The introduction of drug-eluting stents (DES) has made a significant impact on restenosis rates, as shown in many randomized trials (1- 2) and in autopsy studies (3), thus providing an overall benefit to patients by decreasing the incidence of target lesion revascularization. One of the main drawbacks to localized stent-based drug delivery, however, is the associated long-term delay in healing (i.e., lack of re-endothelialization and poor strut coverage) (3), whereby there is an associated infrequent but real risk of late stent thrombosis (LST) (4). Although clinical trials support differences in angiographic late lumen loss in patients receiving first-generation sirolimus-eluting stents (SES) (Cypher, Cordis Corp., Miami Lakes, Florida) and paclitaxel-eluting stents (PES) (Taxus, Boston Scientific, Natick, Massachusetts) (5), it remains unclear whether long-term histologic responses between them exist and how this relates to the time course of arterial healing and mechanism(s) of LST.

Supportive evidence of delayed arterial healing in response to coronary DES implants in humans is mostly derived from autopsy studies in which they are found to exhibit severe suppression of smooth muscle cell infiltration, persistent fibrin, and poor endothelial coverage and the presence of uncovered struts (3,6), particularly those with LST. From a clinical perspective, the pathology data reinforce the importance of these attributes as essential to the maintenance of long-term luminal patency.

The clinical impact of delayed arterial healing becomes significant considering that after initial approvals in 2002, first-generation polymer-based DES were implanted in millions of patients worldwide, and SES and PES still remain viable contenders in the device market. Therefore, understanding the relevance of histopathologic findings from devices used in a global practice of medicine is paramount. Along these same lines, the drawbacks to first-generation DES used in a grand-scale clinical setting can provide valuable information, which will perhaps improve the management of patients receiving current- and next-generation DES in addition to improving future stent designs. Moreover, the occurrence of LST is a real phenomenon and has been described up to 5 years post-implantation and beyond, where appropriate patient management becomes critical to survival. Known fundamental differences exist between SES and PES including the choice of drug, release kinetics, polymer, and stent platform.

The present study emphasizes unique mechanisms of delayed arterial healing and LST in SES and PES deployed for on- or off-label indications using a registry database of autopsy cases. The present findings have important clinical significance considering that our data suggest that the current regimen of dual antiplatelet therapy (i.e., 12 months) after DES implantation may be adequate for the vast majority of DES placed for on-label indications. Conversely, a significantly higher number of DES implanted for off-label indications remained unhealed after 12 months, potentially suggesting the importance of extending the period of dual antiplatelet therapy beyond 12 months in this setting.

Available DES from 174 autopsies (DES for 230 lesions) from the CVPath Institute's stent registry were examined in which 93 of the stented lesions were used in previous reports (3,6- 8). Of these, 47 patients (total of 59 lesions; SES = 25 and PES = 34) survived less than 30 days post-implantation, whereas 127 patients (total of 171 lesions; SES = 77 and PES = 94) survived longer than 30 days. In hearts with multiple stents, overlapping and consecutively implanted DES were treated as 1 lesion, whereas DES showing gaps of >5 mm were considered separate lesions, as described previously (6). Clinical histories and cardiac catheterization reports when available were reviewed. The cause of death was reported as stent-related coronary death, nonstent-related cardiac death, or noncardiac related, as previously defined (6).

The hearts were fixed in 10% neutral-buffered formalin, and the stents were carefully removed and radiographs obtained before submitting for methylmethacrylate embedding. The entire stent was then sawed consecutively from the proximal to distal ends at 2- to 3-mm intervals along the entire length. The sections were prepared at 6 μm and stained with hematoxylin and eosin and modified Movat pentachrome. Depending on the device length, the number of histologic sections available for analysis ranged from 4 sections for 8-mm stents to a maximum of 15 sections for longer, 40-mm lengths. Based on this technique, for lesions with a stent in place for ≥30 days, an average of 8.81 ± 2.43 struts per section were analyzed for a total number of 7,715 struts in 876 stent cross sections.

