ACADEMIC EMERGENCY MEDICINE 2011; 18:613–618 © 2011 by the Society for Academic Emergency Medicine
Objectives: Risk stratification of patients with potential acute coronary syndrome (ACS) is difficult. Patients with prior revascularization are considered higher risk, but they can also have symptoms from noncardiac causes. This study evaluated whether the presenting clinical characteristics were predictive of an increased risk of 30-day cardiovascular events in patients with prior revascularization presenting to the emergency department (ED) with symptoms of potential ACS.
Methods: This was a secondary analysis of the DISPO-ACS study, a 2000-patient, four-site, randomized controlled trial of patients presenting with potential ACS. Process outcomes were evaluated using point-of-care cardiac markers compared to standard laboratory-based markers. Data included demographics, history, presenting symptoms, laboratory and electrocardiogram (ECG) results, hospital course, and 30-day cardiovascular events (death, acute myocardial infarction [AMI], revascularization). The association between presenting characteristics and 30-day cardiovascular events was assessed using univariable analysis and logistic regression; odds ratios (ORs) with 95% confidence intervals (CIs) are given.
Results: Of 2,000 patients enrolled, 611 had prior revascularization (538 percutaneous coronary intervention [PCI], 232 coronary artery bypass graft [CABG], 159 both). The mean (±SD) age was 66 (±14) years, 44% were female, and 22% were black. By 30 days, 101 patients (17%) had cardiovascular events (81 during the index visit, 20 during follow-up). There were four deaths, 28 AMIs, and 67 revascularizations within 30 days; 20 patients had multiple endpoints. Being male (OR = 1.67, 95% CI = 1.07 to 2.62) or nonblack (OR = 1.95, 95% CI = 1.07 to 3.56) or having a family history of coronary artery disease (CAD; OR = 2.09, 95% CI = 1.32 to 3.3), elevated lipids (OR = 1.71, 95% CI = 1.04 to 2.82), prior AMI (OR = 1.79, 95% CI = 1.16 to 2.76), abnormal ECG on arrival (OR = 2.1, 95% CI = 1.33 to 3.34), and a positive initial troponin (OR = 14.7, 95% CI = 6.8 to 32.2) were predictive of cardiovascular events. The multivariable model found family history of CAD (OR = 2.06, 95% CI = 1.26 to 3.36), abnormal initial ECG (OR = 1.89, 95% CI = 1.16 to 3.09), and positive initial troponin (OR = 13.3, 95% CI = 5.9 to 29.6) remained predictive of 30-day cardiovascular events.
Conclusions: In patients with prior revascularization, the initial ECG and early cardiac marker elevations, but not clinical presentation, predict odds of 30-day death, AMI, or revascularization.
More than 6 million patients present annually to U.S. emergency departments (EDs) with chest pain and other symptoms suggestive of acute coronary syndrome (ACS) and myocardial ischemia.1–3 To decrease the number of missed cardiac diagnoses, physicians have adopted a conservative method of evaluating and diagnosing chest pain patients.4 Seventy percent of coronary care unit admissions are due to patients admitted to rule out myocardial infarction (MI).5 The number of patients undergoing revascularization annually is on the rise, and so it stands to reason that a growing number of patients presenting with symptoms suggestive of ACS have also previously undergone revascularization. These patients are traditionally considered to be at higher risk. While many risk stratification techniques have been developed and validated for use in ED patients with potential ACS,7–10 it is unknown whether these techniques are applicable to patients with prior revascularization, since it is unknown which risk factors best predict adverse cardiac events in these patients. Therefore, we explored the association between known risk factors and 30-day adverse cardiovascular events in patients with prior revascularization who presented to the ED and were evaluated for suspected ACS.
This study was a secondary analysis of the multicenter DISPO-ACS (Clinical Trial Number NCT00222352),11 which was a randomized controlled trial comparing process outcomes between patients evaluated for ACS using point-of-care cardiac marker testing (i-STAT cardiac troponin I assay, Abbott Point-of-Care, Abbott Laboratories, East Windsor, NJ) to those being evaluated using central laboratory cardiac marker testing. With the exception of where cardiac markers were assayed, evaluation and treatment was at the physician’s discretion. The current analysis evaluates patients with prior percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG) who presented to the ED with symptoms consistent with potential ACS, to determine predictors of 30-day adverse cardiovascular events. The study was approved by the institutional review boards of all participating sites. All patients provided informed consent.
