Objective Systematic investigation of restenosis after percutaneous coronary intervention (PCI) with bare metal stents (BMS) or first or second generation drug eluting stents (DES) in large scale, broadly inclusive patient populations undergoing follow-up angiography represents a gap in our scientific knowledge. We investigated the incidence of angiographically proven restenosis and its predictors in patients undergoing PCI with stents.
Methods All patients undergoing successful implantation of coronary stents for de novo lesions from 1998 to 2009 and follow-up angiography at 6–8 months at two centres in Munich, Germany were eligible for inclusion. Patients with cardiogenic shock, dialysis dependent renal insufficiency or previous cardiac transplantation were excluded. Data were prospectively collected. The incidence of restenosis, defined as diameter stenosis ≥50% in the in-segment area at follow-up angiography, and its predictors were evaluated.
Results A total of 12 094 patients met inclusion criteria. Angiographic follow-up was available for 10 004 patients (77.5%) with 15 004 treated lesions. Binary restenosis was detected in 2643 (26.4%) patients. Use of first generation DES versus BMS (OR 0.35, 95% CI 0.31 to 0.39) and second generation DES versus first generation DES (OR 0.67, 95% CI 0.58 to 0.77) were independent predictors of lower rates of restenosis. At multivariate analysis, smaller vessel size (OR 1.59, 95% CI 1.52 to 1.68, for each 0.5 mm decrease), total stented length (OR 1.27, 95% CI 1.21 to 1.33, for each 10 mm increase), complex lesion morphology (OR 1.35, 95% 1.21 to 1.51), presence of diabetes mellitus (OR 1.32, 95% 1.19 to 1.46), and history of bypass surgery (OR 1.38, 95% CI 1.20 to 1.58) were independently associated with restenosis and were similar across the spectrum of stent devices.
Conclusions In this large cohort of patients with angiographic surveillance we demonstrated the impact of device development on antirestenotic efficacy, with sequentially improved efficacy from BMS to first generation DES to second generation DES. Predictors of restenosis were small vessel size, increased stented length, complex lesion morphology, diabetes mellitus, and prior bypass surgery.
- INTERVENTIONAL CARDIOLOGY
Statistics from Altmetric.com
In patients undergoing percutaneous coronary intervention (PCI), the use of drug eluting stents (DES) has greatly reduced the need for reintervention when compared with the use of bare metal stents (BMS).1–4 Nevertheless, in-stent restenosis continues to remain the principal reason for treatment failure after contemporary coronary stenting.5 ,6 Moreover, the uptake of percutaneous intervention in larger numbers of patients with increasingly complex lesion characteristics and disease comorbidities means that the number of patients presenting with restenosis remains considerable in absolute terms.5 ,6
The use of angiographic surveillance to detect lumen re-narrowing after coronary stenting is an important tool in the assessment of device efficacy and an integral part of many comparative efficacy studies.7–10 As technology evolves, the identification of clinical, angiographic and procedural factors predicting the risk of restenosis plays a key role in the refinement of interventional techniques, in the assessment of individual patient benefit and risk, and the choice of revascularisation strategies. Previous attempts to identify correlates of restenosis were based on observations from relatively modestly sized patient cohorts.11–14 This limited the generalisability of these findings. In addition, the impact of second generation DES—which now represent the dominant devices in clinical use—on predictors of restenosis remains largely unstudied.
Against this background, we sought to investigate the incidence of angiographically proven restenosis and its predictors in a large, broadly inclusive population undergoing PCI with BMS or first or second generation DES.
