Author + information
- Received March 10, 2009
- Accepted March 26, 2009
- Published online June 1, 2009.
- Saif Anwaruddin, MD⁎,
- Arman T. Askari, MD⁎,
- Hammad Saudye, MD§,
- Lilian Batizy, MS‡,
- Penny L. Houghtaling, MS‡,
- Mohammad Alamoudi, MD∥,
- Michael Militello, PharmD†,
- Kamran Muhammad, MD⁎,
- Samir Kapadia, MD⁎ and
- Stephen G. Ellis, MD⁎,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Stephen G. Ellis, Director, Sones Cardiac Catheterization Laboratory, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, Ohio 44195
Objectives The aim of this study was to assess risk of inpatient surgery at any time after percutaneous coronary intervention (PCI) with drug-eluting stents (DES).
Background Risk of adverse events, including stent thrombosis (ST), in patients undergoing surgical procedures with prior DES remains poorly defined.
Methods Outcomes of consecutive patients having inpatient surgical procedures after PCI with DES, placed from April 28, 2003 until December 31, 2006 at a tertiary-care medical center, were studied. Primary and secondary end points were 30-day post-operative risk of the Academic Research Consortium (ARC) definite and modified probable definitions of ST and combined 30-day post-operative risk of death, nonfatal myocardial infarction (MI), or ST, respectively. Multivariable logistic regression analyses were used to determine independent risk factors.
Results Six hundred six inpatient surgeries on 481 patients with a mean time from PCI to surgery of 1.07 ± 0.89 years were evaluated. The primary and secondary end points occurred after 11 (2.0%) and 56 (9%) surgeries, respectively. Risk of the combined end point and ST decreased significantly in the first 1 to 6 months after PCI (p < 0.0001 and p < 0.014, respectively); however, risk persisted when time between PCI and surgery was >12 months. Independent correlates of the combined end point include emergency surgery, antecedent MI, the pre-operative use of intravenous heparin, and atherosclerotic lesion length treated with DES. Oral antiplatelet status at time of surgery was not a correlate of events.
Conclusions Risk of 30-day post-operative adverse events, including ST, remains significantly higher when surgery is performed soon after PCI, while intermediate-term risk extending at least 2 to 3 years remains important.
Coronary artery stenting has increased in the last several years with the advent of drug-eluting stents (DES) (1). The initial enthusiasm for DES due to reductions in restenosis has been tempered by concerns about stent thrombosis (ST), particularly late ST (2,3).
The risk of noncardiac surgery in weeks following percutaneous coronary intervention (PCI) with bare-metal stents (BMS) has been well-examined (4,5). The risk of DES thrombosis in the perioperative period is thought related to 3 main factors—cessation of dual antiplatelet therapy for surgery, a hypercoagulable state associated with surgery, and non-endothelialization of the stent (6). It is difficult, however, to judge actual risk associated with surgery after DES, because there have been no large-scale systematic reviews of the topic and existing reports to do not provide consistent findings (7–9). Given the currently limited data (10), we sought to better characterize perioperative risk associated with prior DES use.
This study protocol was approved by the Cleveland Clinic Institutional Review Board. All patients having undergone PCI at the Cleveland Clinic with DES from 2003 until the end of 2006 were identified by the Cleveland Clinic PCI database in which consecutive patients are enrolled. These patients were cross-matched with the Cleveland Clinic operating room database to identify those patients having undergone any inpatient surgical procedure at any time point after PCI. Patients were excluded if they had undergone percutaneous surgical procedures, experimental procedures, procedures performed outside the main campus, or heart transplantation (Fig. 1). Surgical risk was defined by a previously validated risk index (11).
Information regarding the resulting cohort of patients was obtained from an interrogation of the medical record. Information was gathered regarding baseline demographic data, past medical history, baseline laboratory values, and medications at time of surgery, including antiplatelet medication status before and at the time of surgery (Online Appendix 1: core variables). Creatinine clearance was calculated with the Cockroft-Gault equation.