In-stent coronary thrombi (nonocclusive or occlusive) were identified by platelet masses occupying >30% of the cross-sectional luminal area, whereas restenosis was defined as >75% cross-sectional luminal area narrowing by neointimal tissue. LST, consistent with the clinical definition, was defined as thrombosis within a stent implanted >30 days (7). Morphometric measurements (IPLab, BD Bioscience Bioimaging, Rockville, Maryland) were performed after digital image capture and included the external elastic lamina, plaque area, stent area, and lumen area in addition to the thickness above each strut, as described previously (8).

The severity of fibrin and inflammation was scored based on a grading scale of 0 to 4, as described previously (3). The number of stent struts associated with giant cells and the maximum number of eosinophils around each strut were also assessed. Uncovered struts were identified by an absence of overlying tissue except for thrombus consisting of platelets/fibrin; the ratio of uncovered-to-total stent struts per section was calculated as described previously (6). In select cases, Luna's method (9) was used to confirm the presence of eosinophils, whereas T lymphocytes were recognized by immunostaining with an antibody directed against CD45RO.

The incidence of unhealed stents was determined for coronary implants distinguished by those in place for <12 and ≥12 months. The cutoff was selected based on the American Heart Association and the American College of Cardiology recommended duration of dual antiplatelet therapy after DES (10). The total number of stent struts per section was counted, and the number of uncovered struts were summed for each consecutive section and reported as a percentage (total number of struts per section). An unhealed stent was defined by a >30% ratio of uncovered-to-total stent struts per cross section, as described previously (6), in which uncovered struts were recognized by the lack of neointimal growth involving smooth muscle cells and/or surface endothelium. A threshold for variations in the ratio of uncovered-to-total stent struts per cross section was derived from receiver-operating characteristic curve analysis by equating sensitivity (75%) and specificity (76%) levels near maximum achievable diagnostic accuracy (79%), as reported previously (6).

Stent comparisons were further performed based on clinical on-label or off-label indications, as summarized in (Table 1). Coronary stents indicated for the treatment of de novo native lesions of lengths <30 mm were considered as on-label use, whereas stents deployed for acute myocardial infarction or bifurcation lesions, left main artery, bypass graft, restenosis, chronic total occlusion, or lesion lengths >30 mm were listed as off-label, as defined previously (11).

Continuous variables with normal distribution are expressed as mean ± SD. The normality of distribution was tested with the Wilk-Shapiro test. Comparison between patients with SES and PES was examined by Student t test for continuous variables with a normal distribution, whereas categorical variables were analyzed by the Fisher exact test. A Wilcoxon rank-sum test was used for comparisons of non-normally distributed data, which were expressed as median and interquartile range. Statistical comparisons among DES with varying implant durations were performed by a Kruskal-Wallis test followed by post hoc analysis using a Wilcoxon rank-sum test to assess significant differences. A value of p < 0.05 was considered statistically significant.

Patient characteristics such as age, sex, and risk factors for coronary artery disease were similar between those receiving SES (19 patients) and PES (28 patients) who died <30 days after stent placement (Table 2). The number of stents per patient and cause of death were also similar between stent types. Analysis of individual lesions with an implant duration of <30 days (SES = 25 days and PES = 34 days) for the time of implantation, indication for stent placement, stent length, lesion location, and prevalence of overlapping stents were comparable between groups. The incidence of early stent thrombosis was equivalent for lesions with SES (44%) and PES (38%) (p = 0.79). Histologically, no differences in the extent of inflammation and fibrin deposition were noted between SES and PES implants at <30 days.