We analysed clinical, angiographic and procedural data of patients with coronary artery disease receiving a coronary stent for de novo lesions between January 1998 and December 2009 in two tertiary referral centres in Munich, Germany (Deutsches Herzzentrum and 1. Medizinische Klinik, Klinikum rechts der Isar). All patients undergoing angiographic surveillance at 6–8 months after successful intervention were eligible for this study. A routine follow-up angiography at this time point after revascularisation is the standard clinical practice at these centres. Patients with cardiogenic shock, chronic renal replacement therapy or previous cardiac transplantation were excluded. BMS were the sole types of stents approved for use from January 1998 to August 2002. Thereafter, DES became available. First generation DES comprised sirolimus eluting stents (polymer based, Cypher—Cordis, Warren, New Jersey, USA; polymer-free, Yukon—Translumina GmbH, Hechingen, Germany), polymer based paclitaxel eluting stents (Taxus—Boston Scientific, Natick, Massachusetts, USA) and phosphorycholine polymer based zotarolimus eluting stents (Endeavor—Medtronic Inc, Santa Rosa, California, USA). Second generation DES were available from January 2006 and included polymer based everolimus eluting stents (Xience V—Abbott Vascular, Santa Clara, California, USA), biolynx polymer based zotarolimus eluting stents (Resolute—Medtronic Inc, Santa Rosa, California, USA), biodegradable polymer based sirolimus eluting stents (Yukon—Translumina GmbH, Hechingen, Germany), and sirolimus and probucol eluting stents (Yukon—Translumina GmbH, Hechingen, Germany).
Some of the patients received BMS in the setting of the early controlled trials comparing DES versus BMS. In patients requiring interventions in multiple lesions in the same setting, typically the same stent was implanted. Antithrombotic and anticoagulant therapies reflected the changing practices during the period of observation. All patients received aspirin at the time of index intervention as well as a loading dose of platelet adenosine diphosphate (ADP) receptor inhibitors. At discharge, aspirin therapy was recommended indefinitely, while platelet ADP receptor inhibitors were prescribed for a period of time ranging from 1 to 24 months, depending on clinical presentation or type of stent implanted. Administration of glycoprotein IIb/IIIa inhibitors was undertaken according to trial protocol or at the operators’ discretion. Anticoagulation during coronary interventions was accomplished through administration of either unfractionated heparin or bivalirudin in all patients. All study subjects received standard cardioactive therapies as required (eg, β-blockers, statins, ACE inhibitors, and other drugs).
Coronary angiography evaluation and definitions
Baseline, post-procedural and follow-up coronary angiograms were digitally recorded and assessed off-line in the quantitative angiographic core laboratory (ISARESEARCH Center, Munich, Germany) with an automated edge detection system (Medis Medical Imaging Systems) by independent experienced operators. Coronary lesions were defined as independent lesions when they were situated in two different coronary segments. Qualitative morphological lesion characteristics were described in accordance with standard definitions.15 ,16 Measurements were performed on angiograms recorded after the intracoronary administration of glyceryl trinitrate (nitroglycerine) using a single worst-view projection at all times. The contrast-filled, non-tapered catheter tip was used for calibration. The angiographic and procedural parameters obtained were vessel size, lesion length, minimal lumen diameter, initial diameter stenosis, maximal balloon diameter (using actual measurement of maximal balloon size), maximal balloon pressure, length of the stented segment, final lumen diameter, and final diameter stenosis. Balloon-to-vessel ratio was calculated as the maximal diameter of the inflated balloon divided by the coronary vessel size. Percentage diameter stenosis was calculated as [1-(minimal lumen diameter/reference vessel diameter)/100]. Restenosis (angiographic or binary) was defined as diameter stenosis ≥50% in the in-segment area (including the stent area and 5 mm segments proximal and distal to the stent edges). Successful stent implantation was achieved if residual stenosis after intervention was <30% with TIMI (thrombolysis in myocardial infarction) flow grade 3.
