Polymer-based sirolimus (Cypher®) and paclitaxel (Taxus®) drug-eluting stents (DES) have reduced rates of restenosis and target lesion revascularization (TLR) compared with bare-metal stents (BMS) and have launched a revolution in the interventional treatment of symptomatic coronary artery disease.1,2 However, this overwhelming enthusiasm has recently been dampened by safety concerns regarding a small but significant increase in the rate of late DES thrombosis compared with BMS.3–5 Clinical studies have identified several patient-related risk factors for late stent thrombosis (LST), such as diabetes, low ejection fraction (EF), renal failure, and discontinuation of anti-platelet therapy.6 Our understanding of the pathophysiology of late DES thrombosis is derived from both preclinical animal and human pathological studies. Little was known about the pathology of DES in humans until a recent study from our laboratory of 40 autopsy cases that demonstrated delayed arterial healing, evidenced by persistence of fibrin, minimal neointimal thickening, and incomplete re-endothelialization compared with 25 matched BMS implants of similar duration.7 This type of detailed morphometric and histological analysis of these specimens has allowed us to understand how the vascular responses of the current generation Cypher and Taxus DES differ from those of BMS, and has also emphasized the importance of assessing healing responses in evaluating the safety of DES.
Lessons from Animal Studies
While it is well recognized that arterial repair after stent placement occurs more rapidly in animals than in man, animal models still hold predictive value since the sequence of biological events associated with arterial repair is remarkably similar.8 Initially, Suzuki and Klugherz published their findings in the pig and rabbit models, respectively, using sirolimus-eluting stents.9,10 Although both showed a significant reduction in neointimal formation, persistence of fibrin was also noted at 28 days. We reported an initial reduction in neointimal formation at 28 days when paclitaxel was delivered on the stent, followed by a loss of benefit at 90 days.11 Moreover, paclitaxel delivery also resulted in persistence of fibrin, increase in inflammation, medial necrosis, and decreased endothelialization compared with BMS at 28 days. These findings were confirmed in a latter study utilizing overlapping, commercially available Taxus and Cypher in the rabbit model.12 In this study, the assessment of endothelialization was carried out on en face using longitudinally cut arteries for scanning electron microscopy analysis, a far superior technique to light microscopy.
Cypher and Taxus stents demonstrated significant delay in endothelialization compared with corresponding BMS at 28 days, and this difference persisted up to 90 days. Furthermore, endothelialization in overlapping segments was further delayed compared with non-overlap segments, probably due to higher doses of the drug and polymer. These data mirrored the findings of delayed healing seen in human DES cases and pointed to the possibility of increased late thrombotic risk with this technology. However, animal models of stenting are limited by their inability to predict the timing of complete healing in man following implantation of DES.
Stent Thrombosis and Arterial Healing in Humans
At autopsy, DES stented arteries are characterized by delayed arterial healing, defined as persistence of fibrin, minimal neointimal thickening, and incomplete re-endothelialization compared with BMS of similar duration (see Figure 1).7 From these data, it remains unclear how long DES remain incompletely endothelialized in humans. In this study, LST was observed in 14 of 23 patients receiving Cypher or Taxus DES for >30 days. Compared with patent DES, significantly poorer endothelial cell coverage of the lumen was consistently documented in all cases of late DES thrombosis (see Figure 1). We also found several procedural and/or pathological factors—such as local hypersensitivity reactions to DES, ostial and/or bifurcation stenting, incomplete apposition of struts, and struts penetration into a necrotic core—to be associated with LST. Similar underlying mechanisms of LST were previously reported for late BMS thrombosis (except for hypersensitivity reactions), which indicates that these lesion and plaque factors may themselves provide additional barriers to healing.11 However, the timing of LST in BMS was significantly earlier than those in DES (BMS: 11 median, 70 days; interquartile range, 33–127, versus DES: 7 median, 173 days; interquartile range, 66–433; p=0.04). Endothelialization is complete at three to four months in BMS and therefore the risk of LST is minimal, whereas in DES the neointima remains unhealed for much longer periods. It remains unknown how much time it takes for DES to be fully healed, as the number of autopsy cases are too few to make reasonable assessments. This may have important implications for the duration of anti-platelet therapy. Since some risk factors for thrombosis are related to the procedure or patient factors, conditions such as bifurcation stenting, excessive stent length, overlapping stents, or incomplete apposition due to underdeployment should be avoided (see Figure 2). Other patient factors such as previous history of bleeding and non-compliance with anti-platelet therapy must be ascertained and use of DES in such patients should be avoided.
