD;
William Wijns, MD
From the University of Geneva, Geneva, Switzerland (E.C.); Groupe Hospitalier Bichat-Claude-Bernard, Paris, France (P.G.S.); and Cardiovascular Center, OLV Hospital, Aalst, Belgium (W.W.).
Correspondence to Edoardo Camenzind, MD, University of Geneva, 1 rue Michel-Servet, 1211 Geneva, Switzerland. E-mail edoardo{at}camenzind-cardio.net
 |
Introduction
|
|---|
Percutaneous coronary intervention has become the most frequently
used method of myocardial revascularization.
1,2 The advent of
coronary stenting led to a significant decrease in the complications
seen after balloon angioplasty, resulting in improved patient
outcome.
3,4 Yet, stented angioplasty has been plagued from the
onset by early stent thrombosis (<30 days after index procedure)
and late in-stent restenosis (ISR). Initially, stent thrombosis
rates as high as 24% raised serious doubts as to the viability
of the therapy.
5 With the combined prescription of thienopyridines
and aspirin for 4 to 8 weeks,
6,7 together with proper stent
deployment techniques,
8 early stent thrombosis rates decreased
to what was felt to be an unavoidable and acceptable 1% to 1.5%.
At the same time, efforts to reduce the 30% late ISR rates through
systemic pharmacological approaches remained unsuccessful until
local radiation, a strong antiproliferative therapy, was applied
to prevent or treat ISR.
913 Vascular brachytherapy was
the first illustration that delayed healing might portend an
increased risk of thrombosis together with the expected reduction
in restenosis. Indeed, stent thrombosis rates increased again
up to 5.3%, and the time window of event occurrence was extended
beyond 1 year so that the initial clinical benefit would eventually
erode as time went by.
13,14 Today, first-generation drug-eluting
stents (gen1-DES: Cypher, Cordis, Johnson & Johnson, Miami
Lakes, Fla [sirolimus-eluting stent, SES] and Taxus, Boston
Scientific Corp, Natick, Mass [paclitaxel-eluting stent]) releasing
an antiproliferative compound (sirolimus or paclitaxel, respectively)
via a nonbioerodable polymer have been proved to reduce the
incidence of ISR by up to 75%.
1539 Since the publication
in 2002 of the first randomized trial
16 comparing DES and bare
metal stents (BMS) in highly selected patients and lesions,
the use of DES in clinical practice has expanded to the majority
of coronary lesion subsets (eg, de novo complex lesions, long
lesions, small vessels), to high-risk patients (multivessel
angioplasty, patients with diabetes mellitus),
2527,3339 and more recently, to primary percutaneous coronary intervention
for ST-segment elevation acute myocardial infarction (MI).
40,41 From the literature, Cypher and Taxus appear to yield similar
rates of repeat revascularization, although some studies suggest
that the luminal preservation achieved by the Cypher stent and
measured by coronary angiography may be slightly superior to
that of the Taxus stent.
4245 At the same time, late stent
thrombosis (LST; >30 days after index procedure), although
rare, is again emerging as a cause for concern with both types
of gen1-DES.
4650
Response by Serruys and Daemen p 1455
Understanding the pathophysiology of late thrombosis of gen1-DES seems essential for assessing the clinical relevance of these unpredictable and potentially lethal events.51,52 On the basis of the evidence that is available from preclinical, autopsy, and clinical studies, we propose that the pathophysiological mechanisms known as Virchows triad may be responsible for LST with gen1-DES.53 Furthermore, we comment on the clinical evidence derived from industry-sponsored and investigator-driven trials that suggests a small but relevant incremental risk of LST with gen1-DES compared with BMS.
 |
Delayed Vascular Healing as a Consequence of Gen1-DES Implantation
|
|---|
Autopsy studies have shown that after BMS deployment an inflammatory
reaction takes places in the vessel wall that involves macrophages
and T lymphocytes with few B lymphocytes and giant cells.
5456 After implantation of gen1-DES, a more pronounced inflammatory
response has been described that may occasionally be associated
with a local hypersensitivity reaction and eosinophilic infiltration.
48,50 The synthetic nonbioerodable polymer containing the drug may
be an important trigger of local coronary inflammation,
48,50,5759 even though the metal struts
60 or the drug itself may participate
in this phenomenon.
61 Coronary inflammation is responsible for
both delayed reendothelialization of stent and vessel wall
56,62 and destruction of medial vessel wall layers, causing positive
regional remodeling, all of which eventually result in a delayed
vascular healing response.
48,50
Clinically, local inflammatory response and delayed or incomplete arterial healing manifest on angiography by aneurysm formation but also can be suspected in the presence of late acquired stent malapposition (LASMA), as was observed with imaging modalities such as intravascular ultrasound (IVUS).6366 Incomplete strut apposition by IVUS (which includes both persistent malapposition and LASMA) has been observed in 21% of the patients assigned to SES in the Randomized Study With the Sirolimus-Eluting Velocity Balloon-Expandable Stent in the Treatment of Patients With De Novo Native Coronary Artery Lesions (RAVEL) trial.67 The incidence of LASMA was close to 10% in both the Sirolimus-Eluting Stent in De Novo Native Coronary Lesions (SIRIUS)68 (8.7%) and TAXUS II69 (8.7%) trials. According to these trials,
1 of 10 to 20 lesions with LASMA will develop an angiographically visible aneurysm.64,69 This incidence has been confirmed in the TAXUS V trial in which 1.4% of the patients developed late acquired aneurysms in the paclitaxel-eluting stent arm.38
 |
Pathophysiology of LST
|
|---|
These local changes in vascular biology, in combination with
systemic alterations of coagulation pathways, are reminiscent
of the pathophysiological mechanisms described in 1856 by Rudolf
Virchow and could be responsible for LST with gen1-DES
53: (1)
an abnormal vessel wall lining (eg, incomplete endothelialization),
(2) an abnormal blood-flow pattern (eg, slow flow), and (3)
altered blood constituents (eg, increased blood thrombogenicity).
