A meta-analysis of published studies was conducted to evaluate the incidence, predictors, and clinical outcomes of stent fractures. Eight studies with 108 stent fractures in 5,321 patients were analyzed using the Bayesian method. Study end points included in-stent restenosis (ISR) and target lesion revascularization (TLR). The mean incidence of stent fracture per patient was 4.0% (95% confidence interval 0.4% to 16.3%). All cases, except 1, were reported with sirolimus-eluting stents. The incidence of stent fracture was 30.4% in the left anterior descending coronary artery, 10.9% in the left circumflex coronary artery, 56.4% in the right coronary artery, <0.01% in the left main coronary artery, and 1.7% in saphenous vein grafts. The probability of stent fracture was significantly higher in the right coronary artery than in the left anterior descending and left circumflex lesions (p <0.01). Left main stents were less likely to fracture compared to those in all other vessels (p <0.01). The probability of stent fracture was significantly increased in overlapping stents (7.5% vs 2.1%, p = 0.01) and long stents (46 vs 32.5 mm, p <0.01). Lesions with stent fractures had higher rates of ISR (38% vs 8.2%, p <0.01) and TLR (17% vs 5.6%, p <0.01). Conversely, the probability of stent fractures was higher in patients with ISR (12.8% vs 2.1%, p <0.01) and TLR (8.8% vs 2.7%, p <0.01). In conclusion, although not always associated with clinical sequelae, the probability of ISR and TLR is increased with stent fracture. Conversely, the probability of stent fractures is increased in lesions with ISR or TLR, thus raising the need for surveillance and management guidelines for at-risk patients.
Coronary stent fracture is a relatively recently recognized phenomenon, first reported after a bare-metal stent implantation in a venous bypass graft in 2002. Soon thereafter, a report of adverse events with bare-metal stents included 9 cases of radial fracture associated with recurrent angina, repeat intervention, or coronary artery bypass grafting. The first case of drug-eluting stent (DES) fracture in coronary arteries did not appear until 2004. Since then, several additional reports have appeared. These reports are limited by their small size and/or single-center designs. We conducted a meta-analysis as well as a systematic critical review of the published research with the goal of providing perspective on the incidence and the clinical, angiographic, and procedural variables associated with coronary stent fractures and their clinical sequelae.
Methods
We searched the PubMed, Medline, and Cochrane databases for published studies, prospective and retrospective, in English from 2000 to 2008 using the Medical Subject Heading terms “coronary stent fracture,” “stent fracture,” “sirolimus-eluting stent (SES) fracture,” “paclitaxel-eluting stent (PES) fracture,” and “DES restenosis.” Abstract lists from the 2007 and 2008 scientific meetings of the American Heart Association, the American College of Cardiology, and Transcatheter Cardiovascular Therapeutics were searched. Published reviews and editorials were also reviewed.
Studies were included if they met the following criteria: (1) a follow-up duration of ≥6 months, (2) >50 patients in each study, (3) the use of DES for percutaneous coronary intervention, (4) ≥50% of patients with follow-up repeat angiography with or without intravascular ultrasound (IVUS), and (5) publication in a peer-reviewed journal or presentation at a major national cardiology meeting.
Stent fracture was defined as either strut fracture or complete separation of the stent segments, as assessed by coronary angiography, plain fluoroscopy, and/or IVUS at follow-up, which was absent at the time of the index procedure. Reported clinical outcomes included in-stent restenosis (ISR), stent thrombosis, and need for target lesion revascularization (TLR). ISR was defined as a >50% diameter stenosis within the stented segment visualized on follow-up angiography. Stent thrombosis was defined as an angiographically documented filling defect in proximity to a previously placed stent on repeat coronary angiography. TLR was defined as treatment for recurrent angina or signs of ischemia and a >50% diameter stenosis at the target lesion on follow-up angiography.