Patient characteristics such as age, sex, and risk factors for coronary artery disease for stent implants with a duration of ≥30 days were similar between SES (61 patients) and PES (66 patients) (Table 3). The cause of death was also similar between groups, with 30% of patients dying of stent-related causes. In the lesion-based analysis (SES = 77 lesions and PES = 94 lesions), stent duration, indication for stenting, stent length, lesion location, and prevalence of overlapping stents were comparable in those with DES (Table 3). Of the total 171 lesions, 93 (54%) were treated for off-label indications (SES, 40 of 77 [52%]; PES, 53 of 94 [56%]; p = 0.64). The incidence of LST was similar among patients with DES and was observed in 21% of SES compared with 27% of PES (p = 0.47). Chronic total occlusion of the stent with organized thrombus was also equally observed in both SES and PES (6% each; p > 0.99). Finally, restenosis was documented in 8% of SES and 5% of PES lesions (p = 0.55).

Although external elastic lamina stents and plaque areas were compared between lesions with SES and PES, the mean and maximum neointimal thicknesses were significantly greater in PES (mean 0.13 mm; range 0.07 to 0.20 mm and maximum 0.23 mm; range 0.13 to 0.37 mm) than SES (mean 0.10 mm; range 0.04 to 0.15 mm; p = 0.04 and maximum 0.17 mm; range 0.06 to 0.28 mm; p = 0.04) (Table 4). The heterogeneity in neointimal thickness between sections (neointimal thickness variation in Table 4) was also significantly greater for PES (0.14 mm; range 0.08 to 0.31 mm) versus SES (0.10 mm; range 0.03 to 0.22 mm; p = 0.02). Moreover, there was a progressive and significant increase in neointimal thickness beyond 9 and 18 months in lesions with PES without evidence of LST (Figure 1A); although similar trends were observed for SES, findings were of borderline significance (Figure 1A).

Accumulated fibrin was significantly greater in PES compared with SES (fibrin score: PES 1.8; range 1.0 to 2.5 vs. SES 0.8; range 0.0 to 2.0; p = 0.001) (Table 4). On the contrary, SES implants were associated with a significantly greater inflammatory score compared with PES (SES 1.3; range 0.5 to 2.0 vs. PES 1.0; range 0.5 to 1.5; p = 0.007). The contributing cells resulting in greater inflammation observed with SES were predominantly eosinophils, lymphocytes, and giant cells, and although the incidence of malapposition was comparable in SES and PES, the underlying mechanism is likely related to excessive inflammation for SES as opposed to persistent fibrin in the case of PES.

The overall incidence of uncovered struts was similar between devices, and both were associated with a decreased frequency of unhealed stents beyond 12 months (Figure 1B). A further analysis revealed nearly complete healing in stents placed for on-label indications with implant durations of >12 months, whereas the majority of DES with off-label use remained unhealed beyond this similar time point (Figure 1B).

Underlying pathologic causes of in-stent thrombosis were determined by histologic review (Table 5). There were no significant differences in variables associated with early stent thrombosis (i.e., <30 days) between SES and PES. Significant differences, however, were found for the underlying tissue reactions associated with LST for DES implanted for ≥30 days. For SES, the incidence of LST was significantly associated with localized hypersensitivity consisting of eosinophils, lymphocytes, and giant cells throughout the stented segment (p = 0.0005) relative to PES, whereas the latter was attributed to malapposition secondary to excessive fibrin deposition on the abluminal surface of the stent (p = 0.03).

Localized hypersensitivity was documented in 5 cases involving 7 lesions treated by SES in which LST developed ((Table 6),Figure 2A). The patient mean age was 52 years (range 40 to 62 years), and the mean stent length was 27 mm (range 18 to 41 mm). The majority of patients receiving SES died in the very late phase (>1 year) in which the mean implant duration was 649 days (range 112 to 990 days). Malapposition was observed in 5 lesions (71%) with a mean section of struts from the vessel wall of 944 μm (range 320 to 1,620 μm). Consistently, histologic sections of SES with LST showed severe inflammation consisting of eosinophils, lymphocytes, and giant cells throughout the entire stented segment; however, granulomatous inflammation was only localized to a few stent struts. In most SES with severe inflammation, there was positive remodeling of the vessel resulting in a malapposition.