Any of these elements alone or in combination favors intravascular
thrombus formation.
53,70 This generic mechanism of disease can
be applied specifically to the coronary arteries after implantation
of gen1-DES
4549 and can be delineated as follows.
Abnormal Vessel Wall Lining
In the first month after the implantation of a BMS, a new endothelial layer covers the stent struts, reestablishing a "normal" coronary vessel wall lining, thus reducing the risk of LST secondary to strut or vessel surface thrombogenicity.5456 Gen1-DES inhibit or may even abolish this physiological vessel wall healing, leaving the struts in direct contact with flowing blood and blood elements.48,50,71 Complete or partial lack of reendothelialization of stent struts and vessel wall generates a long-lasting,62 if not permanent, unhealed vessel wall surface favoring platelet adhesion and aggregation, which may eventually cause thrombus formation.48,50 With BMS, incomplete reendothelialization also has been observed,55 but unlike the case with DES, it was not seen in series of comparative case reports.50
Abnormal Blood-Flow Pattern
Within 8 months, the inflammatory vessel wall response to gen1-DES implantation may induce a positive regional vascular wall remodeling, as shown by histology48,50 or IVUS,63 may be responsible for LASMA,6769 or even, in extreme cases, may cause angiographically visible aneurysm.36,6466 These structural changes (vessel widening) may induce, according to the law of continuity,72 slow flow velocities and low shear stress particularly in the vicinity of the stent struts73 and possibly on the abluminal side between the stent struts and the coronary wall. Slow flow velocity is known to promote thrombogenesis,53,72 as demonstrated by the increased propensity for acute MI to occur in aneurysmal-ectasic vessels74 and suggested by thrombotic occlusion of gen1-DES at the site of aneurysmal dilatation.48,50,72 Therefore, regional positive vascular remodeling within the stent increases the surface-related prothrombogenicity of both vessel wall and stent struts.75
Abnormal Blood Constituents
Increased blood thrombogenicity plays a crucial role in favoring LST. The prothrombotic effect of the reduction or discontinuation of aspirin,76 ADP receptor inhibitors46,47 (eg, clopidogrel), or both49 may be further potentiated by an ongoing systemic inflammatory reaction77,78 (eg, fever, postoperative course, malignancy) or by dehydration.79
 |
The Link Between the Antirestenosis Effect of Gen1-DES and LST
|
|---|
The mechanism of the antirestenosis properties of gen1-DES is
best appreciated by detailed measurements of arterial dimensions
with either coronary angiography or IVUS. Late loss (LL) is
a quantitative coronary angiography parameter used to measure
indirectly the inhibitory potential of DES on neointimal formation.
It is defined as the minimal luminal diameter after the procedure
minus the minimal luminal diameter at follow-up and is frequently
plotted on frequency-distribution curves. From randomized comparisons
with BMS, it appears that the frequency-distribution curve of
LL with gen1-DES is shifted to the left (
Figure 1), reflecting
their strong inhibitory effect on intimal hyperplasia.
80 The
left end of the curvebelow the 0-mm point of LL (negative
LL)represents the patient subset with increased minimal
luminal diameter at follow-up that is thus likely to have experienced
positive remodeling.
80

View larger version (24K):
[in this window]
[in a new window]
|
Figure 1. Frequency distribution of LL values after SES (Cypher) vs BMS. The SES distribution curve is shifted to the left, indicating the inhibitory effect on intimal hyperplasia. Modified from Lemos et al,80 with permission of the publisher. Copyright © 2004, the American Heart Association.
|
|
Widening of the coronary lumen over time may reduce both intrastent flow velocity72 and wall shear stress.73 Segmental slow flow may be caused by a local intravasal abnormality (eg, LASMA, aneurysm, or bifurcational stenting)51,6466 and by a global coronary perfusion abnormality (eg, diastolic coronary perfusion determined by variables such as tachycardia, increased telediastolic pressure, microangiopathy, distal embolization) and thereby give rise to prolonged interaction between vessel wall and blood constituents.81 Finally, abnormal blood constituents influenced by systemic factors such as dehydration,79 inflammation (eg, increased fibrinogen),77 and hemostatic balance (eg, anticoagulation/antiplatelet treatment)76 will increase intrastent thrombogenicity. Because these determinants of the risk of LST vary over time and from patient to patient, it is not surprising that LST remains largely unpredictable.