We used Bayesian methods to fit all models in the present study because of the unbalanced structure of the data, nonlinearity of the models, and flexibility of the Bayesian paradigm. All priors were vague but proper and were neutral relative to any hypotheses that might be entertained. Models were fit in WinBUGS (MRC Biostatics Unit, Cambridge University, Cambridge, United Kingdom) using Markov-chain Monte Carlo techniques. Convergence of the Markov-chain Monte Carlo algorithms appeared to be immediate; we decreased 1,000 samples for burn-in and ran all Markov-chain Monte Carlo chains for 200,000 iterations. We report posterior medians for most parameters because many posterior distributions were highly skewed; we report 95% central intervals. Reported p values are 1 sided.
We tried fixed- and random-effects models for the probability that a patient or a lesion had a fracture, and the random-effects models fit substantially better. We then included percentage follow-up as a covariate in these models. In further models, the lesion was the unit of observation, because this was how data were reported. Yang et al reported the number of stents rather than number of lesions for the nonfractured stents. We fit a model to data from Aoki et al, Umeda et al, Chung et al, and Okumura et al to estimate the number of lesions per patient and applied it to Yang et al’s number of subjects to calculate Yang et al’s number of lesions without fractures, and we used that estimated number in all further calculations. The model assumed that the number of additional lesions per subject was distributed as a Poisson with mean λ times the number of patients in the study; λ was estimated to be 0.196 (SE 0.015, 95% confidence interval [CI] 0.168 to 0.227), and we estimated that Yang et al had 552 lesions without fractured stents.
Vessel (left anterior descending coronary artery, left circumflex coronary artery, right coronary artery, left main coronary artery, and saphenous vein graft [SVG]) is a polytomous event. We calculated the probability of a fractured stent given the vessel using Bayes’s theorem (inside the already Bayesian analysis) as P(fracture | vessel) = P(vessel | fracture) × P(fracture)/{sum vessel [P(vessel | fracture) × P(fracture)]} and similarly for nonfractured stents. The probability of a fractured stent, P(fracture), in a lesion was calculated using our random-effects model, and the probabilities of a given vessel given that the stent fractured or did not fracture were estimated using a fixed-effects model. Different sets of studies were used for each part of the calculation. We fit a similar model for dichotomous events ISR, TLR, bifurcating stents, and overlapping stents.
Stented length is continuous and was reported by studies as mean ± SD. We used a random-effects model with separate grand means and variances for lengths of fractured stents and lengths of nonfractured stents. We used fixed effects for the conditional models.
Results
We identified 8 eligible studies ( Table 1 ) for meta-analysis. Most studies reported clinical and angiographic follow-up data at 6 to 12 months. All the studies used angiography for follow-up, with or without IVUS. A total of 5,321 patients with 108 stent fractures from 8 studies contributed data to this meta-analysis.
Variable | Aoki et al | Lee et al | Chung et al | Okumura et al | Yang et al | Umeda et al | Kim et al | Yamada et al |
---|---|---|---|---|---|---|---|---|
Number of patients with follow-up angiography | 256 | 530 | 3,065 | 138 | 479 | 382 | 415 | 56 |
Criterion for follow-up angiography | Prospective follow-up | Clinically driven repeat angiography | Retrospective collection of SF cases from 13 centers | Prospective follow-up | Retrospective analysis | Prospective follow-up | Retrospective collection of data | Prospective follow-up |
Average follow-up time (months) | 8 | 7 | 6.2 | 12 | 6–9 | 6–9 | 6 | 6 and 12 |
Percentage follow-up angiography | 91.4% (256/280) | 19.4% (530/2,728) | 50% | 91% (138/151) | 71% (479/675) | 90.5% (382/422) | 83% (415/500) | 100% |