In contrast, the primary contributor to LST in PES implants was malapposition secondary to excessive strut fibrin, as observed in 7 lesions in 6 patients ((Table 6), Figure 2B) with a mean age of 51 years (range 37 to 69 years). The mean stent length was 29 mm (range 16 to 45 mm), and the mean implant duration was 611 days (range 75 to 1,200 days). The mean distance separating the struts from the vessel wall was 404 μm (range 180 to 620 μm). The luminal surface generally lacked endothelial cell coverage as well as evidence of granulation tissue consisting of macrophages, smooth muscle cells, or proteoglycan matrix (Figure 2B).

Despite delayed arterial healing and a similar incidence of LST in first-generation coronary DES obtained at autopsy, the underlying composition of the neointima associated with SES and PES implants is clearly distinct. In patients with SES, there was greater inflammation involving eosinophils, lymphocytes, and giant cells (hypersensitivity reaction) resulting in positive remodeling and malapposition. In contrast, excessive para-strut fibrin and malapposition was associated with PES implants. Intuitively, these findings support distinct mechanisms of LST, coinciding with very diverse stent platforms.

Similar to clinical studies, SES showed significantly greater neointimal suppression compared with PES; however, both DES were associated with increased neointimal growth with time, so-called late catch-up. The prevalence of unhealed stents in patients surviving ≥12 months was different based on the indication for placement as nearly complete healing was largely confirmed for on-label use, whereas a significant number of stents (irrespective of the platform) deployed for off-label indications essentially remained unhealed.

As previously reported, mechanism(s) of LST are likely multifactorial (3). Although there is a certain commonality in the mechanism of LST for both SES and PES in that all patients demonstrated poor endothelialization, our findings indicate the final stimulus for thrombus development may be different based on DES type. The disparities in vascular responses observed in the present study regarding the stent milieu are reminiscent of preclinical models and are undoubtedly attributable to differences in the drug load, polymer coating, and unique elution profile of each device.

Only mild inflammation was exhibited by SES in patients with short-term survival; however, with longer implant durations, inflammation became more prevalent as the majority of hypersensitivity cases were documented in devices >1 year old (Table 4). The findings are remarkably consistent with preclinical DES implants in porcine coronary arteries, which typically show escalating amounts of inflammation over time (12). In SES implants in humans, the inflammatory infiltrate is composed predominantly of eosinophils, multinuclear giant cells, and lymphocytes, which leads to positive remodeling, malapposition, and potentially LST. In our cases, multiple SES implanted in the same patient invariably demonstrated these findings irrespective of location, suggesting the existence of a localized immunologic reaction. As previously postulated (13), the hypersensitivity reaction is likely attributed to the polymer rather than drug, which is presumably completely eluted by 3 months.

Our finding that greater fibrin accumulated around struts in PES also remains consistent with the vascular responses seen in previous preclinical studies. For example, Farb et al. (14) showed a dose-dependent increase in fibrin deposition and medial necrosis after deployment of PES in rabbit iliac arteries, and similar dose escalatory findings were reported in a porcine model (15). Therefore, we believe that paclitaxel itself is responsible for excessive fibrin deposition.

Despite a marked difference in neointimal composition between DES platforms, all implants exhibiting LST were associated with positive vessel remodeling and, in the majority, stent malapposition. The reported occurrence of late acquired stent malapposition after DES implantation exists in approximately 10% of lesions (16- 17). Cook et al. (18) reported that stent malapposition was more frequent in patients with LST (77%) than without (12%). Recent meta-analysis data also support the finding of late stent malapposition as significantly higher in DES compared with bare metal stents (BMS) (19) and was associated with LST. In the present study, the overall incidence of malapposition was similar between SES and PES. Accordingly, however, the underlying mechanisms of this phenomenon are fundamentally different between platforms.