The link between the antirestenosis effect or healing response measured by LL and clinical events (LST and ISR) after gen1-DES implantation can be described as a J-shaped curve relationship: Both negative and high LL are linked to an excess in clinical events (Figure 2).82 LST is rare (flatter portion of the J curve) and more likely to occur in the patient population with negative LL (larger lumen by angiography as a result of delayed healing or nonhealing). Repeat revascularization is frequent (steeper portion of the J curve) and occurs more often as the LL values increase (smaller angiographic lumen as a result of exuberant healing and neointimal growth). Any increase in the population with negative LL is expected to increase the absolute number of patients potentially subject to LST (Figure 3). Therefore, the paradigm that predicates the use of quantitative coronary angiography (or IVUS) as a surrogate end point for both efficacy and safety outcome may no longer be valid in the setting of technologies that interfere with vascular healing.83,84

View larger version (40K):
[in this window]
[in a new window]
|
Figure 2. J-curve relationship between LL and clinical events. A negative LL (left part of the curve) and an increasing positive LL (right part of the curve) are both linked to more clinical events. Events such as late thrombosis are more likely to occur in the population with a negative LL (left arm of the curve), and events such as restenosis are more likely to occur with progressively increasing LL (right arm of the curve). Modified from Camenzind,82 with permission of the publisher. Copyright © 2006, the Massachusetts Medical Society.
|
|

View larger version (25K):
[in this window]
[in a new window]
|
Figure 3. Derived from the frequency-distribution curve of the LL values, the cases below the 0-mm mark of LL represent the group having a positive regional vascular remodeling at follow-up. Therefore, the area under the frequency-distribution curve of LL below the 0-mm mark represents a collection of patients at risk for LST for having at least 2 criteria of the Vichows triad (an abnormal vessel wall lining and an abnormal blood-flow pattern). The population at risk for LST is 5 time larger in the DES group vs the BMS group. Modified from Lemos et al,80 with permission of the publisher. Copyright © 2004, the American Heart Association.
|
|
 |
Identifying the Population at Risk for LST
|
|---|
To identify patients at higher risk for LST with gen1-DES and
to assess the magnitude of the problem, one should focus on
the patient population with delayed or nonhealing response.
The obvious way to identify these patients is to evaluate positive
regional coronary remodeling, which is the morphological expression
of a local inflammatory response and delayed healing. According
to quantitative coronary angiography, the proportion of cases
below the 0-mm LL point, represented by the area under the negative
part of the frequency-distribution curve of LL values, is close
to 50% (
Figure 3).
80 According to IVUS and the incidence of
LASMA, the patient population at risk for LST after gen1-DES
can be estimated on average at 10% (RAVEL, 13 of 91; SIRIUS,
7 of 80; TAXUS II, 20 of 229; Diabetes and Sirolimus-Eluting
Stent [DIABETES], 11 of 75; total, 51 of 475 or 11%).
64,68,69,85
The fact that the shift to the left of the frequency-distribution curve of LL after SES implantation seems more pronounced in the subgroup at highest risk for ISR is noteworthy. The mean in-segment LL is 0.02 mm in patients with insulin-dependent diabetes mellitus versus 0.09 mm in patients with noninsulin-dependent diabetes, as opposed to 0.56 versus 0.42 mm after BMS implantation. Therefore, this subset of patients at higher risk of ISR also is more likely to develop incomplete healing and positive remodeling.86 The fact that LASMA secondary to gen1-DES is more frequent in patients with an increased risk of restenosis seems counterintuitive because some degree of resistance to the antiproliferative action of DES would be expected in patient/lesion subsets with higher propensity for ISR.85,86 However, this observation has recently been confirmed by serial IVUS analyses in diabetic patients in whom LASMA was observed in 14.7% of the cases after SES compared with 0% after BMS implantation (P<0.001).85
 |
Recognizing LST
|
|---|
Assessing the true incidence of rare safety-related events such
as LST is not trivial. One can focus on the potential clinical
consequences of the event (eg, death and nonfatal MI) or on
the event itself using angiographic or pathological evidence
of stent thrombosis. The clinical end points are less specific
for LST but more sensitive than angiographic assessment and
will thus achieve a more "inclusive" recognition of LST. Reporting
only the LST event itself may appear more specific but represents
a less sensitive and a definition-dependent approach that pictures
a more "restrictive" recognition, with the potential for underestimation.
When dealing with safety issues, we can make a case for using
broad and inclusive assessment methods. These methodological
issues complicate the assessment of LST. Rates of death and
MI may be affected by potential confounding factors during the
process of "adjudication to prespecified event definitions"
or by "partial reporting," leading to an underestimation of
clinically serious events. Predefined adjudication of events
such as cardiac and noncardiac death and Q-wave and nonQ-wave
MI is almost systematically used in gen1-DES trial programs.
1539 Event adjudication is particularly problematic for death (ie,
cardiac or noncardiac death) because sudden-onset cardiac events
in patients with cancer or intercurrent infection may truly
be related to stent thrombosis and yet adjudicated as noncardiac.