Follow-up and clinical events
Follow-up was established by evidence of clinical visit at or after 30 days after the surgical procedure or by occurrence of a primary or secondary end point. A clinical visit to an affiliate hospital or satellite clinic was deemed acceptable for follow-up as electronic medical records were available for review. Death was established either by the Social Security Death Index (last updated on February 2, 2008) and/or by review of the paper and electronic medical record.
Events such as myocardial infarction (MI) or ST were adjudicated (S.A.) on the basis of available clinical data. Diagnosis of perioperative MI was established with biomarkers, electrocardiographic evidence of ischemia or infarction, and clinical course as previously published (12). “Definite” ST events were classified on the basis of the Academic Research Consortium (ARC) definition (13). “Probable” ST was a modified version of the ARC definition. The modification involved excluding any unexplained death within 30 days of PCI as a probable ST and requiring evidence of an MI with evidence of ischemia in the territory of the prior stent without another obvious cause within the 30-day post-operative period. The rationale for modification of the definition was to avoid arbitrary assignment of ST as a cause of death in patients having undergone surgery within 30 days of PCI.
The primary end point for this study was prospectively defined as the 30-day post-operative incidence of ST (as in the preceding), and the secondary end point was defined as the 30-day post-operative combined incidence of ST, nonfatal MI (non-stent thrombosis), or death.
Continuous data are expressed as mean ± SD or with median and 15th and 85th percentile values (comparable to 1 SD) as appropriate. For continuous variables, Wilcoxon rank sum tests were used to analyze 2-level group differences and Kruskal-Wallis tests were used to analyze 3-level group differences. Categorical data are displayed as frequencies and percentages, and comparisons are made with chi-square tests (Fisher exact tests if appropriate).
Multivariable logistic regression analyses were used to determine independent risk factors for the combined end point. A stepwise selection approach was used with criteria of p ≤ 0.05 for retention of variables to form an initial model. Bootstrap bagging methods were then used to determine important reliable predictors with random resampling and automated analysis of 1,000 datasets for each end point. Covariates entered into ≥50% of the bootstrapping runs were retained in the final parsimonious model.
The unit of observation for this study is surgery following DES. To allow for at least a 15-day uninterrupted follow-up after index surgery, for patients with more than 1 temporally related surgery, the index surgery was taken as the last surgery in the sequence for the primary analysis. Generalized estimating equations (GEE) (14) were developed to account for within-patient correlation of repeated measurements (multiple surgeries/patient). Final parameter estimates are the results of GEE-adjusted models.
Transformations of scale for continuous variables were used to meet model assumptions and for proper calibration of predictive variables with outcome. Univariable trend plots were used to display covariate relationships with outcomes. Statistical analyses were performed with SAS version 9.1 (SAS Institute, Cary, North Carolina).
Five hundred forty-one patients underwent 752 inpatient surgeries from April 28, 2003 to December 31, 2006 at the Cleveland Clinic at any time after PCI with DES. Eighty-nine surgeries from 60 patients were excluded due to missing core variables including perioperative medication/antiplatelet use at or before the time of surgery, and 57 surgeries were not used in the analysis, because they occurred in the 15 days before the index surgery. Results from these patients were analyzed separately. Therefore, 481 patients who underwent 606 surgeries comprise the principal study cohort. Types of surgery are shown in Table 1. Results of 57 surgeries not included in the analysis, due to temporal proximity to the index procedure, are shown in Table 2. Mean time from PCI to surgery was 1.07 ± 0.89 years with a median of 0.88 years (15th and 85th percentiles of 70 days and 2 years, respectively).
Patients who had surgery within 1 month of PCI were different in many ways compared with those who had surgery at a later time point. Patients were more frequently taking aspirin or clopidogrel within 24 h of surgery, less often taking statins, and less likely to be hypertensive. Furthermore, duration of time (days) not taking either aspirin or clopidogrel before surgery was significantly shorter when time from PCI to surgery was <1 month (Table 3).
There were 381 surgeries (63%) performed on patients taking neither aspirin or clopidogrel, 126 surgeries (21%) performed on patients taking both oral antiplatelet agents, 86 surgeries (15%) performed on patients taking only aspirin, and 6 (1%) surgeries performed on patients taking only clopidogrel.
