Risk of Noncardiac Surgery and Adverse Events After Stents
We performed a matched retrospective cohort study to assess the incremental risk of surgery on postoperative adverse cardiac events following coronary stent implantation. Patients with a coronary stent implanted at Veterans Affairs (VA) medical centers between October 1, 1999, and September 30, 2009, were identified in the VA Patient Treatment Files by International Classification of Diseases-Ninth Edition (ICD-9) procedure codes of 36.06 for bare-metal stent (BMS) or 36.07 for drug-eluting stent (DES). Following the identification of the study cohort, information on noncardiac procedures in the 24 months following coronary stent implantation was obtained from the VA National Surgery Office and the Centers for Medicaid & Medicare Services. Noncardiac surgery was defined by CPT codes, and detailed information on the construction of the study cohort and study variables has been previously published. Admission status was defined as elective if a patient was admitted to the hospital from home, and all other admission sources were considered nonelective.
We compared all patients with coronary stents undergoing noncardiac surgery in the 24 months following stent implantation to patients with coronary stents not undergoing subsequent surgery. Each patient undergoing surgery was matched to two patients who did not undergo surgery. Matching was on the basis of patient age, race, stent type (BMS or DES), year of stent placement, each of the six variables included in the revised Cardiac Risk Index (RCRI), and those identified in our previous analyses as significant predictors of major adverse cardiac events, including myocardial infarction (MI) 6 months prior to stent implantation and peripheral vascular disease. Patients in the nonsurgical cohort were required to be alive during the time interval following stent placement when their matched counterpart underwent surgery. That is, if a surgical patient had surgery 200 days following stent placement, their two matched nonsurgical patients had to be alive for at least 200 days following their stent placements.
Our primary outcome was a composite endpoint of acute MI and/or coronary revascularization by percutaneous coronary intervention or coronary artery bypass grafting within 30 days following surgery in the surgical cohort or the equivalent post-stent time period for the nonsurgical cohort. We did not include all-cause mortality in our composite endpoint as the cohorts were not matched on probability of death at the time of stent placement. We excluded patients from the study who underwent surgery within 2 weeks following coronary stent placement as it was difficult to attribute an MI to surgery versus an MI associated with receiving a coronary stent on the basis of administrative data.
Assessment of the matched cohort. The matched cohort was compared by univariate and bivariate frequencies to describe patient characteristics and the composite outcome of adverse cardiac events. Bivariate frequencies were compared using chi-square tests and continuous variables were compared using Wilcoxon rank-sum tests. To ensure effective matching on cardiac event outcomes, we calculated the cumulative risk of the composite cardiac endpoint using Kaplan-Meier survival curves comparing time to first outcome by cohort and by stent type.
Assessment of the Incremental Risk of Noncardiac Surgery. To determine the incremental risk of surgery on adverse cardiac events, we estimated the risk difference for adverse events in the 30 days following procedure by calculating the rate for the nonsurgical patients during the same 30-day time interval following stent as their matched surgery. For example, in a nonsurgical cohort patient matched to a surgical cohort patient at 370 days post-stent, we only examined adverse cardiac events during the time period of 370 to 400 days post-stent. Unadjusted risk differences and 95% confidence intervals (CI) for all outcomes were determined both overall and stratified by time since coronary stent placement using binomial regression models and incorporating generalized estimating equations (GEE).
Contribution of Time Post-stent to Incremental Risk of Surgery. We sought to determine how surgical risk for adverse cardiac events changes over time since stent placement. Examination of unadjusted plots strongly suggested that the relationship between cardiac risk and time from stent is nonlinear. To fully understand this, we considered time as a continuous variable in the main analyses. Additional analyses considered timing intervals to simplify clinical interpretation.
To plot changes in risk difference across time as a continuous variable, adjusted risk differences for adverse cardiac events were first determined by calculating a baseline risk of events across time since stent using a generalized additive model. Covariates in the baseline risk model included history of ischemic heart disease, congestive heart failure, cerebral vascular disease, chronic kidney disease, and insulin-dependent diabetes mellitus. Next, the predicted probability of adverse cardiac events for each surgical patient was calculated using a second generalized additive model that adjusted for surgery type, surgery admission status, and work relative value unit as a measure of surgical complexity. The final adjusted risk difference for each surgical case was determined by taking the difference between the baseline risk of the matched nonsurgical patient and the predicted probability of adverse cardiac events following surgery for the surgical patient.
To test whether specific factors contributed to the reduction in risk by time intervals, we used an interaction term within the GEE model to determine significant changes in risk difference by stent type, procedure type, surgical complexity, and case status across time of coronary stent placement (<6 weeks, 6 weeks to 6 months, and >6 months). Bivariate frequencies and GEE models were completed using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina). Adjusted risk differences were calculated and plotted using R packages MGCV and GGPLOT.