Indeed, prothrombotic changes (eg, inflammatory status, dehydration)
in patients with advanced noncardiac disease (eg, malignancies)
may trigger cardiac death. Of interest is the fact that in the
Cypher clinical program
1527 all-cause death was reported,
whereas in the Taxus program,
2839 mortality events were
restricted to adjudicated cardiac death (except for TAXUS V,
VI [oral presentation at 2 years], and the pilot trial TAXUS
I
28,29,3639) (
Tables 1 through 4


). Overall in the Cypher
and Taxus randomized trials, mortality rates were higher in
the gen1-DES arm compared with the BMS arm (4.67% versus 3.33%
and 2.15% versus 2.01%;
Tables 1 through 4


). The lower absolute
mortality rate in the Taxus versus the Cypher trial programs
(2.15% versus 4.67%) may simply be related to the adjudication
process. Thus, it would seem advisable to report both all-cause
death and adjudicated cardiac death. Of note, in recently reported
direct comparisons via randomized trials, serious events (death
and MI) were more common in the Taxus than the Cypher group.
4245
View this table:
[in this window]
[in a new window]
|
TABLE 2. Serious Adverse Events in Cypher Versus BMS Trial Program According to Intention-to-Treat Analysis, Pooled Trials Up to Latest Follow-Up
|
|
View this table:
[in this window]
[in a new window]
|
TABLE 4. Serious Adverse Events in the Taxus Versus BMS Trial Program According to Intention-to-Treat Analysis, Pooled Trials Up to Latest Follow-Up
|
|
When stent thrombosis events are reported with the more specific approach, the use of predetermined definitions has proved to be problematic. Early stent thrombosis is defined as an ischemic event up to 30 days after the index procedure, an event which can include unexplained death, Q-wave MI, or (sub)abrupt closure requiring revascularization. In contrast, LST is defined as ischemic event >30 days after the index procedure that includes solely MI attributable to the target vessel with angiographic documentation of thrombus or total occlusion at the target site and freedom from an interim revascularization of the target vessel (so-called late angiographic stent thrombosis [LAST]). Thus, unlike the case with early stent thrombosis, the definition of LAST does not take into account all potential clinical presentations of stent thrombosis by excluding death and even ECG-documented Q-wave MI in the absence of angiography. Requiring angiographic documentation of intracoronary thrombosis assumes that direct percutaneous coronary intervention will be used universally to treat acute MI patients, which is far from being the case. Instead, with reperfusion after successful thrombolysis and antithrombotic therapy, the diagnosis of LAST may be dismissed, even when delayed angiography is available. Using this restrictive definition likely underestimates the true rate of LST and smoothes out differences between devices over short observation periods.87 Recognizing these issues, a proposal has been put forward to categorize stent thrombosis (Dublin or ARC definitions) according to timing (acute, subacute, late, and very late) and level of documentation (definite, probable, and possible), which illustrates that LST is highly dependent on definition and adjudication (D.E. Cutlip, personal communication, September, 2006). This "unifying" definition improves the comparability across trials but will not necessarily guarantee a more accurate assessment of the incidence of LST with gen1-DES. An example is that including stent thromboses that follow interim revascularization of the target vessel has a major impact on LST rates (excluded according to study protocol definitions). As a result, thrombosis events that are consecutive to the treatment of ISR in the control arm (eg, with in-stent implantation of gen1-DES or brachytherapy) may counterbalance the spontaneously occurring LST in the gen1-DES arm. Finally, because of the small sample size of many of these trials, even a small number of patients excluded from analysis (eg, because of lost to follow-up, revoked patient consent, or follow-up out of the predefined time-window) may have a large impact on comparative outcomes. These methodological issues illustrate the complexities entailed by the analysis, presentation, and comparison of such trials and databases.
 |
Assessing the Safety of Gen1-DES: From LST to All-Cause Mortality
|
|---|
Although we recognize the limitations in our ability to detect
rare side effects and the difficulties in comparing the different
studies, we have attempted to evaluate the relative safety profile
of gen1-DES and BMS from the analysis of the following events:
all-cause mortality, with and without excluded patients, per
protocol and per intention-to-treat analysis (
Tables 5 and 6
)
and LST as the target safety event, defined in a pathogenetically
pertinent and clinically relevant manner (
Tables 1 through 4


).
All official data sources have been consulted and compared to
verify data consistency.
1539 The following data sources
were retained: (1) published peer-reviewed articles, (2) presentations
at major meetings (American College of Cardiology, American
Heart Association, European Society of Cardiology, EuroParis
Course on Revascularization, and Transcatheter Cardiovascular
Therapeutics), (3) printed information distributed by the industry,
and (4) latest updated data on file from the industry. It is
important to realize that practice guidelines prepared by scientific
societies are using solely peer-reviewed articles as data sources.
88,89 The additional data sources were included to increase the analysis
sample and to capture the longest available follow-up period.
Indeed, it is unusual when the peer-reviewed publication appears
within 1 year after the official first presentation,
16,21,25,27,28,31,36,38 and follow-up presentations are frequently not published.