Categorical data are presented as counts and proportions (%). Continuous data are presented as median and IQR (25th; 75th centiles) or as mean±SD, as appropriate. Data distribution was tested for normality using the Kolmogorov-Smirnov test. For patient level data the differences between groups were checked for significance using the Student t or Kruskal-Wallis tests (continuous data) or the χ2 or Fisher exact tests where the expected cell value was <5 (categorical variables). For lesion level data, the differences between groups were checked for statistical significance using generalised estimating equations for non-normally distributed data in order to address intra-patient correlation in patients who underwent multi-lesion interventions.17 We included in a multivariate regression analysis all clinical, angiographic and procedural features reporting a value of p<0.05 at univariate analysis. Then, separate multivariate analyses addressed predictors of restenosis in the cohort of patients receiving BMS, first generation DES or second generation DES, with assessment of interaction between variables included and the treatment with first or second generation DES (two-tailed value of p<0.05 was significant). The adjusted ORs with 95% CI were used as summary statistics and were derived from generalised estimating equation models.17 The statistical software package R V.2.15.1 (R Foundation for Statistical Computing, Vienna, Austria) was used for analyses. The R package gee was used to perform multivariate analyses.18
Among all patients undergoing coronary stent implantation during the study period, 12 904 met the enrolment criteria. Of these, 10 004 patients (77.5%) with 15 004 lesions underwent angiographic surveillance at 6–8 month follow-up: 4649 patients were treated with BMS (6521 lesions), while 5355 were treated with DES (8483 lesions). Among those not receiving 6–8 month follow-up angiography, 379 patients (13.1%) had died. Patients without angiographic surveillance were older (mean age 68.9±10.9 vs 65.4±12.3 years, p<0.001), with a higher proportion of diabetics (30.8% vs 24.0%, p<0.001) and with a similar proportion of DES implanted (54.3% vs 53.5%, p=0.47) in comparison to patients with invasive surveillance. Follow-up angiography was performed at a median of 198.7 days (182.1; 219.7) after the index procedure. The study diagram flow is shown in figure 1.
Figure 2 shows the utilisation of stent type and the incidence of restenosis over the duration of the study period. The higher the proportion of DES used, the lower the proportion of lesions with restenosis at angiographic surveillance; overall the proportion of lesions with restenosis was 30.1%, 14.6%, and 12.2% in patients treated with BMS, first generation DES, and second generation DES, respectively.
Angiographic restenosis was detected in 2643 (26.4%) patients and 3098 (20.6%) lesions.
Baseline patient characteristics of patients with and without restenosis are shown in table 1. Baseline angiographic and procedural characteristics of lesions with and without restenosis are shown in table 2.
Female gender, presence of diabetes mellitus, history of bypass surgery, clinical presentation with myocardial infarction, and multivessel disease were more likely to be present in patients with restenosis. In addition treatment for lesions in the left circumflex artery, lesions with complex morphology, chronic occlusions, long lesions, lesions in small vessels, and high grade baseline stenosis were more likely in patients with restenosis. Treatment for left main stem disease was less likely in patients with restenosis.
The model included all the characteristics with a value of p<0.05 at univariate analysis. Figure 3 shows the complete set of variables included in the multivariate analysis. The presence of diabetes mellitus, history of bypass surgery, left main stenting, complex lesion morphology, small vessel size, high grade baseline stenosis, and longer stent length were independently associated with higher likelihood of restenosis. Higher balloon-to-vessel ratio, treatment with first generation DES versus BMS, and treatment with second generation DES versus first generation DES were independently associated with lower likelihood of restenosis. Figure 4 shows the proportion of lesions with restenosis at follow-up according to stent treatment type.
Bare metal and drug eluting stent cohorts
The online supplementary table provides the rates of restenosis within the first and second generation DES cohorts. Subgroup analyses demonstrated that patients treated with BMS, first generation DES or second generation DES were associated with similar predictors of restenosis (see online supplementary figure).
In the current analysis we reported the incidence and predictors of angiographic restenosis in more than 10 000 patients with coronary artery disease treated with stent implantation at two centres over a 12 year period. The dataset analysed represents the largest population of patients with angiographic surveillance since the introduction of PCI. The salient findings are: (1) intervention with first generation DES is an independent predictor of lower rates of restenosis as compared with BMS; (2) intervention with second generation DES predicts less restenosis in comparison with first generation DES or second generation DES: small vessel size, longer stented length, complex lesion morphology, presence of diabetes mellitus, and history of bypass surgery are the most important predictors of angiographic restenosis.