Study Setting and Population
The multicenter study was conducted between December 2004 and November 2006 within four EDs (Jewish Hospital of Cincinnati, Cincinnati, OH; University of Pennsylvania, Philadelphia, PA; William Beaumont Hospital, Troy, MI; and Stanford University, Stanford, CA), with annual ED censuses ranging from 33,000 to 116,000 patients. The proportion of these ED visits with possible ACS ranged from 4% to 8%. During the study period, trained research personnel were present in the ED from 8 am to midnight 7 days per week at two sites and 9 am to 5 pm or 8 am to 11:30 pm 5 days per week at two sites.
Participants were required to be 21 years of age or older with symptoms suggestive of possible ACS who had cardiac biomarkers ordered by their physician. Patients were excluded if they had an electrocardiogram (ECG) diagnostic for ST-segment elevation myocardial infarction (STEMI) because they have already been identified as being high risk and meeting an outcome (i.e., acute myocardial infarction [AMI]) prior to enrollment. Patients were randomized to having cardiac markers measured in the central laboratory only or at the point of care and in the central laboratory. This analysis includes only the cohort who had a self-reported history of previous coronary artery revascularization.
Patient information was obtained at the time of presentation to the ED using a case report form that was designed to comply with the standardized reporting guidelines for ED chest pain risk stratification studies.12 Data included demographics, medications, initial vital signs, physical exam, characteristics of chest pain and associated symptoms, cardiac risk factors, prior cardiac testing, ECG interpretation, calculated Thrombosis in Myocardial Infarction (TIMI) risk score, and final ED diagnosis. Prior coronary revascularization procedures were defined as PCI and CABG. Admitted patients were followed during their hospital stay and any complications or interventions were recorded. Patient follow-up was obtained via telephone at 30 days after presentation. Patients or their proxies were questioned about the occurrence of death, MI, and revascularization.
Biomarkers. Because only 50% of participants in the DISPO-ACS study had a point-of-care cardiac biomarker, central laboratory assay results were used for this study. Sites used the DADE Dimension (Siemens, Deerfield, IL), Bayer Centaur platform (Siemens, Deerfield, IL), or Abbott AxSYM (Abbott Laboratories, Abbott Park, IL) for TnI assays and the Elecsys 2010 (Roche Diagnostics, Indianapolis, IN) for the TnT assays. The cutoff used for TnT was 0.1, while that for TnI was 0.3.
The primary outcome was the occurrence of a cardiovascular event, which included all-cause mortality, nonfatal AMI (STEMI and NSTEMI), and new revascularization within 30 days of presentation. As all patients presented with symptoms consistent with ACS, either ECG evolution of AMI or elevation in cardiac markers with a typical rise and fall were required for final diagnosis of AMI. This decision was made locally rather than through the use of a central adjudication. Outcomes included those occurring during the initial index hospitalization and during the 30-day follow-up period.11 STEMI present on admission was excluded, as the patient already met an easily identified outcome. New revascularization was defined as CABG or PCI with or without stent placement. Follow-up end points, including death, were determined by record review or communication with the patient or next of kin.
Continuous data are presented as either means with standard deviations (SD) or medians with ranges, based on the distribution of the data. Categorical data are presented as the frequency of occurrence and percentages. The association between presenting characteristics and 30-day outcomes was assessed using univariable and multivariable logistic regression; odds ratios (ORs) with 95% confidence intervals (CI) are given. The multivariable analysis used prespecified models for predicting the occurrence of a cardiovascular event. The variables were chosen based on clinical logic regarding variables that affect cardiovascular event rates. Variables entered in the prespecified model included age, race, sex, cardiac risk factors, the symptoms noted in Table 1, and ECG. ECGs were classified as shown in the ECG diagnostic categories in Table 1. These categories were then collapsed into normal or not normal. Normal was considered to include normal and nonspecific findings. Everything else was considered abnormal. In the final model, we explored colinearity and outliers. The correlation matrix indicated no correlation greater than 0.3. Excluding the top 10% most influential cases did not change the magnitude or direction of effects, although statistical significance decreased due to smaller sample sizes. The p-values all remained below 0.1.