22,29,34
View this table:
[in this window]
[in a new window]
|
TABLE 5. Calculated Death and All-Cause Mortality in the Cypher Versus BMS Trial Programs Up to Latest Available Follow-Up in Intention-to-Treat Patients According to Data on File at Cordis
|
|
View this table:
[in this window]
[in a new window]
|
TABLE 6. Calculated Death and All-Cause Mortality in the Taxus Versus BMS Trial Programs Up to Latest Available Follow-Up in Per-Protocol Patients According to Data on File at Boston Scientific
|
|
The most appropriate manner to assess safety is under debate. A more rigorous approach to assess safety to maximize the detection of rare adverse events, at least in device trials in which compliance is not really a confounding factor, seems to be the use of as-treated analysis (better option) or per-protocol analysis. On the contrary, a more rigorous approach to assess efficacy is an intention-to-treat analysis. An inclusive approach to mortality is to use calculated mortality defined as the sum of all-cause mortality plus the excluded patients, assuming that their death is a worst-case hypothesis (Tables 5 and 6
). For the Cypher program, a per-protocol analysis was not performed, and the number of patients excluded from the intention-to-treat analysis were kindly provided on request (Table 5; D. Donohoe, data on file, Cordis). Focusing on the intention-to-treat patient population, calculated mortality presents as a
2-times-lower risk difference (0.72% versus 1.34%) compared with all-cause mortality when evaluating the Cypher arm and the BMS arm (Table 5). The same comparison in the Taxus program using the more rigorous per-protocol population showed a
2-times-higher risk difference (0.85% versus 0.31%) in the calculated compared with the all-cause mortality (Table 6; J. Koeglin, data on file, Boston Scientific). These results may be interpreted in the following manner: The intention-to-treat population of the Cypher program may not only level out but also invert the gradient between calculated and all-cause mortality. Conversely, the higher gradient in calculated mortality versus all-cause mortality in the Taxus program may reflect both the use of the per-protocol population, which amplifies the difference between the active treatment arm and the control arm, and a larger proportion of excluded patients in the Taxus arm compared with the control arm. In the Taxus program, excluded patients can be calculated from the decreasing patient numbers, as indicated in Table 3. However, the specific cause of exclusion, whether death, loss to follow-up, revoked patient consent, or follow-up outside the predefined time window, was not specified. Of interest is the fact that both excluded patients and as-treated or per-protocol analysis are rarely presented in publications or presentations, which makes it difficult to readily assess safety. Most important, the proportion of the excluded patient population is of a magnitude similar to the all-cause mortality rate, which emphasizes the potential impact that the collection of excluded patients could have on the global safety assessment.
Tables 7 and 8
show all-cause mortality rates according to different data sources for the Cypher and Taxus trial programs. Of major concern is the variability of the reported all-cause mortality according to the different sources that we have scrutinized.1539,90,91 As shown in Tables 7 and 8
, it is of particular concern that all-cause mortality rates appear to differ between published peer-reviewed literature and data on file at the companies, especially when the latter are lower than the former.29,91 The biggest limitation in comparing these data derives from the fact that all-cause mortality is not systematically reported (eg, adjudicated cardiac mortality is reported instead). Further subtle differences in data reporting may be encountered (eg, results are reported as percentages but without mention of the excluded patients, precluding calculation of the absolute number of events).
To assess LST in a clinically relevant manner, definitions should be respectful of the mechanism of disease leading to the event. Most appropriately, LST should be defined in the same way as acute thrombosis (<30 days). For the time period >30 days after the index procedure, death and MI are the most frequently encountered clinical presentations of LST.51,75,92 However, nonST-segment elevation MI generally is secondary to revascularization procedures, mainly as a result of the treatment of ISR (see case narratives for the Cypher and Taxus program when available). Accordingly, the incidence of the combined serious events (all-cause death and Q-wave MI) up to the latest available follow-up time point was 60% higher in the Cypher than in the BMS group (6.26% versus 3.91%; Table 2), suggesting a prothrombotic effect of gen1-DES. The risk was consistently higher throughout all available clinical follow-up time points (6 to 9 months, 1 year, 2 years, 3 years) with >15% of total death and >95% of Q-wave MI in the gen1-DES group, a finding that suggests that the increased risk persists up to 3 years (Tables 9 and 10
). Accordingly, the incidence of stent thrombosis can be estimated to range from
2% per year (according to the follow-up of the combined data of RAVEL, SIRIUS, and E-SIRIUS up to 3 years) to 3.4% per year (according to the follow-up data up to 4 years of RAVEL solely) compared with 1.34% and 1.07% per year after randomization to BMS in the Cypher and Taxus programs, respectively (Tables 9 through 11
). A similarly consistent gradient of risk, albeit of smaller amplitude for the reasons discussed above, was observed in the Taxus program (Table 11).
View this table:
[in this window]
[in a new window]
|
TABLE 9. Global Serious Adverse Events in the Cypher Versus BMS Trial Program (RAVEL, SIRIUS, E-SIRIUS, C-SIRIUS) Stratified by Time Point of Follow-Up (Data According to Maximal Difference SC Reported in the Literature or at Official Meetings*)
|
|
View this table:
[in this window]
[in a new window]
|
TABLE 10. Global Serious Adverse Events in the Cypher Versus BMS Trial Program (RAVEL, SIRIUS, E-SIRIUS, C-SIRIUS) Stratified by Time Point of Follow-Up (Data According to Minimal Difference SC Reported in the Literature or at Official Meetings*)
|
|
View this table:
[in this window]
[in a new window]
|
TABLE 11. Global Serious Adverse Events in the Taxus Versus BMS Trials (TAXUS I, II, IV, V, and VI) Stratified by Time Point of Follow-Up
|
|
Worth noting, the investigator-driven follow-up analysis of the Basel Stent Kosten Effektivitats Trial (BASKET),93 (BASKET-late94), showed that cardiac death or nonfatal MI occurred in 4.9% of patients with gen1-DES versus 1.3% of patients with BMS, from months 7 to 18 after stenting, after discontinuation of clopidogrel. Late thrombosis-related events (ie, angiography- or autopsy-confirmed stent thrombosis or target vessel related death or MI) were twice as frequent after gen1-DES than after bare metal stents (2.6% versus 1.3%).