Methods
Study Design and Data Sources
We performed a matched retrospective cohort study to assess the incremental risk of surgery on postoperative adverse cardiac events following coronary stent implantation. Patients with a coronary stent implanted at Veterans Affairs (VA) medical centers between October 1, 1999, and September 30, 2009, were identified in the VA Patient Treatment Files by International Classification of Diseases-Ninth Edition (ICD-9) procedure codes of 36.06 for bare-metal stent (BMS) or 36.07 for drug-eluting stent (DES). Following the identification of the study cohort, information on noncardiac procedures in the 24 months following coronary stent implantation was obtained from the VA National Surgery Office and the Centers for Medicaid & Medicare Services. Noncardiac surgery was defined by CPT codes, and detailed information on the construction of the study cohort and study variables has been previously published. Admission status was defined as elective if a patient was admitted to the hospital from home, and all other admission sources were considered nonelective.
Study Population
We compared all patients with coronary stents undergoing noncardiac surgery in the 24 months following stent implantation to patients with coronary stents not undergoing subsequent surgery. Each patient undergoing surgery was matched to two patients who did not undergo surgery. Matching was on the basis of patient age, race, stent type (BMS or DES), year of stent placement, each of the six variables included in the revised Cardiac Risk Index (RCRI), and those identified in our previous analyses as significant predictors of major adverse cardiac events, including myocardial infarction (MI) 6 months prior to stent implantation and peripheral vascular disease. Patients in the nonsurgical cohort were required to be alive during the time interval following stent placement when their matched counterpart underwent surgery. That is, if a surgical patient had surgery 200 days following stent placement, their two matched nonsurgical patients had to be alive for at least 200 days following their stent placements.
Outcome Variables
Our primary outcome was a composite endpoint of acute MI and/or coronary revascularization by percutaneous coronary intervention or coronary artery bypass grafting within 30 days following surgery in the surgical cohort or the equivalent post-stent time period for the nonsurgical cohort. We did not include all-cause mortality in our composite endpoint as the cohorts were not matched on probability of death at the time of stent placement. We excluded patients from the study who underwent surgery within 2 weeks following coronary stent placement as it was difficult to attribute an MI to surgery versus an MI associated with receiving a coronary stent on the basis of administrative data.
Analytical Methods
Assessment of the matched cohort. The matched cohort was compared by univariate and bivariate frequencies to describe patient characteristics and the composite outcome of adverse cardiac events. Bivariate frequencies were compared using chi-square tests and continuous variables were compared using Wilcoxon rank-sum tests. To ensure effective matching on cardiac event outcomes, we calculated the cumulative risk of the composite cardiac endpoint using Kaplan-Meier survival curves comparing time to first outcome by cohort and by stent type.
Assessment of the Incremental Risk of Noncardiac Surgery. To determine the incremental risk of surgery on adverse cardiac events, we estimated the risk difference for adverse events in the 30 days following procedure by calculating the rate for the nonsurgical patients during the same 30-day time interval following stent as their matched surgery. For example, in a nonsurgical cohort patient matched to a surgical cohort patient at 370 days post-stent, we only examined adverse cardiac events during the time period of 370 to 400 days post-stent. Unadjusted risk differences and 95% confidence intervals (CI) for all outcomes were determined both overall and stratified by time since coronary stent placement using binomial regression models and incorporating generalized estimating equations (GEE).
Contribution of Time Post-stent to Incremental Risk of Surgery. We sought to determine how surgical risk for adverse cardiac events changes over time since stent placement. Examination of unadjusted plots strongly suggested that the relationship between cardiac risk and time from stent is nonlinear. To fully understand this, we considered time as a continuous variable in the main analyses. Additional analyses considered timing intervals to simplify clinical interpretation.
To plot changes in risk difference across time as a continuous variable, adjusted risk differences for adverse cardiac events were first determined by calculating a baseline risk of events across time since stent using a generalized additive model. Covariates in the baseline risk model included history of ischemic heart disease, congestive heart failure, cerebral vascular disease, chronic kidney disease, and insulin-dependent diabetes mellitus. Next, the predicted probability of adverse cardiac events for each surgical patient was calculated using a second generalized additive model that adjusted for surgery type, surgery admission status, and work relative value unit as a measure of surgical complexity. The final adjusted risk difference for each surgical case was determined by taking the difference between the baseline risk of the matched nonsurgical patient and the predicted probability of adverse cardiac events following surgery for the surgical patient.
Changes in Incremental Risk of Surgery by Time Intervals
To test whether specific factors contributed to the reduction in risk by time intervals, we used an interaction term within the GEE model to determine significant changes in risk difference by stent type, procedure type, surgical complexity, and case status across time of coronary stent placement (<6 weeks, 6 weeks to 6 months, and >6 months). Bivariate frequencies and GEE models were completed using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina). Adjusted risk differences were calculated and plotted using R packages MGCV and GGPLOT.