Pathological risk factors are also important in terms of minimizing the risk of LST. Penetration of necrotic core is frequently documented in patients with acute myocardial infarction (AMI) or even acute coronary syndromes. Although two pivotal studies showed similar rates of stent thrombosis in AMI patients treated with BMS versus DES,13,14 the safety of this practice cannot be determined due to the limited duration of follow-up (i.e. one year) since recent clinical studies of registries demonstrated an increased risk of cardiac events after one year.3–5 Our own observational pathological studies show a tendency for lesions with penetration of necrotic core by DES struts to be associated with delayed arterial healing and LST. Similarly, clinical studies have demonstrated the feasibility, efficacy, and safety of percutaneous coronary intervention (PCI) with DES for bifurcation lesions;15 however, pathological findings indicate that bifurcation lesions impede arterial healing. Detailed morphometric analysis shows a greater delay in arterial healing at flow dividers, which is likely related to blood flow patterns and shear stress.16 Another pathological risk factor is incomplete stent apposition, which we have observed in significantly more patients with LST compared with those without. Our data suggest that the lack of strut apposition to the vessel wall retards arterial healing and may be the result of positive remodeling due to excessive drug concentration or secondary to a local hypersensitivity reaction. Both result in uncovered struts, which serve as the nidus for stent thrombosis. Similar to procedural risk factors, some of these pathological risk factors may be avoided by an appropriate patient selection (or intravascular ultrasound-guided intervention). Many of these underlying pathological morphologies fall into the category of ‘off-label’ use, which is estimated to represent the majority indication for DES today.17
More recently, we have reported the specific morphometric and histological parameters that significantly correlate with late thrombosis.18 Multiple logistic generalized estimating equations (GEEs) modeling demonstrated that endothelialization is the best predictor of thrombosis. The ratio of uncovered to covered stent struts (RUTSS), one of the morphometric parameters assessed, was the best correlate of endothelialization, suggesting that neointimal coverage of struts could be used as a surrogate for endothelialization. A univariable logistic GEE model of occurrence of thrombus in a stented section versus RUTSS demonstrated a marked increase in risk for LST as the number of uncovered struts increased. We also found heterogeneity of coverage of stent struts, both within individual cross-sections as well as between sections from the same stent. Within the same stent, while some struts showed healing as demonstrated by neointimal growth, others remained bare and served as a nidus for mural thrombus formation.
The middle section of the stent (versus the proximal and distal ends) was the most common location of stent struts lacking neointimal coverage and this was also the most common site of stent thrombosis. This heterogeneity in healing is likely derived from uneven distribution of the drugs, which is caused by multiple factors such as stent design, inter-strut spacing, strut apposition to the vessel wall, and underlying plaque (i.e. plaque composition). Thus, the extent of arterial healing cannot be easily predicted; however, a careful selection of lesion morphology is important to avoid delayed arterial healing.
Endothelial Function Following Drug-eluting Stent Implantation
In addition to the examination of endothelial coverage of the stented segment, understanding the impact of DES on endothelial function is paramount. Recent clinical reports suggest that DES may also impair endothelial responses to acetylcholine- and exercise-mediated vasodilation in humans.19.20 The ability of an intravascular device to facilitate functional endothelial regrowth is crucial since these cells provide essential anti-thrombotic factors in addition to structural integrity to maintain stent patency. Recent in vitro studies examining the effects of paclitaxel and sirolimus on endothelial cells demonstrate increased tissue factor mRNA and protein expression.21,22 Another report suggests that these drugs also increase local production of PAI-1 in cultured endothelial cells.23 Although we have not observed any cases of LST with a complete endothelial layer so far, these findings indicate that the thrombotic risk of DES may be a function of both the extent of endothelial coverage as well as the ability of these cells to conduct their normal functions.
Conclusions
Drug-eluting stents result in delayed arterial healing and poor endothelialization, which are the primary substrates underlying all cases of LST. Additional barriers to healing also increase the risk of thrombosis and include bifurcation lesions, necrotic core penetration, incomplete apposition, and long lesion stenting and overlapping stents. Furthermore, endothelial dysfunction from drugs used on the stent may also play a role in the exaggerated thrombotic risks.
Conflict of Interest Disclosure
Gaku Nakazawa, MD: None; Aloke V Finn, MD: None; Renu Virmani, MD: Company Sponsored Research Support for Medtronic AVE, Abbott Vascular, W.L. Gore, Atrium Medical Corporation, Boston Scientific, NDC Cordis Corporation, Novartis, Orbus Medical Technologies, Biotronik, BioSensors, Alchimer, and Terumo, and Consultant for Medtronic AVE, Guidant, Abbott Laboratories, W.L.Gore, Terumo, and Volcano Therapeutics, Inc.