From independent registries75,95 of unrestricted use of gen1-DES, the rate of LAST (the restrictive definition of LST) was 0.17% to 0.35% per year (incidence calculated as LAST over the mean follow-up), which also reflects a 2- to 3-times-higher incidence than with BMS (0.1% per year). Finally, an analysis of pooled data from 2 high-volume institutions (the ThoraxCentre, Rotterdam, the Netherlands, and University Hospital, Bern, Switzerland), representing >8000 patients treated with gen1-DES, has suggested that the risk of LAST may accrue at a steady rate of
0.6% per year, with no indication that event curves are reaching a plateau at up to 3 years.96 These findings from real-life practice that include up to 60% off-label use of the devices are in agreement with the present analysis, which is restricted to the data reported from randomized clinical trials1539 (Tables 9 through 11
).
 |
Clinical Implications
|
|---|
Administration of dual antiplatelet therapy, usually maintained
for 3 to 6 months after gen1-DES deployment, generally can counterbalance
the increased thrombogenicity favored by delayed vessel healing
and local inflammation.
16,21,25,27,28,31,36,38 The importance
of dual antiplatelet therapy is underscored by the finding that
its interruption appears to be the single most potent correlate
of LST,
46,47,49,51,92 with an impressive hazard ratio of 163
(95% CI, 26 to 998;
P<0.001).
49 This phenomenon had already
been observed when stenting was associated with brachytherapy.
9799 In addition, in the BASKET-late analysis, thrombosis-related
events occurred between 15 and 362 days after discontinuation
of clopidogrel,
94 indicating that intercurrent events that interfere
with blood thrombogenicity may still trigger LST. Thus, the
excess risk of LST may not be confined to the days after clopidogrel
discontinuation.
7779 The 2005 European Society of Cardiology
and the 2005 American College of Cardiology/American Heart Association/Society
for Cardiovascular Angiography and Inverventions recommendations
88,89 suggest maintaining dual antiplatelet therapy for 12 months,
at least "in patients who are not at high risk of bleeding."
Given the relation between discontinuation of dual antiplatelet
therapy and LST, clinicians may be tempted to continue such
therapy even longer. However, the extent to which protracted
dual antiplatelet therapy with aspirin and clopidogrel confers
protection against LST is unknown, although such therapy was
shown to portend genuine bleeding risks, as demonstrated in
the Management of Atherothrombosis With Clopidogrel in High-Risk
Patients With TIA or Stroke (MATCH) and Clopidogrel for High
Atherothrombotic Risk and Ischemic Stabilization, Management
and Avoidance (CHARISMA) trials,
100,101 and carries significant
costs that may render very long-term prescription unpractical
or inaccessible to many patients. Even in the short-term, the
Prospective Registry Evaluating Myocardial Infarction: Events
and Recovery (PREMIER) study has shown that up to 14% of patients
will not be compliant with the prescribed dual antiplatelet
therapy for a variety of reasons.
92 In addition, when noncardiac
surgical, dental, or biopsy proceduresall intercurrent
events frequently occurring in the adult patient population
treated with percutaneous coronary intervention and even more
so in elderly with multiple comorbiditiesare contemplated,
difficult dilemmas arise as to the most appropriate clinical
management strategy.
Instead of prolonging the duration of dual antiplatelet therapy in all patients, it would be more helpful to be able to identify subgroups of patients at higher risk of LST in relation to an increased incidence of delayed or nonhealing vascular response after implantation of gen1-DES. Invasive surrogate markers such as negative LL by quantitative coronary angiography or LASMA by IVUS are helpful for understanding mechanisms but cannot be used to screen all patients.
At the same time, it is acknowledged that continuing interaction exists between gen1-DES and the vessel wall, resulting in dynamic changes in size of plaque and vessel dimensions, for at least 2 years after implantation.102 Therefore, quantitative coronary angiography or IVUS obtained at a single time point (eg, 6 to 9 months) may not be predictive for the long term in the individual lesion/patient. Among the potential clinical predictors, the available data suggest that patients with diabetes mellitus appear to be at higher risk of LST perhaps for longer periods of time. Another group of patients at higher risk are those with overlapping gen1-DES.103 Multivariable analysis of pooled databases that include individual patient data would be most helpful in determining which patient groups may benefit most from treatment with gen1-DES and those in whom the risk of LST, and its potentially catastrophic consequences, is unacceptably high.
Given the current concerns about the long-term safety of gen1-DES and the uncertainties about the potential duration of the incremental risk, the indiscriminate daily use of gen1-DES implantation in all patients undergoing percutaneous coronary intervention no longer seems advisable. In patients at low risk for ISR or in patients at high risk for LST, the clinical benefit in terms of ISR reduction may be offset by an increased risk of LST, which carries significant morbidity and mortality.51 Therefore, when gen1-DES are preferred over BMS, physicians and patients may be trading relatively benign events such as restenosis and repeat revascularization for a rare but potentially fatal event caused by stent thrombosis. These undeniable safety concerns should not be viewed as detracting from the benefit of stented angioplasty with DES but rather as a reminder of the need to accumulate large data sets with long-term follow-up to identify subgroups with balanced safety and efficacy outcome, as well as additional information on the optimal duration of antiplatelet therapy.
