Drug-eluting stents have revolutionized percutaneous coronary intervention by dramatically reducing the incidence of in-stent restenosis (ISR). Stent fracture (SF), although a common finding with significant negative clinical implications in the peripheral vasculature (1- 4), has never been reported in multicenter randomized clinical coronary stent trials (5- 14). However, recent data (15- 32) have suggested that SF, as defined by catheter coronary angiography (CCA) or intravascular ultrasound (IVUS) evidence of a gap, may be a significant contributor to ISR, particularly in sirolimus-eluting stents (15- 16). ISR has been extensively evaluated by 64-detector computed tomographic angiography (CTA) (33). This study was designed to determine the characteristics and relationship of stent gaps on CTA to ISR, SF, and overlap failure (OF) by CCA, in those selected for invasive angiography.

Two hundred ninety-two consecutive patients, with implantation of 613 stents, undergoing 64-detector CTA were evaluated retrospectively. The patient demographics are shown in (Table 1). All were referred for evaluation of symptoms. Of the 292 patients, CCA data were available in 143, allowing comparison of CTA and CCA findings for 384 stents.

Metoprolol 50 to 100 mg by mouth and/or 5 mg intravenously ×4 was administered to reduce the heart rate to <60 beats/min. The CTA were acquired on the Philips Brilliance-64 scanner (Philips Medical Systems, Cleveland, Ohio) using the 64 × 0.625-mm detector configuration, 120 kVp, 600 to 1,050 mA, 0.2 pitch, and standard or sharp filters (Philips CC and CD filters). Nonionic contrast (Ioversol 350 mg/ml at 5 to 6 ml/s) was used, followed by 50 ml of saline at the same rate using a double-head injector (Optivantage DH, Mallinkrodt, Cincinnatto, Ohio). Estimated effective radiation dose was 13 mSv for men and 18 mSv for women. The cardiac phase best demonstrating each artery (usually 75% of the R-R interval) was analyzed using a dedicated CT workstation (Philips CT Extended Brilliance Workspace, Philips Medical Systems) and a cardiac adaptive multisegment reconstruction algorithm. Curved and straightened multiplanar reformatted images were constructed and evaluated for stent separation and ISR. All stents were evaluated, irrespective of quality.

Stent gap was diagnosed when both of the following criteria were fulfilled on the curved multiplanar reformatted images, and on cross-sectional analysis of the straightened multiplanar reformatted images: 1) partial or complete (circumferential) gap or a “crush” pattern on visual inspection (Figure 1); and 2) confirmation of Hounsfield units (HU) <300 (the lowest HU in the normal stent areas) at the site of separation, consistent with the absence of metallic stent material (Figures 2, 3, 4, 5, 6, 7). The length of the separation (i.e., the distance between the normal stent edges surrounding the separation) was measured. All stents were analyzed; none were considered unevaluable because of motion artifact or adjacent very dense calcification capable of producing a gap secondary to shadowing. The CTA analysis was performed by 2 independent observers who did not partake in the CCA analysis. ISR by CTA was evaluated as previously described (33).

Selective coronary angiography was performed for clinical indications using standard techniques in 143 patients. The reasons for the nonreferral for invasive angiography of the remainder of the patients cannot be accurately ascertained since the patients were referred for CTA by many different community physicians with different thresholds and criteria for proceeding to invasive procedures. Stented areas were reviewed by a separate observer who did not participate in the CTA interpretation. Stents specifically described as single or overlapped were classified accordingly. If specific information was unavailable, stented lengths >40 mm were considered overlapped; the remainder were classified as unknown. Under 3-fold magnification, all stents were evaluated for ISR, defined as >50%, by caliper measurement of percent diameter stenosis, and for separation consistent with SF or OF. Lesions were defined according to the American College of Cardiology/American Heart Association classification (34). SF was classified as partial or complete separation of stent segments. Excessive tortuosity was defined as the presence of 2 or more bends >75° proximal to the target lesion; at least 1 proximal bend >90° (35). Conformability was defined as the degree to which a stent can bend around its longitudinal axis after deployment.

IVUS was acquired in only 5 patients, a number too few to permit meaningful analysis.