Angiographic restenosis is usually defined as a binary measure being adjudicated in the presence of ≥50% lumen diameter stenosis at follow-up angiography, typically 6–9 months after intervention.19 ,20 Although the cut-off percentage diameter stenosis may be considered arbitrary, a significant body of evidence supports its robustness in the assessment of the efficacy of coronary stents.8–10 ,21 The introduction of DES therapy represented an important milestone in coronary intervention, significantly lowering both clinical and angiographic restenosis as compared with bare metal stenting.1–4 Indeed, the improved clinical outcomes seen with DES facilitated the expansion of percutaneous intervention to increasingly complex patient and lesion populations.22 However, as rates of restenosis increase with disease complexity, the absolute number of patients presenting with restenosis remains considerable and the identification of predictive factors is an issue of broad clinical importance. Moreover, although second generation DES now represent the dominant devices in clinical use, the impact of second generation DES on predictors of restenosis remains unclear.
A number of previous studies, based on relatively small patient cohorts, investigated the determinants of angiographic restenosis in patients undergoing coronary stenting. Following BMS implantation, the strongest predictors of restenosis were small vessel size, long lesion and long stented length, and presence of diabetes mellitus.23–25 In the DES era, the strongest predictors of restenosis were found to be small vessel size, long stented segments, final percentage of diameter stenosis, and the different antirestenotic potency of DES.12–14 Some studies also identified diabetes mellitus, in-stent restenosis and ostial lesion location as important predictive factors.12–14
The present study assessed predictors of restenosis in the largest population undergoing coronary stenting and follow-up angiography since the introduction of PCI. We studied a broadly inclusive patient population treated with BMS, first generation DES and second generation DES. In addition we focused on angiographically proven restenosis as this likely represents the most objective parameter of efficacy after stent implantation, without the risk of bias associated with adjudication of clinical restenosis.21
Our data confirm that across the spectrum of patients undergoing coronary stenting, small vessel size and long segment stenting continue to be strong predictors of restenosis, irrespective of the type of stent.26 Indeed, vessel size remains the factor most closely correlated with restenosis. Conceptually this is easy to understand: at an individual patient level, for any given degree of neointimal overgrowth after stenting the probability that the resultant diameter stenosis exceeds 50% is a straightforward function of vessel size and residual stenosis post-stenting.27 Although this association holds true regardless of stent type, by virtue of higher absolute suppression of neointimal hyperplasia, DES remain the most efficacious treatment modality for small vessel disease.28 Previous studies with BMS found the length of stented lesion strongly correlated with restenosis, although the relation between lesion length and restenosis after DES has been clear.26 The present dataset study underlines that total stented length remains a strong independent predictor of restenosis after both BMS and DES.
Left main stenting appears protective against restenosis in the univariate analysis, probably by virtue of a larger vessel diameter. After adjustment for other covariates (such as vessel size), left main stenting predicts higher restenosis risk, mainly due to lesion complexity (ie, bifurcation). Further investigation of the impact of left main stenting on restenosis is certainly warranted.