These concerns should also be put into context. Given the relatively small total population with late follow-up available from randomized trials and the low rate of the events, there remains considerable uncertainty about the true rates of LAST with DES and BMS in real-life practice. It remains possible that the (small) risk of LAST with DES may be offset by the clinical benefits derived from reduced need for repeat revascularizations.104
 |
Research and Regulatory Implications
|
|---|
The observed safety issues are inherent to the mechanisms of
action of gen1-DES, and in the future, every new DES will have
to be analyzed not only by angiographic metrics but also in
a clinically pertinent manner. The vascular healing response
to any new device or treatment modality is the pivotal mechanism
that will determine its long-term safety.
82 Therefore, more
focus should be placed on this issue during preclinical testing,
which thus far was concentrating primarily on efficacy metrics.
82 Specific experimental animal settings should be developed to
better understand and test the antihealing or prohealing properties
of newer-generation DES.
105 The translation from preclinical
studies to clinical testing and evaluation should initially
be performed in small pilot trials to determine the healing
response in human atherosclerotic vessels and therefore provide
some indication as to the propensity for LST and hence the desirable
duration of dual antiplatelet therapy. To visualize and quantify
vessel healing response, sophisticated imaging methods and functional
testing may prove useful when used alone or in combination (eg,
angiography, IVUS, optical coherence tomographic imaging, angioscopy,
in vivo studies of endothelial function). After assessment and
confirmation of the vascular healing response, adequately sized
clinical trials should be performed to determine clinical benefit
and to confirm safety. When a new device is approved for clinical
use, long-term monitoring of safety outcomes should be mandated
by regulatory agencies, monitoring that should include device
traceability. As is obvious from recent history, it is particularly
important that adjudication of events, data monitoring, management,
and analysis, as well as long-term follow-up, be performed independently
from sponsoring device companies. It is important to monitor
clinical safety events (such as all-cause mortality) and not
solely definition-dependent and adjudication-dependent events,
the incidence of which may be reduced by the use of stringent
definitions. Furthermore, it is worrisome to see new DES no
longer tested for superiority against BMS but rather for noninferiority
with approved devices in relatively small trials and with angiographic
end points. As a result, potential small increases in death
and nonfatal MI rates may become part of the "background" event
rate in both study arms and no longer be noticeable.
 |
Summary
|
|---|
Several lines of evidence converge in suggesting that the use
of gen1-DES is associated with a small but incremental risk
of potential LST. Three years ago, occasional case reports alerted
us to the potential for very LST (beyond 1 year) after gen1-DES
placement.
46,47,49 Although rare, these events bear severe clinical
consequences with case-fatality rates as high as 45%.
51 LST
appears to be temporally related in part to discontinuation
of antiplatelet therapy.
48,52 Because most LST events occur
outside hospitals, angiographic or autopsy documentation of
definite stent thrombosis is often missing, with the potential
for underestimation of the problem. Because of their clinical
presentation, death and MI have been suggested as the ultimate
clinical outcome and benchmark for LST. A critical review and
pooled analysis of the available randomized clinical trial data
comparing Cypher or Taxus with BMS show a consistent trend toward
more frequent death and Q-wave MI at late follow-up with gen1-DES.
These findings are consistent with a recently published meta-analysis
106 and with the results of the BASKET-late trial
94 showing a 3-
to 4-fold increase in late deaths or MI. The analysis of pooled
registries from 2 high-volume institutions has suggested that
the risk of LAST may accrue steadily without plateauing for
up to 3 years.
96 Taken together, these data provide cause for
serious concern about the long-term safety of gen1-DES, particularly
given the size of the global population that is potentially
exposed (nearly 6 million gen1-DES are already implanted worldwide).
It appears that the relevance of altered vascular healing with
incomplete vessel wall and stent reendothelialization and regional
positive remodeling has not been fully recognized. Research
efforts should focus on developing site-specific treatment modalities
that combine antirestenotic and prohealing properties, allow
vascular healing, and preserve late patient safety. Large trials
powered for clinical outcomes, including death and nonfatal
MI, are direly needed to ensure that the frequent but rather
benign restenosis has not been exchanged for the rare but unpredictable
and potentially lethal LST. In the meanwhile, concerns for patient
safety should remain the primary guide for decision making and
caution against indiscriminate use of gen1-DES.
107,108
 |
Acknowledgments
|
|---|
We are indebted to Dr David DeMets and Dr Tom Cook for their
critical review and suggestions, to Nicolas Masson for his precious
technical support in preparing the figures, and to Dr Pierre-André
Dorsaz for expert statistical support.