|Demographics, Characteristics||Total (N = 611)||No Outcome (n = 497)||Outcome (n = 114)|
|Age (yr)||65.9 (±14.1)||65.6 (±14.1)||67.3 (±14.0)|
|Black||136 (22.3)||119 (23.9)||17 (14.9)|
|White||442 (72.3)||352 (70.8)||90 (78.9)|
|Other||33 (5.4)||26 (5.2)||7 (6.1)|
|Male||343 (56.1)||268 (53.9)||75 (65.8)|
|Female||268 (43.9)||229 (46.1)||39 (34.2)|
|Family history of CAD||335 (54.8)||257 (51.7)||78 (68.4)|
|Hypertension||475 (77.7)||388 (78.1)||87 (76.3)|
|Hyperlipidemia||417 (68.2)||327 (65.8)||90 (78.9)|
|Diabetes mellitus||195 (31.9)||153 (30.8)||42 (36.8)|
|Current or former smoker||352 (57.6)||287 (57.7)||65 (57.0)|
|Current or former cocaine use||12 (2.0)||12 (2.4)||0 (0.0)|
|Prior CABG||232 (38.0)||188 (37.8)||44 (38.6)|
|Prior MI||296 (48.4)||226 (45.5)||70 (61.4)|
|Prior PCI||538 (88.1)||438 (88.1)||100 (87.7)|
|Dyspnea||340 (55.6)||274 (55.1)||66 (57.9)|
|Diaphoresis||154 (25.2)||124 (24.9)||30 (26.3)|
|Nausea||166 (27.2)||137 (27.6)||29 (25.4)|
|Weakness||222 (36.3)||178 (35.8)||44 (38.6)|
|Dizziness||153 (25.0)||123 (24.7)||30 (26.3)|
|Palpitations||93 (15.2)||79 (15.9)||14 (12.3)|
|Old||14 (2.3)||11 (2.2)||3 (2.6)|
|New||3 (0.5)||1 (0.2)||2 (1.8)|
|Old||45 (7.4)||29 (5.8)||16 (14.0)|
|New||30 (4.9)||22 (4.4)||8 (7.0)|
|Old||105 (17.2)||86 (17.3)||19 (16.7)|
|New||75 (12.3)||58 (11.7)||17 (14.9)|
|Old||3 (0.5)||3 (0.6)||0 (0.0)|
|New||1 (0.2)||0 (0.0)||1 (0.9)|
|Old||67 (11.0)||54 (10.9)||13 (11.4)|
|New||34 (5.6)||28 (5.6)||6 (5.3)|
|ECG diagnostic category|
|STEMI||1 (0.2)||0 (0.0)||1 (0.9)|
|Ischemia not known to be old||68 (11.1)||50 (10.1)||18 (15.8)|
|Ischemia known to be old||72 (11.8)||56 (11.3)||16 (14.0)|
|Abnormal/nondiagnostic||200 (32.7)||156 (31.4)||44 (38.6)|
|Nonspecific||144 (23.6)||122 (24.5)||22 (19.3)|
|Normal||126 (20.6)||113 (22.7)||13 (11.4)|
|Initial troponin positive||33 (5.4)||0 (0.0)||33 (28.9)|
Post hoc adjustment for site of enrollment was performed at the request of the reviewers, but site did not have independent predictive value and therefore was not added to our prespecified model. At the request of reviewers we also performed a sensitivity analysis including only those patients who were troponin negative at presentation in the same prespecified model.