Definition of SF | Angiographic or IVUS evidence of SF | Minor, moderate, and severe | Disruption, avulsion, and displacement | Angiographic or IVUS evidence of SF | Angiographic or IVUS evidence of SF | Complete and partial fracture | Stent strut discontinuity with identifiable gaps | Stent strut dissociation |
Type of stent | SES | SES/PES | SES | SES | SES | SES | 6 SES and 1 PES | SES |
Incidence | 3.1% per patient (8 SFs/256 patients) or 2.6% per lesion (8 SFs/307 lesions) | 1.9% (10 SFs/530 patients) | 0.84% (26 SFs/2,065 SES) | 2.8% per patient (4 SFs/138 patients) or 2.4% per lesion (4 SFs/169 lesions) | 5.6% per patient (27 SFs/479 patients) or 3.2% per lesion (27 SFs/686 lesions) | 8.4% per patient (32 SFs/382 patients) or 7.7% per lesion (33 SFs/430 lesions) | 1.6% per patient (7 SFs/415 patients) | 5.3% (3 SFs/56 patients) |
ISR | 50% (4 restenoses/8 SFs) | 60% (6 restenoses/10 SFs) | 65% (16 restenoses/37 SFs) | 100% (4 restenoses/4 SFs) | 22.7% (5 restenoses/22 SFs) | 15.2% | 14.3% (1 restenosis/7 SFs) | 100% (3 restenoses/3 SFs) |
TLR | 50% (4/8 SFs) | 70% (7/10 SFs) | 30% (11/37 SFs*) | NA | 9.1% (2/22 SFs) | 9.4% (3/32 SFs) | 0 | NA |
In-stent thrombosis | 0 | + (1 thrombosis/10 SFs) | 0 | 0 | 0 | 0 | 0 | + (1 thrombosis/10 SFs) |
Proportion of restenosis related to SF | 9.09% (4 SF restenoses/44 SES restenoses) | NA | 6.3% (16 SF restenoses/252SES restenoses) | 31% (4 SF restenoses/13 SES restenoses) | 7.4% (5 SF restenoses/67SES restenoses) | NA | 3.3% (1 SF restenoses/30 restenoses) | NA |
IVUS confirmation | + | 0 | + (43% of cases) | + | + | + (operator decision) | + | + |
Risk factors | Implanted stent length, SVG location, right coronary location | Placement in tortuous vessels, overlapping stents, balloon overexpansion | Stent length, overlapping stent, severe angulation | NA | Long stented segment, right coronary location, metal overlap | Total stent length, change in angulation of the lesion after stent implantation, right coronary location | Right coronary location | NA |
Overall, the mean incidence per patient of stent fracture was 4.0% (95% CI 0.4% to 16.3%). Although percentage patient follow-up was not similar among individual studies, the incidence of stent fractures was not found to be significantly affected by the percentage follow-up.
Among the total number of stent fractures, 30.4% (95% CI 22.2% to 39.5%) were estimated to occur in the left anterior descending coronary artery, 10.9% (95% CI 6.0% to 17.7%) in the left circumflex coronary artery, 56.4% (95% CI 46.9% to 65.5%) in the right coronary artery, <0.01% (95% CI 0.0% to 0.9%) in the left main coronary artery, and 1.7% (95% CI 0.3% to 5.2%) in SVGs. The probability of stent fracture was significantly lower in left main lesions compared to all other vessel lesions (p <0.01 for all 4 comparisons), and left anterior descending and left circumflex lesions had significantly lower fracture probabilities than right coronary lesions (p <0.0001 for both comparisons). Lesions with overlapping stents had a higher probability of stent fracture than those without overlapping stents. The probability of fracture for overlapping stents was 7.5% (95% CI 0.8% to 28%) versus 2.1% (95% CI 0.2% to 9%) (p = 0.0002). Fractured stents (average length 46 mm, 95% CI 38 to 54) were longer on average (p <0.01) than nonfractured stents (average length 32.5 mm, 95% CI 24.4 to 40.6).
Lesions with stent fractures had a higher (38% vs 8.2%) rate of ISR (p <0.0001; Table 2 ) as well as a higher rate (17% vs 5.6%) of TLR (p <0.0001; Table 3 ). Similarly, the probability of stent fractures was higher in patients with ISR and TLR.
Parameter | Estimate | 95% CI | p Value |
---|---|---|---|
P(stent fracture, given ISR) | 12.8% | 2.0%–49.4% | |
P(stent fracture, given no ISR) | 2.1% | 0.3%–12.4% | |
Difference in probabilities | 10.6% | 1.7%–36.9% | <0.0001 |
P(ISR, given stent fracture) | 38.0% | 29.2%–47.4% | |
P(ISR, given no stent fracture) | 8.2% | 7.0%–9.6% | |
Difference in probabilities | 29.7% | 20.8%–39.2% | <0.0001 |