It is also notable that diabetes mellitus remains an important independent predictor of restenosis across the entire population enrolled in our study. Whether or not the high efficacy of DES attenuates or even eliminates the propensity of patients with diabetes to restenosis has remained a topic of some debate.6 ,12 ,29 Divergent definitions of diabetes30 and small sample sizes available for previous analyses 12 ,21 ,26 ,31–33 may have had an impact on the true effect of this comorbidity in patients undergoing coronary stenting.26 ,34 However, the present large scale analysis confirms the importance of diabetes as a risk factor for restenosis after coronary stenting. Nevertheless, it might be observed that in the subgroup of patients treated with first generation DES, the presence of diabetes mellitus did not predict restenosis risk. Although the value of this finding in the setting of multiple comparisons may be debateable, the observation is consistent with previous data.13 ,32 ,33
The current dataset also confirms that prior coronary bypass surgery remains a predictor of restenosis after stent implantation. This is in keeping with data from previous studies.13 The reason for this observation remains poorly defined. It is conceivable that this finding reflects the fact that these patients are likely to have more aggressive or advanced atherosclerotic disease, which confers reduced efficacy after percutaneous intervention.5 ,26
We find that predictors of binary restenosis are similar across the spectrum of stent devices. However, we also demonstrate that intervention with first generation DES is a strong independent predictor of lower rates of restenosis as compared with BMS, and that intervention with second generation DES is a predictor of less restenosis in comparison with first generation DES. This supports observations from large scale meta-analysis of randomised trials and confirms the impact of device development on antirestenotic efficacy with sequential improvement from BMS to first generation DES to second generation DES.35–38
The current analysis has some limitations. First, adjudication of restenosis was based on the percentage diameter of lumen renarrowing at surveillance angiography. As such, information on vessel wall morphology and restenotic tissue characterisation was not available. Secondly, angiographic follow-up was missing in 22.5% of patients. While these data are deemed ‘missing not at random’,39 it is a consistent feature of all large angiographic follow-up studies. Moreover the validity of such data has been shown to be high when rates of angiographic follow-up approach 80%40 and numerous studies have shown that angiographic surrogates are robust markers of device efficacy.9 ,10 ,21 These considerations notwithstanding, a possible selection bias, due to higher frequency of patients with recurrence of symptoms undergoing angiography, cannot be excluded. Thirdly, this study lacks a randomised design with stent selection based only on the period specific availability of the devices. As such, the findings in relation to comparative efficacy of different stent generations should be interpreted with caution. Fourthly, the grouping of stents as bare metal, first generation and second generation efficacy is necessarily arbitrary and cannot fully capture differences between different BMS or DES platforms. Finally, when analysing the predictive variables among BMS, first generation and second generation DES subgroups, the impact of errors introduced by multiple testing should be considered.
In this large cohort of patients with angiographic surveillance, we demonstrated the impact of device development on antirestenotic efficacy, with sequentially improved efficacy from BMS to first generation DES to second generation DES. Predictors of restenosis were small vessel size, increased stented length, complex lesion morphology, diabetes mellitus, and prior bypass surgery and remained similar across the spectrum of stent devices implanted.
What is already known about this subject
The rates and predictors of restenosis after percutaneous coronary intervention (PCI) with bare metal stents (BMS) or first or second generation drug eluting stents (DES) in large scale, broadly inclusive patient populations undergoing follow-up angiography represent a gap in our scientific knowledge.
What does this study add
This is by far the largest experience with systematic assessment of restenosis after PCI in the literature worldwide. We demonstrate that PCI with first generation DES is a strong independent predictor of lower rates of restenosis as compared with BMS, and that less restenosis is predicted following intervention with second generation DES compared with first generation DES. Predictors of restenosis were small vessel size, increased stented length, complex lesion morphology, diabetes mellitus, and prior bypass surgery, and remained similar across the spectrum of stent devices.
How might this impact on clinical practice
The improved clinical outcomes seen with DES facilitated the expansion of PCI to increasingly complex patient and lesion populations. However, as rates of restenosis increase with disease complexity, the absolute number of patients presenting with restenosis remains considerable and the identification of predictive factors is an issue of broad clinical importance. This study confirms the impact of device development on antirestenotic efficacy with sequential improvement from BMS to first generation DES to second generation DES. Predictors of restenosis remain similar across the spectrum of stent devices.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
SC and RAB contributed equally.
Contributors SC and RAB conceived and designed the study and wrote the paper. AK supervised the conduct of the trial and data collection. SC and AK provided statistical advice on study design and analysed the data. TT, SP, MJ, TI, LAK, MF, K-LL, contributed substantially to its revision. SC, RAB and AK take responsibility for the paper as a whole. SC, RAB, K-LL and AK had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All the authors approved the manuscript.
Competing interests AK has submitted patents in relation to a number of DES technologies and received consulting or lecture fees from Abbott, Biotronik, and Biosensors. The other authors declare no potential conflict of interest.
Ethics approval Routine clinical practice in patients undergoing coronary stenting at our institutions.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.