Among 2,000 patients presenting with symptoms consistent with ACS enrolled in the DISPO-ACS study, 611 had prior revascularization (538 had PCI, 232 had CABG, 159 had both) and were included in the analysis. There was a mean (±SD) age of 66 (±14) years, 44% were female, and 22% were black. Patient demographics and presenting characteristics are given in Table 1. Of the 611 study patients, 517 (84.5%) were admitted, 90 (15%) were discharged from the ED, and three (0.5%) were transferred to another facility. During their hospital course, 195 (31.9%) received an echocardiogram, 101 (16.5%) received a nuclear stress test, 38 (6.2%) received graded exercise stress testing, and 121 (19.8%) underwent diagnostic catheterization. Testing performed and ED disposition are summarized in Table 2.
|Male vs. female||1.55||(0.96–2.51)||0.073|
|Family history of CAD vs. no family history||2.06||(1.26–3.36)||0.004|
|Abnormal/ischemic changes vs. normal/nonspecific||1.89||(1.16–3.09)||0.011|
|Positive baseline troponin||13.25||(5.94–29.55)||<0.001|
By 30 days, 101 (17%) patients had cardiovascular events (81 during the index visit and 20 during follow-up). There were four deaths, 28 AMIs, and 67 revascularization procedures. An additional 13 patients had a positive troponin. A positive troponin has not been included as a cardiovascular outcome but as a primary risk stratification variable. Statistical models suggested that being male (OR = 1.67, 95% CI = 1.07 to 2.62), nonblack (OR = 1.95, 95% CI = 1.07 to 3.56), having a family history of coronary artery disease (CAD; OR = 2.09, 95% CI = 1.32 to 3.3), elevated lipids (OR = 1.71, 95% CI =1.04 to 2.82), prior AMI (OR = 1.79, 95% CI = 1.16 to 2.76), abnormal ECG on arrival (OR = 2.1, 95% CI = 1.33 to 3.34), and a positive initial troponin (OR = 13.3, 95% CI = 5.9 to 29.6) were predictive of adverse cardiovascular events. The multivariable model (Table 2) had a moderate fit with a c-statistic of 0.716 (95% CI = 0.66 to 0.77) and included family history of CAD (OR = 2.06, 95% CI = 1.26 to 3.36), abnormal initial ECG (OR = 1.89, 95% CI = 1.16 to 3.09), and positive initial troponin (OR = 13.3, 95% CI = 5.9 to 29.6). Sensitivity analysis of the 578 patients who were troponin negative at presentation revealed 78 with an adverse cardiovascular outcome. The best-fit multivariable included male sex (OR = 1.57, 95% CI = 0.95 to 2.61), family history of CAD (OR = 2.11, 95% CI = 1.26 to 3.54), and abnormal initial ECG findings (OR = 1.87, 95% CI = 1.12 to 3.12). The effect sizes and model c-statistic (0.647, 95% CI = 0.584 to 0.711) were very similar to that of the original analysis.
Prior research has primarily focused on risk stratification tools for identifying high-risk patients who would benefit from rapid transition to a catheterization laboratory, or conversely identifying the very low-risk patient who, despite symptoms of potential ACS, may be eligible for rapid discharge or a brief observation period. Common clinical characteristics incorporated in risk stratification schemes include demographics, presenting symptoms, past medical history, initial ECG, and cardiac marker results. There is no risk score that focuses on patients with prior revascularization. From the emergency care provider perspective, these patients are generally considered to be among the high-risk cohort. Logic, however, would dictate that patients with prior revascularization may be more attuned to symptoms and actually more likely to seek medical care with lesser severity of symptoms. Additionally, this cohort would still also be subject to conditions that can be difficult to distinguish from ACS, such as gastroesophageal reflux disease. Therefore, although they already have demonstrated coronary disease, they may also present to the ED with symptoms that are noncardiac. How best to risk stratify patients known to have revascularization has not been well addressed in prior studies.
The TIMI risk score categorizes patients on a scale ranging from 0 (low risk) to 7 (high risk) based on age >65 years, presence of three or more conventional cardiac risk factors (including hypertension, diabetes, cholesterol elevation, family history of CAD/MI, and history of tobacco use), aspirin use within 7 days prior to ED presentation, two or more anginal events in the past 24 hours, ST-segment elevation or depression of >1 mm, elevated cardiac biomarkers, and known CAD (including PCI, CABG, cardiac catheterization with known disease, and documented prior MI). Patient risk increases as the overall TIMI score rises. While the TIMI risk score includes consideration of prior coronary stenosis, including PCI and CABG, it does not weigh this factor higher than any one of the others, and the overall risk depends on the sum of all criteria.7 The 30-day cardiovascular event rate for patients with prior revascularization in our study suggests also that taken together, patients with revascularization are higher risk than others, but our analysis suggests that this cohort can be further risk stratified.