Descriptive statistics were used to characterize demographic and peri-procedural data. Differences between the 2 groups (gap present vs. gap absent) were examined with the Fisher exact test for categorical variables or the independent-samples t test for continuous variables. Degree of agreement in the stent gap designation between the 2 observers was computed using Cohen's kappa coefficient. In order to minimize the type I error rate, which could result from examining multiple hypotheses, p < 0.01 was considered a priori to indicate statistical significance. Multivariate analysis utilized 2 separate stepwise logistic regression procedures to identify potential predictors of fracture. The first model included the following demographic and patient characteristics as predictors: sex, age, hypertension, hyperlipidemia, smoking, diabetes mellitus, stroke, history of myocardial infarction, and use of statins, aspirin, clopidogrel, and beta-blockers. The second model included all coronary angiography characteristics as presented in (Table 2). A value of p < 0.05 was used to indicate statistical significance in the multivariate analyses. All statistical analyses were performed with SPSS version 16.0.2 (SPSS, Chicago, Illinois). The study was approved by the Institutional Review Board of Lenox Hill Hospital.

The patient demographics are displayed in (Table 1). Stent gap was noted in 14.4%; diabetic patients were more frequently found in the SF/OF group (p = 0.001) but was not a significant predictor in a multivariate analysis. The mean ± SD interval between the CCA and CTA studies was 57.4 ± 130 days.

There were 384 stents in the 143 patients with both CCA and CTA data, Stent gap was noted in 16.9% of the stents by CTA. There were 4 stents with the crush pattern on CTA and 1 with total separation; the remainder had the partial gap pattern. SF was observed in only 1.0% by CCA; 2 had total and 2 had partial separation. There was a highly significant association of stent gap by CTA with ISR on CCA (Tables 2, 3). There were 229 stents in the 159 patients who did not proceed to CCA; stent gap was noted in 6.6% (p < 0.001 compared with those with CCA follow-up).

ISR on CCA was noted in 46.1% of stent gaps (p < 0.001), and stent gaps accounted for 27.8% of the total ISR (p < 0.001). The HU for the gap area was 196.9 ± 81.1 compared with 481.9 ± 161.8 for the intact portion. The gap length was 2.3 ± 0.9 mm. Stent gap agreement between the 2 observers was very strong (kappa = 0.904). Stent implantation information was available for 124 patients; there were no differences in the interval between implantation and the CCA in stents with (median 618 days, range 1,851 days) and without (median 559 days, range 2,369 days) stent gaps. The only CCA characteristic significantly associated with a stent gap by univariate analysis was stent diameter ≥3 mm; stent gaps were present in 20.0% of ≥3 mm stents compared with 3.4% of <3-mm stents (p = 0.002). By multivariate analysis, stent diameter was not a significant predictor. Stent type, location, length, underlying calcification, post-dilation, and bifurcation stents were not predisposing factors (Table 2). Excessive tortuosity and lack of conformability were not associated with stent gaps; however, their frequency was insufficient to permit meaningful analysis. There were no differences in the frequency of stent gaps noted in single versus overlapped stents (Table 2). However, 26.6% of the stents were in the unknown category and could not be classified as single or overlapped; 46.6% of the gaps were noted in this group. Due to sample-size limitations, it is possible that there was not sufficient power to detect potentially significant predictors in the multivariate analysis. The sensitivity and specificity of the CTA for detection of ISR by CCA were 89.3% and 79.2%, respectively.

(Figures 2, 3, 4, 5, 6, 7) demonstrate the CTA and catheter angiographic characteristics of stent gaps. In (Figure 2)A, there is obvious separation at an overlap site on the CTA. Highlighting the difficulty inherent in the limited sampling of catheter angiographic analysis is the clear gap noted in 1 noncontrast frame (Figure 2C), which is totally unapparent in a second frame (Figure 2D). There was only mild ISR (Figure 2B).

In (Figure 3), there are 2 areas of stent separation by CTA (Figure 3A), with significant ISR at the more proximal site. As in (Figure 2), a single noncontrast frame revealed the stent separation on catheter angiography (Figure 3C); all other frames revealed a normal-appearing stent. In both cases, cross-sectional analyses ((Figure 2)E and Figure 3E) revealed HU at the gap sites that were below the threshold for metallic stent material.

A stent gap is noted on CTA in (Figure 4)A with catheter angiography revealing only moderate ISR (Figure 4B) and an apparently intact stent (Figure 4C). Cross-sectional analysis (Figure 4D) confirmed the more common gap fracture pattern and decreased HU associated with the fracture site.

A totally occluded proximal stent with distal filling of the stent by collaterals, and 2 stent gaps are seen on CTA in (Figure 5)A. Catheter angiography confirmed the total occlusion (Figure 5B), but the stent was intact in noncontrast frames (Figure 5C). Cross-sectional analysis confirmed the gap pattern with clearly decreased HU in the gap areas. In (Figure 6)A, CTA demonstrated obvious separation without ISR. Catheter angiography revealed only mild ISR, and a single noncontrast frame suggested partial fracture (Figure 6B). Cross-sectional analysis ((Figure 6)C and Figure 6D), as in the previous cases, confirmed the gap with decreased HU.