Disclosures
None.
 |
References
|
|---|
- Mack MJ, Brown PP, Kugelmass AD, Battaglia SL, Tarkington LG, Simon AW, Culler SD, Becker ER. Current status and outcomes of coronary revascularization 1999 to 2002: 148,396 surgical and percutaneous procedures. Ann Thorac Surg. 2004; 77: 761768.[Abstract/Free Full Text]
- Lenzen MJ, Boersma E, Bertrand ME, Maier W, Moris C, Piscione F, Sechtem U, Stahle E, Widimsky P, de Jaegere P, Scholte op Reimer WJ, Mercado N, Wijns W, for the European Society of Cardiology. Management and outcome of patients with established coronary artery disease: the Euro Heart Survey on coronary revascularization. Eur Heart J. 2005; 26: 11691179.[Abstract/Free Full Text]
- Serruys PW, de Jaegere P, Kiemeneij F, Macaya C, Rutsch W, Heyndrickx G, Emanuelsson H, Marco J, Legrand V, Materne P, Belardi J, Sigwart U, Colombo A, Goy JJ, van den Heuvel P, Delcan J, Morel M-A, for the Benestent Study Group. A comparison of balloon-expandable-stent implantation with balloon angioplasty in patients with coronary artery disease. N Engl J Med. 1994; 331: 489495.[Abstract/Free Full Text]
- Mercado N, Boersma E, Wijns W, Gersh BJ, Morillo CA, de Valk V, van Es GA, Grobbee DE, Serruys PW. Clinical and quantitative coronary angiographic predictors of coronary restenosis: a comparative analysis from the balloon-to-stent era. J Am Coll Cardiol. 2001; 38: 645652.[Abstract/Free Full Text]
- Serruys PW, Strauss BH, Beatt KJ, Bertrand ME, Puel J, Rickards AF, Meier B, Goy JJ, Vogt P, Kappenberger L. Angiographic follow-up after placement of a self-expanding coronary-artery stent. N Engl J Med. 1991; 324: 1317.[Abstract]
- Morice MC, Zemour G, Benveniste E, Biron Y, Bourdonnec C, Faivre R, Fajadet J, Gaspard P, Glatt B, Joly P. Intracoronary stenting without Coumadin: one month result of a French multicenter study. Cathet Cardiovasc Diagn. 1995; 35: 17.[Medline]
[Order article via Infotrieve]
- Bertrand ME, Legrand V, Boland J, Fleck E, Bonnier J, Emmanuelson H, Vrolix M, Missault L, Chierchia S, Casaccia M, Niccoli L, Oto A, White C, Webb-Peploe M, Van Belle E, McFadden EP. Randomized multicenter comparison of conventional anticoagulation versus antiplatelet therapy in unplanned and elective coronary stenting: the Full Anticoagulation Versus Aspirin and Ticlopidine (FANTASTIC) study. Circulation. 1998; 98: 15971603.[Abstract/Free Full Text]
- Colombo A, Hall P, Nakamura S, Almagor Y, Maiello L, Martini G, Gaglione A, Goldberg SL, Tobis JM. Intracoronary stenting without anticoagulation accomplished with intravascular ultrasound guidance. Circulation. 1995; 91: 16761688.[Abstract/Free Full Text]
- Condado JA, Waksman R, Gurdiel O, Espinosa R, Gonzales J, Burger B, Villora G, Acquatella H, Crocker IR, Seung KB, Liprie SF. Long-term angiographic and clinical outcome after percutaneous transluminal coronary angioplasty and intracoronary radiation therapy in humans. Circulation. 1997; 96: 727732.[Abstract/Free Full Text]
- King SB 3rd, William DO, Chougule P, Klein JL, Waksman R, Hilstead R, Macdonald J, Anderberg K, Crocker IR. Endovascular beta-radiation to reduce restenosis after coronary balloon angioplasty: results of the Beta Energy Restenosis Trial (BERT). Circulation. 1998; 97: 20252030.[Abstract/Free Full Text]
- Verin V, Popowski Y, de Bruyne B, Baumgart D, Sauerwein W, Lins M, Kovacs G, Thomas M, Calman F, Disco C, Serruys PW, Wijns W, for the Dose-Finding Study Group. Endoluminal beta-radiation therapy for the prevention of coronary restenosis after balloon angioplasty: the Dose-Finding Study Group. N Engl J Med. 2001; 344: 243249.[Abstract/Free Full Text]
- Waksman R, Ajani AE, White RL, Pinnow E, Dieble R, Bui AB, Taaffe M, Gruberg L, Mintz GS, Satler LF, Pichard AD, Kent KK, Lindsay J. Prolonged antiplatelet therapy to prevent late thrombosis after intracoronary gamma-radiation in patients with in-stent restenosis: Washington Radiation for In-Stent Restenosis Trial Plus 6 months of clopidogrel (WRIST PLUS). Circulation. 2001; 103: 23322335.[Abstract/Free Full Text]
- Serruys PW, Wijns W, Sianos G, de Scheerder I, van den Heuvel PA, Rutsch W, Glogar HD, Macaya C, Materne PH, Veldhof S, Vonhausen H, Otto-Terlouw PC, van der Giessen WJ. Direct stenting versus direct stenting followed by centered beta-radiation with intravascular ultrasound-guided dosimetry and long-term anti-platelet treatment: results of a randomized trial: Beta-Radiation Investigation With Direct Stenting and Galileo in Europe (BRIDGE). J Am Coll Cardiol. 2004; 44: 528537.[Abstract/Free Full Text]
- Waksman R, Ajani AE, White RL, Chan R, Bass B, Pichard AD, Satler LF, Kent KK, Torguson R, Deible R, Pinnow E, Lindsay J. Five year follow-up after intracoronary gamma radiation therapy for in-stent restenosis. Circulation. 2004; 109: 340344.[Abstract/Free