There were 11 ST events (2.0%, 95% confidence limits: 0.8% to 3.3%; 4 definite, 7 probable) during the 30-day post-operative period. They occurred at a median of 2 days after surgery. One probable ST event occurred in the cohort of the 57 surgeries (1.8%) that was excluded from the primary analysis due to surgery within 15 days before the index surgery. Clinical descriptions of those patients having an ST event are shown in Table 4. There was a higher incidence of ST when surgery was performed within 1 month after PCI as compared with surgery performed after 30 days after PCI (p = 0.04) (Fig. 2). Over time, however, despite a reduction in the incidence of ST, risk seems to persist (Fig. 3A).
Univariate correlates for the 30-day risk of ST include: a lower pre-operative ejection fraction (p = 0.004), use of IV heparin in the pre-operative period (up until the time of surgery) (p = 0.006), absence of hypertension (p = 0.02), use of aspirin within 24 h of surgery (p = 0.03), and a lower mean creatinine clearance (p = 0.05). A multivariable model was not constructed for the end point of ST, due to the small absolute number of events.
The 30-day post-operative combined end point occurred after 56 (9%, 95% confidence interval: 7% to 12%) surgeries in this cohort. There were 32 (5.0%) deaths and 23 (4.1%) nonfatal MIs (non-ST) within the 30-day post-operative period. Surgeries that were performed within 1 month of PCI were associated with a much higher risk (p = 0.0001) (Fig. 2). As with ST, despite a reduction in risk over time, risk persisted even when surgery was performed as far out as 2 to 3 years after PCI (Fig. 3B).
Emergency surgery and high-risk surgeries were significant univariate predictors of the combined end point (p < 0.001), as was use of clopidogrel (p = 0.002) or aspirin (p = 0.008) in the 24 h before surgery. These variables were often interrelated. Other univariate correlates of the combined end point included lower pre-operative creatinine clearance (p = 0.01), lower pre-operative ejection fraction (p = 0.001), lower pre-operative hemoglobin (p = 0.001), and elevated pre-operative white blood cell count (p = 0.0002). Furthermore, a shorter interval (days) of not taking aspirin before surgery was also noted to be a significant univariate correlate (p = 0.043).
Independent correlates for the 30-day post-operative combined end point include MI in the 30 days before surgery (p = 0.0004), use of pre-operative IV heparin (p = 0.0004), emergency surgery (p = 0.004), and length of the atherosclerotic lesion where DES was placed (p = 0.02) (Table 5). Once these variables were considered, use of aspirin or clopidogrel in the 24 h before surgery was not a significant predictor of the combined end point (p = 0.4).
A separate analysis of the 30-day post-operative combined end point was performed, having excluded coronary artery bypass grafting surgeries (n = 483). The incidences of ST and the combined primary end point were 2.1% and 9.3%, respectively, and were not significantly different from the overall population (p > 0.4). Independent correlates for the 30-day post-operative combined end point included MI in the 30 days before surgery (p < 0.0001), use of pre-operative IV heparin (p = 0.0004), and higher surgical risk (p = 0.003) (Table 6). Again, use of aspirin or clopidogrel in the 24 h before surgery was not a significant correlate of the combined end point (p = 0.8).
This study reports upon the heretofore incompletely described 30-day post-operative risk associated with prior DES use. Overall risk of ST in this cohort was 2.0%, with 4 definite and 7 probable adjudicated events. A similar risk of ST was noted when coronary artery bypass grafting surgeries were excluded from the analysis. Overall risk of the combined end point of death, nonfatal MI, or ST was 9.0%. For ST, risk seemed to be significantly higher when the time interval between PCI and surgery was shorter. Furthermore, although the risk of both the primary and secondary end points decreased for the increasing interval between PCI and surgery, the risk of an event seems to persist—with an approximately 1% risk of thrombosis occurring when the time interval between PCI and surgery was between 2 to 3 years.