The “crush” pattern is shown in (Figure 7)A and (Figure 7)D in an obtuse marginal branch with severe ISR at a hinge point, a location prone to fracture. Nonetheless, fracture was not visible on catheter angiography ((Figure 7)B and Figure 7C).

This study is the first to systematically evaluate the significance of stent gaps defined by CTA. The strong relationship between stent gaps and ISR has significant implications for the occurrence of SF and OF.

SF has recently been implicated in ISR. In 530 patients undergoing repeat angiography, Lee et al. (15) noted 10 (1.8%) with SF by angiography. Binary ISR for the SF patients was 70%; all required target lesion revascularization. Predisposing factors were sirolimus stents (100%), excessive tortuosity (40%), and stent overlap (50%) with increased rigidity that may act as a fulcrum for metal deformation. Aoki et al. (16) evaluated 307 sirolimus-eluting stents in 280 patients and noted 8 (2.6%) with SF on catheter angiography within 8 months of implantation, confirmed by IVUS. Binary ISR for the SF patients was 37.5%, and 50% underwent target lesion revascularization. Of the 8 with SF, 7 were at overlapped areas. All were located at hinge points. Predisposing factors were RCA SF location (odds ratio [OR]: 10.00) with greater vessel deformation with cardiac motion, saphenous vein graft location (OR: 35.88), and longer stent length with associated higher radial forces (OR: 1.04). Lee et al. (17) evaluated 366 patients with sirolimus stents; SF was noted in 10 (2.7%). Of the 26 with ISR, 10 (38.5%) were associated with SF (7 by angiography and 3 by IVUS). ISR was present in 44% of the SF; 3 (30%) were in overlapped stents. SF was not found in 30 patients with ISR after bare-metal Bx Velocity (Cordis Corporation, Bridgewater, New Jersey) stent implantation. In addition to the above series, there are 13 case reports evaluating 14 patients with SF: 10 involved sirolimus-eluting stents (18- 32). Two patients were evaluated by CTA (28,31).

Proposed mechanisms for SF, in addition to those discussed above, are low stent conformability, overexpansion during post-dilation, and bifurcation lesion with high angulation. In the present study, stent diameter ≥3 mm was the only significant predisposing factor for stent gaps by univariate analysis (Table 2); stent gaps were present in 20.0% of ≥3-mm stents compared with 3.4% of <3-mm stents (p = 0.002). However, it was not a significant predictor in multivariate analysis. Stent length, location and type, overlapped stents, post-dilation, underlying calcification, and bifurcation stents were not significantly related. Excessive tortuosity and lack of conformability were rarely noted, and their contribution could not be evaluated.

The high incidence of ISR associated with stent gaps most likely results from the absence of drug-elution protection from neointimal hyperplasia, or a drug-free zone, at the gap site. Other possibilities include broken struts causing local mechanical stimulation of the vessel wall, resulting in inflammation and development of intimal hyperplasia, as well as local uncovered unstable plaque.

The ability of CTA to confidently identify stent gaps is dependent not just on demonstration of a “gap,” which may be more apparent than real, depending on the window settings. Rather, the gap must be associated with HUs below the minimum density of stent material, which is independent of window center and width, and the absence of artifact that may contribute to this finding. The study must be carefully evaluated for shadowing effects of adjacent dense calcification and motion artifacts.

In the absence of artifacts, the most likely explanations for the hypodense gap are fracture- or OF-related absence of strut material. Overinflation, with spreading of struts without true fracture, cannot be excluded. Bifurcation stents, with inflation into a side branch and possible strut damage, were not associated with a higher incidence of stent gaps (Table 2). The absence of strut material for a distance greater than the normal average interstrut distance (1 mm) is convincing evidence for an uncovered portion of the artery. The 46% incidence of ISR in stent gaps and the 28% association of stent gaps with ISR support the pathologic significance of this observation and the likely presence of SF or OF, even though there is no confirmatory gold standard.

The most common gap was a hypodense partial gap suggestive of incomplete SF. The crush pattern, manifested by a flattening of the stent with a hypodense gap, was less frequent. Only 1 case of total separation was noted.

The infrequency of catheter angiographic identification of stent separation in the present series (1.0%) is similar to the 1.8% to 2.7% in previously reported studies (15- 17). The discrepancy between the CTA (16.9%) and catheter angiography frequency in this and previous reports has several possible explanations.