The Sanchis score predicts mortality or MI at 1 year in low-risk patients.13 Similar to the TIMI risk score, the Sanchis scale categorizes patients from 0 (very low risk) to ≥4 (very high risk). The score is derived from a chest pain score ≥10 points, the number of chest pain episodes in past 24 hours, age, presence of diabetes, and prior revascularization.13 Because this tool focuses solely on low-risk patients and long-term outcomes, its potential for identifying those revascularized patients at high risk of 30-day cardiovascular outcomes is limited. The Goldman scale, however, focuses on risk categorization and can guide immediate care decisions.9 The scale uses aspects of the patient history, physical examination, and the initial ECG to determine the potential risk of 72-hour adverse cardiovascular outcomes.14,15 This scale, however, lacks any risk factor related to prior CAD or revascularization and therefore has unknown utility for risk stratification in our patient cohort.
Since the development of these tools (TIMI, Sanchis, and Goldman), several studies have been performed to determine their utility. In 2006, the predictive value of the TIMI risk score for predicting 30-day cardiovascular outcomes was found to be high, but insufficient for use on its own to determine disposition of ED patients.7,16 Test characteristics, including sensitivity, specificity, and predictive values, have been shown to be poor for all three scores for predicting 30-day cardiovascular outcomes,9 emphasizing the need for more accurate risk stratification tools.
Revascularization is generally considered an important cardiovascular risk factor driving clinical decisions to admit ED patients with suspected ACS, despite the limitations in existing risk stratification algorithms. Our study demonstrates that within the cohort of patients who have already undergone revascularization, there are potential benefits of considering other clinical characteristics when evaluating their 30-day risk of adverse cardiovascular events. We found that in our multivariable model, family history of CAD, abnormal initial ECG, and early troponin I elevations were predictive of 30-day cardiovascular events. These risk factors are similar to those most predictive of risk in patients without revascularization; the presenting ECG and initial cardiac markers values are the most informative factors within the TIMI score.7 In patients with prior revascularization, other presenting clinical characteristics were not shown to be clinically significant risk factors.
Our study does have certain limitations that are worth discussing. This study was not intended to compare patients with prior revascularization to patients who have not received revascularization, as this is not an important decision point when a patient presents to the ED. Individual patients are either revascularized or not. Although it was multicenter with a diverse group of patient demographics, we did not enroll consecutive patients, and thus have the limitations in generalizability that stem from selection of a cohort of patients who consented to a biomarker study, rather than an all-comers patient population. The study is also subject to work-up bias due to variations in clinical practice, as patients did not receive a standardized pathway of care. Patients who did not receive diagnostic testing are unlikely to be identified as candidates for new revascularization; however, our other end points, death and AMI, were uniformly assessed. We excluded STEMI present on admission, as these patients are easily identified as meeting an outcome by the time of arrival, and it is well known that most of these patients also receive revascularization. Finally, we did not examine whether different forms of revascularization (bare metal stent, drug eluting stent, or bypass) may present differently; rather, we considered revascularization as whole. The possibility that risk of 30-day cardiovascular events differs based on the type of prior revascularization is real, but was not addressed in our study.
As the number of coronary revascularization procedures performed continues to increase, the proportion of patients presenting to the ED with a history of prior revascularization will continue to grow. Our findings suggest that in patients with a history of percutaneous coronary intervention or coronary artery bypass graft, the initial electrocardiogram and early troponin I elevations are highly predictive of 30-day death, nonfatal acute myocardial infarction, or revascularization. Clinical presentation was not predictive of 30-day cardiovascular events. The risk stratification of patients with a history of prior percutaneous coronary intervention or coronary artery bypass graft who present to the ED with symptoms suggestive of possible acute coronary syndrome still remains a challenge. Further research needs to be conducted to enable early identification of those revascularized patients at high risk of subsequent cardiovascular events.