Emergency surgery and MI in the 30 days before surgery were independent correlates of the 30-day post-operative combined end point. They serve to validate the model for the combined end point, because they have been shown to predict complications in noncardiac surgery (11,15,16). Longer atherosclerotic target lesion length was also an independent correlate of the 30-day post-operative combined end point. Longer atherosclerotic coronary lesions might be a surrogate for more extensive coronary artery disease burden as well as use of longer stents and/or overlapping/bifurcating stents, all of which might impact endothelialization of the DES. In this cohort, however, direct PCI-related characteristics such as stent length or diameter were not independent predictors of the combined end point.
The use of IV heparin in the pre-operative period also seems to be a robust independent predictor of the 30-day combined end point. A possible explanation for this is the well-described heparin rebound phenomenon (17). Intravenous unfractionated heparin also might cause platelet activation (18,19). The combination of these effects, in conjunction with the hypercoagulable state known to occur after surgery, could serve to explain the apparent correlation between heparin use and risk (20).
Premature cessation of oral antiplatelet therapy, particularly within 1 year after PCI with DES, has been identified as a factor associated with increased risk of ST (21). In the perioperative setting, however, continuation of oral antiplatelet therapy might not be considered feasible, due to increased risk of significant bleeding (22). In this analysis, surgery within 1 month of PCI was more likely to have been performed on patients taking aspirin or clopidogrel. Aspirin use in the perioperative period was also identified as a univariate predictor of both end points. Although this is likely due to timing of surgery after PCI, it might also reflect a potentially sicker patient population needing more urgent or emergent surgery shortly after PCI.
Attempting to dissect out the interplay and importance of these variables, the multivariable model for the 30-day post-operative combined end point demonstrates that taking aspirin and/or clopidogrel in the 24 h before surgery was not protective (p = 0.3), irrespective of interval between PCI and surgery. Although the reason for this remains uncertain, it suggests oral antiplatelet therapy in the perioperative period might not protect against adverse events, including ST. In this cohort, there were 3 ST events (Patients #7, #8, and #10 in Table 4) observed in surgeries performed on patients taking dual antiplatelet therapy (2 taking aspirin, 1 taking aspirin and clopidogrel) at an average of 323 days after PCI.
That IV heparin was a correlate of adverse outcomes and oral antiplatelet therapy was not apparently protective of such events suggests that these agents might be surrogates for clinical instability, surgical risk, or perceived risk of ST. For example, heparin was more often used before higher-risk surgeries (vascular, cardiovascular), and oral antiplatelet therapy was more likely to be continued when time from PCI to surgery was <1 month. Furthermore, emergent surgery often precluded any planned discontinuation of oral antiplatelet therapy. However, within the multivariable model, effects of IV heparin seemed to be independent of surgical risk and timing of surgery—even when higher-risk surgeries were removed from the analysis (data not shown). It might be that unmeasured variables, such as non-endothelialization of the stent, might be vitally important in determining risk—particularly in those undergoing surgery soon after PCI and while taking oral antiplatelet therapy.
The implications of these findings are important and widely applicable, particularly because surgery after PCI can be unpredictable. The incidence of a surgical procedure within 1 year after PCI with DES is approximately 4.4% (P. Berger, personal communications, May 2008).
Given widespread use of DES in PCI, the role of proper pre-operative risk stratification remains crucial in those with previous DES placement. The primary objective should be to delay elective surgery as far from the time of PCI as possible for risk reduction. In this cohort, the rate of ST was 7% when surgery was within 1 month of PCI, 2.9% when performed between 1 and 6 months after PCI, 1.4% between 6 and 12 months, and 0.78% when surgery was >12 months from PCI. One caveat is that delaying surgery does not eliminate the risk of adverse post-operative events, including ST. Furthermore, performing surgery with patients taking dual antiplatelet therapy might not confer protection from adverse post-operative events, including ST.
This study remains limited in that it is derived from a single-center, retrospective cohort. The retrospective nature of the study combined with the need to combine DES and surgical databases to derive the patient cohort might have resulted in an underestimation of true incidence of perioperative ST, because events occurring during the period of antiplatelet therapy withdrawal before surgery would have been missed. Furthermore, although the incidence of ST in this cohort was high, the absolute number of events remains low. This precluded any robust statistical analysis to further define the relationships among variables.