First, successful detection of SF or OF by catheter angiography is directly related to the gap length and inversely related to deviation of image acquisition from the plane perpendicular to the gap; overlap of strut edges on nonperpendicular acquisitions may render the gap invisible. This proof of concept is clearly illustrated in (Figure 2) and (Figure 3), in which the gap was visible only in very few frames and unapparent in all others, in (Figure 4) in which the absence of struts was apparent only by IVUS (not shown), and in (Figure 7) with partial SF suggested on a single catheter angiographic frame yet evident on IVUS (not shown). The problem is magnified by the limited number of acquisitions, typically 5 to 8 for the left coronary artery and 2 to 4 for the right coronary artery. The 3-dimensional quality of CTA renders it immune to the issue of image acquisition; the gap site can be inspected from every conceivable angle.

Second, lesser degrees of separation may have been present in this series compared with prior reports. CTA, as discussed above, is very likely more sensitive in detecting lesser separation if present.

Third, the CTA stent gap findings may not represent true SF or OF. This is highly unlikely since, as discussed in the previous text, there is no other plausible explanation for the decreased HU (<300). Adjacent calcified plaque was not of sufficient density to produce HU reduction by shadowing, and all the studies were of sufficient quality to eliminate other sources of artifacts as contributors. In addition, the high frequency of catheter angiography-proven ISR associated with CTA stent gaps is similar to the previously reported association of ISR with SF diagnosed by catheter angiography and intravascular ultrasound (15- 17). IVUS would seem to be well suited for SF/OF identification but has not been extensively evaluated. Careful frame-by-frame analysis is essential, and the almost ubiquitous presence of superimposed calcification may render differentiation of stent material from calcified plaque difficult, if not impossible. In addition, the pullback may skip over a fracture site. Nonetheless, expert IVUS evaluation might be expected to yield results similar to CTA. Its highly invasive nature restricts its use to patients already undergoing catheter angiography.

The lower incidence of stent gap in patients not referred for CCA compared with those who underwent invasive evaluation (6.6% vs. 16.9%, p < 0.001) may reflect a lower incidence of ISR-related symptoms requiring further testing.

In patients with a gap in a single stent, SF is the most likely explanation. In prior reports of fracture at an overlap site, stent separation has always been attributed to fracture rather than to the possibility that the stents were never completely overlapped or migrated over time. The lack of detection by catheter angiography suggests that the stents appeared successfully overlapped at the time of their implantation. However, the intrinsic limited acquisition issue is exaggerated by the even more limited number (1 to 2) of post-stent acquisitions, and inadequate overlap at the time of implantation cannot be excluded. In the present study, stent information was missing for 26.6% of the stents, which could not be classified as single or overlapped; 46.6% of the gaps were noted in this group. Consequently, the true frequencies of possible SF and OF could not be determined. In the 73.4% with implantation data, stent gaps were equally present in single versus overlapped stents.

The patient population represents a retrospective analysis of a consecutive series of stented patients who underwent CTA for clinical indications, with only 49% undergoing CCA, rather than a prospective, consecutive series of patients who underwent stenting with follow-up CCA and CTA. Consequently, there is significant selection bias, and the study very likely overestimates the incidence of SF or OF in the general population of stented patients. Reflecting the problems inherent in tertiary referral centers, in which data regarding stents implanted elsewhere may not be available, is the incomplete stent implantation data in those who underwent CCA. However, sufficient numbers of stents were available for meaningful analysis, and this study is the first to address this topic in a large series of patients. SF and OF cannot be absolutely confirmed in the absence of postmortem examination. However, the absence of other plausible explanations, the 46% incidence of ISR in stent gaps, and the 28% association of stent gaps with ISR support the pathologic significance of this observation.

The incidence of stent gaps with possible fracture or OF in a large series of stents evaluated by CTA and CCA was 16.9%, representing 28% of the total ISR population; ISR was present in 46% of SF/OF. This very strong association of stent gaps with ISR suggests that, in patients with drug-eluting stents, it may not be failure of the drug-eluting compound to prevent neointimal hyperplasia that is responsible for all ISR, but rather lack of exposure of the arterial segment to the compound at a gap site in a substantial number. Greater emphasis on those factors that promote SF and on manufacturing techniques that prevent SF appear to be in order. In overlapping stent implantations, IVUS may prove to be a reliable tool for verifying the accuracy of overlap at the time of insertion. The inability of catheter angiography to detect stent gaps and the highly invasive nature of IVUS, and its potential confounding by calcified plaque, suggest that CTA is the diagnostic procedure of choice.