In the present study, early stent thrombosis and LST were equally observed in SES and PES, which is consistent with the clinical experience (20- 21). Although PES are associated with greater neointimal formation, the incidence of uncovered struts, the most powerful predictor of risk of LST, was similar between DES. This, however, was associated with greater heterogeneity in the healing response to PES relative to SES, as evidenced by greater disparities in neointimal thickness. Because histologic sections of PES demonstrating the thinnest neointima are typically accompanied by persistent fibrin, the heterogeneity of arterial healing may result from an uneven distribution of drug and polymer. Further support for variations in available paclitaxel in a single stent comes from scanning electron microscopy studies demonstrating webbing and delamination of polymer, which is a frequent finding in PES (22). Finally, angioscopy studies by Takano et al. (23) also demonstrate heterogeneity of stent strut coverage in PES compared with SES.

The gradual increase in neointimal thickness over time for both first-generation platforms reinforces the notion that the arterial responses to DES is a dynamic process lasting well beyond the 6- to 8-month time point recommended for angiographic follow-up in clinical trials. This time course of arterial healing contrasts with the pattern of neointimal formation after BMS placement, which typically peaks at 6 to 12 months with regression thereafter (24- 25). Clinical studies of DES also report a gradual increase in target lesion revascularization and neointimal volume by intravascular ultrasound occurring over 4 years (26- 29).

Although slow to develop, progressive neointimal growth in DES is likely related to persistent fibrin and inflammation. In both DES platforms, biological signs of a drug effect such as fibrin remain beyond the reported durations of drug release. Fibrin degradation products, in particular fibrin fragment E, are a known initiator of smooth muscle cell migration and proliferation (30- 31), a phase that generally occurs early after BMS placement. In addition to fibrin, the presence of inflammation is yet another plausible explanation for the late increases in neointimal formation associated with DES. Nonerodible polymers used in first-generation DES, namely, poly(styrene-b-isobutylene-b-styrene) for PES and polyethylene-covinyl acetate, and poly n-butyl methacrylate for SES are associated with chronic inflammation and, in particular for the latter, eosinophils (3,13,22,32) in which the release of various growth factors as a result of inflammation might serve to promote smooth muscle cell proliferation (33). In a previous study of human coronary stents from our laboratory, greater intimal inflammation was found to be an independent predictor of restenosis (34). Although a dramatic late catch-up in restenosis has not been reported in DES, target vessel revascularization rates do increase with time, which is consistent with the gradual growth of neointima seen in the present study (Figure 1A). Therefore, the possibility exists that late restenosis may erode into the benefit of DES compared with BMS. Consistent with the gradual increase in neointima formation over time, it was also found that arterial healing in both DES platforms was continuous over long periods. Despite the decreased prevalence of unhealed stents (defined as >30% uncovered struts) for implants in place for >12 months, the finding of unhealed struts was heavily contingent on the clinical indication for stent placement regarding on- versus off-label use. In our series, better outcomes regarding stent healing were found with DES deployed for on-label indications (demonstrating nearly complete healing after 12 months) compared with off-label use, although strut coverage was still numerically greater in off-label DES implanted for >12 months. These findings may have implications for the duration of antiplatelet therapy after DES placement. Our data suggest that for the vast majority of DES placed in the context of on-label, the current (i.e., 12-month) regimen of dual antiplatelet therapy for DES implants may be adequate. It is also noteworthy that a considerably higher number of DES deployed for off-label indications remained unhealed beyond 12 months, and, therefore, continued dual antiplatelet therapy for these patients is advisable.

Although the pathologic finding of stents at autopsy might not entirely represent those deployed in survivors, the large number of cases in our registry, however, should allow us to provide a more detailed and reliable assessment of histopathologic findings representative of the general population. Despite the fact that clinical data such as antiplatelet therapy were not available in all cases, the aim of the current study was to investigate the time course and differences in arterial healing between SES and PES; therefore, the absence of complete clinical information does not preclude our assessment.

The present data suggest that DES exhibit divergent mechanisms of LST in which hypersensitivity likely plays a significant role in SES, whereas for PES, the etiology seems to be an association with excessive fibrin deposition on the abluminal surface with malapposition. Another important finding was the nearly complete healing in DES placed for >12 months with confirmed on-label use, whereas off-label indications for both stents resulted in incomplete healing, even in DES in place for >12 months.