For purposes of the principal statistical analysis and to allow at least a 15-day follow-up without another surgery, we excluded all surgeries performed within 15 days of a subsequent “index” surgery. After the 57 surgeries that were excluded due to this logic employed to allow for a 30-day follow-up after any surgery, one probable ST (1.7%) occurred before the index surgery and was therefore censored from the principal analysis. This incidence is similar to that observed for ST events in those surgeries that were included in the analysis (p = 0.97). Furthermore, there were only a few differences in baseline characteristics between included surgeries and excluded surgeries (Table 2). Given the reason for exclusion of these patient surgeries, these differences are expected. As such, we do not believe that selective exclusion of these surgeries would have significantly altered the interpretation of patient outcomes.
We also acknowledge the limitations of our definition of ST and its application in the perioperative period. A modified version of the ARC definitions was used to exclude perioperative events unrelated to ST. Because the perioperative setting can complicate the adjudication of an accurate diagnosis, the secondary combined end point was also evaluated, but it is possible that the relationship between death or MI and ST was misclassified in some patients
Finally, we acknowledge the possibility that not all follow-up events were captured. Routine follow-up was not protocolized. However, given the large catchment area of the hospital and affiliate hospitals linked by electronic medical records, 87.2% of patients had documented follow-up within 30 days, and 85% had a post-procedural electrocardiogram. All patients had known vital status at 30 days. Furthermore, those patients who received both cardiac and noncardiac care at this institution were more likely to return for follow-up or complications related to either issue.
We present the experience from a retrospective cohort having undergone inpatient surgical procedures after prior PCI with DES. The 30-day post-operative risk of ST was 2.0%, and that of ST, nonfatal MI, or death was 9.0%. Risk seems to be highest when the interval from PCI to surgery is <1 to 2 months, but some appreciable risk endures over at least 2 to 3 years. Use of perioperative oral antiplatelet agents might not protect against and use of pre-operative IV heparin might increase the risk of events in the post-operative period. Risk seems to be higher with antecedent MI, longer atherosclerotic lesion length, or emergency surgery. These findings need to be validated in a large prospective data set but, pending this, assist the clinician in making decisions about surgeries in patients with DES.
The authors acknowledge the invaluable assistance of John Smetanko with regard to medical record/data acquisition.
For a supplementary Methods section, please see the online version of this article.
Dr. Askari has served as a speaker for Bristol-Myers Squibb and Sanofi Aventis. Dr. Ellis has served as a consultant, <$10K annually, for Abbott Vascular, Boston Scientific, and Cordis, and has received research funding from Centocor and Eli Lily & Co.
- Abbreviations and Acronyms
- Academic Research Consortium
- bare-metal stent(s)
- drug-eluting stent(s)
- myocardial infarction
- percutaneous coronary intervention
- stent thrombosis
- Received March 10, 2009.
- Accepted March 26, 2009.
- American College of Cardiology Foundation
- Ong A.T.,
- McFadden E.P.,
- Regar E.,
- de Jaegere P.P.,
- van Domburg R.T.,
- Serruys P.W.
- Wilson S.H.,
- Fasseas P.,
- Orford J.L.,
- et al.
- Kaluza G.L.,
- Joseph J.,
- Lee J.R.,
- Raizner M.E.,
- Raizner A.E.
- Riddell J.W.,
- Chiche L.,
- Plaud B.,
- Hamon M.
- Schouten O.,
- van Domburg R.T.,
- Bax J.J.,
- et al.
- Lee T.H.,
- Marcantonio E.R.,
- Mangione C.M.,
- et al.
- Thygesen K.,
- Alpert J.S.,
- White H.D.,
- et al.
- Cutlip D.E.,
- Windecker S.,
- Mehran R.,
- et al.
- Hanley J.A.,
- Negassa A.,
- Edwardes M.D.,
- Forrester J.E.
- Xiao Z.,
- Theroux P.
- Hodgson J.M.,
- Stone G.W.,
- Lincoff A.M.,
- et al.