Optimal Timing of Intervention in Non–ST-Segment Elevation Acute Coronary Syndromes
http://www.100md.com
《循环学杂志》
the Cardiovascular Division, Beth Israel Deaconess Medical Center, Boston, Mass (J.W.R., D.J.C.)
Duke Clinical Research Institute and Division of Cardiology, Duke University Medical Center, Durham, NC (E.D.P., A.Y.C., M.T.R., E.M.O., P.B.B., E.R.D.)
Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Mass (C.P.C.)
University of Oklahoma Health Sciences Center, Oklahoma City, Okla (J.F.S.)
Department of Health Care Policy, Harvard Medical School (S.-L.N.)
Pennsylvania Hospital, University of Pennsylvania School of Medicine, Philadelphia, Pa (C.V.P.).
Abstract
Background— Recent studies indicate that a routine invasive approach for patients with unstable angina (UA) and non–ST-segment elevation myocardial infarction (NSTEMI) yields improved outcomes compared with a conservative approach, but the optimal timing of this approach remains open to debate.
Methods and Results— We used day of hospital presentation as an instrumental variable to study the impact of timing of cardiac catheterization and revascularization therapy on acute outcomes (death, reinfarction, stroke, cardiogenic shock, or congestive heart failure) among patients with UA and NSTEMI. Between January 2001 and September 2003, 56 352 patients with UA or NSTEMI were treated at 310 US hospitals participating in the CRUSADE national quality improvement initiative. Weekend patients were defined as those who presented to the hospital between 5 PM on Friday and 7 AM on Sunday. All other patients were classified as weekday. Weekday patients were similar to weekend patients in terms of demographics, clinical characteristics, and the use of medical therapies in the first 24 hours. Although overall rates of cardiac catheterization and revascularization were similar for the 2 groups, median time to catheterization was significantly longer for weekend than for weekday patients (46.3 versus 23.4 hours, P<0.0001). This delay was not associated with increased in-hospital adverse events, including death (weekend 4.4% versus weekday 4.1%, P=0.23), recurrent MI (2.9% versus 3.0%, P=0.36), or their combination (6.6% versus 6.6%, P=0.86). These findings were not affected by risk adjustment or use of alternative definitions of weekend versus weekday presentation. When weekend presentation was used as the basis for an instrumental variable analysis, we found that catheterization within the first 12 hours of presentation was associated with a nonsignificant trend toward reduced in-hospital mortality (absolute risk reduction 1.9%; 95% CI 6.7% lower to 2.9% higher; P=0.43) that decreased with longer treatment delays.
Conclusions— Although weekend presentation is associated with a delay in invasive management among patients with UA and NSTEMI, in the context of contemporary medical therapy, this does not increase adverse events. Weekend presentation appears to fulfill accepted criteria as an instrumental variable for studying the optimal timing of invasive management for acute coronary syndrome patients. Using weekend status as an instrumental variable, we found no significant benefit to early catheterization, although we could not exclude an important risk reduction, particularly for catheterization within 12 hours of presentation.
Key Words: coronary disease myocardial infarction angioplasty catheterization revascularization
Introduction
The syndromes of unstable angina (UA) and non–ST-segment elevation myocardial infarction (NSTEMI) account for much of the morbidity and mortality of cardiovascular disease.1 Until recently, there was considerable debate regarding whether these patients should be treated conservatively or undergo routine invasive assessment and revascularization therapy. With the completion of several landmark trials, there is now general consensus regarding the value of an early invasive strategy, particularly in higher-risk patient subsets.2–5
Nonetheless, there remains considerable uncertainty about the optimal timing of cardiac catheterization and revascularization for UA/NSTEMI patients. The definition of "early" intervention differed from as brief as 24 to 48 hours after presentation in the Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy–Thrombolysis in Myocardial Infarction 18 (TACTICS-TIMI 18) trial3 to as long as 3 to 5 days in the Randomized Intervention Trial of Unstable Angina-3 (RITA-3) and the Fast Revascularization During Instability in Coronary Artery Disease-2 (FRISC-2) trial.4,5 Moreover, post hoc analysis of the TACTICS-TIMI 18 trial failed to identify a benefit of very early intervention (<48 hours) versus later intervention,6 whereas data from the Global Registry of Acute Coronary Events (GRACE) have suggested that very early intervention is associated with worse clinical outcomes.7
Recently, a single small trial directly compared very early intervention (within hours of presentation) versus delayed intervention (5 days) in high-risk UA/NSTEMI patients.8 In that study, despite use of aggressive anticoagulation and antiplatelet therapy, very early treatment was superior to a prolonged cooling off strategy. If this finding were confirmed in other studies, it would suggest that high-risk non–ST-segment elevation acute coronary syndrome (NSTE-ACS) patients should be managed with immediate coronary angiography and revascularization when appropriate, similar to the current primary angioplasty treatment paradigm for patients with ST-segment elevation MI. Given that admissions for NSTE-ACS currently outnumber admissions for STEMI by nearly a 4:1 margin,9 widespread adoption of such a policy would have important implications for hospital staffing, workflow, and regionalization of cardiovascular services.
The Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the ACC/AHA guidelines (CRUSADE) quality improvement initiative is a voluntary nationwide program designed to examine processes and outcomes of care for patients hospitalized with UA/NSTEMI.10 We used the CRUSADE database to study the optimal timing of cardiac catheterization and coronary revascularization in these patients by means of 2 related nonexperimental, quasi-randomized study designs: a natural experiment and an instrumental variable analysis.
Methods
Patient Population
The patient population for the present study was derived from the CRUSADE initiative. As described previously, CRUSADE is a voluntary, industry-supported, nationwide quality improvement program that enrolls unselected high-risk patients hospitalized with UA or NSTEMI.10 To be eligible, patients must arrive at a participating facility either via the emergency department or by transfer within 24 hours of onset of symptoms. In addition, their initial evaluation must reveal 1 or more high-risk features, including ST-segment depression on ECG, transient ST-segment elevation, or positive cardiac biomarkers. The study protocol was approved by the institutional review boards at each institution.
Data Collection
For each patient, data were collected based on chart review by trained personnel using standardized data collection forms and definitions. Data collected included baseline clinical characteristics, laboratory results, use of acute medications (within 24 hours of hospital arrival), use and timing of invasive cardiac procedures (coronary angiography, percutaneous coronary revascularization, or bypass surgery), and clinical outcomes through hospital discharge. Contraindications to specific therapies were also recorded.
All adverse events were investigator-reported. Postadmission myocardial infarction was defined as clinical signs and symptoms of a new infarction confirmed by new ECG changes or (re)elevation of cardiac enzymes. Stroke was defined as a new focal neurological defect that lasted >24 hours. Congestive heart failure was defined as exertional dyspnea, orthopnea, rales over more than one third of lung fields, elevated jugular venous pressure, or pulmonary congestion on chest radiograph thought to be related to cardiac dysfunction. Cardiogenic shock was defined as systolic blood pressure <90 mm Hg for more than 1 hour believed to be secondary to cardiac dysfunction.
General Analytic Approach
The usual approach to using observational (ie, nonrandomized) data to assess the impact of treatment delays on outcomes for patients with UA/NSTEMI would involve classifying patients according to the timing of invasive therapy and comparing outcomes according to this classification with use of standard risk-adjustment techniques, such as logistic regression or propensity scores.11 Although commonly employed, this approach inevitably suffers from residual confounding, in particular because standard risk-adjustment techniques can only account for measured confounders.12,13
Several alternative approaches that have been termed "quasi-randomized" take advantage of natural variation in practice patterns and allow researchers to control for both measured and unmeasured confounding in observational research.14 One such method is the "natural experiment." In this approach, patients are segregated and compared according to an observed characteristic that is related to the overall care pattern of interest but is otherwise unrelated to any underlying patient characteristics.
Another related alternative to standard risk-adjustment is instrumental variable analysis. An instrumental variable is an observable factor that is associated with a specific treatment pattern but is otherwise unrelated to patient characteristics and does not directly affect the outcome of interest.15 By utilizing the instrumental variable (rather than the treatment itself) as a covariate in a 2-stage regression analysis, one can take advantage of the generalizability and statistical power afforded by large data sets while overcoming many of the intrinsic limitations of observational studies.
Natural experiments and instrumental variable analysis have the advantage of allowing the investigator to eliminate both observed and unobserved confounding as an explanation for a study’s findings, similar to the benefits of randomization in an experimental study.15 Although these quasi-experimental approaches are attractive in theory, identification of an appropriate natural experiment or instrumental variable is often challenging. In the case of timing of invasive therapy for patients with NSTE-ACS, however, day of hospital admission is an attractive candidate for both approaches. On the basis of preliminary clinical observations from our center, we hypothesized that patients admitted on the weekend would be less likely to undergo very early catheterization and revascularization than patients admitted during the work week. If patient characteristics and other treatment patterns did not differ substantially on the basis of weekday versus weekend presentation, this characteristic could serve as the basis for a natural experiment comparing the impact of differential treatment delays on patient outcomes. Furthermore, weekend status could also serve as an instrumental variable for more direct examination of the impact of timing of cardiac catheterization and revascularization on outcomes among patients with NSTE-ACS.
Weekend Status
For both our natural experiment and instrumental variable analyses, we defined weekend patients as those who presented to the admitting hospital between 5 PM. Friday and 7 AM Sunday. All other patients were considered weekday patients. These definitions were chosen so as to maximize the number of weekend patients presenting >24 hours from Monday morning, at which time we expected catheterization laboratory facilities would be fully operational. Alternative definitions of weekend and weekday were considered in sensitivity analyses.
Statistical Analysis
Baseline characteristics, treatment patterns (medication use within 24 hours of presentation and use and timing of coronary angiography and revascularization procedures), and in-hospital outcomes were compared between the weekend and weekday presentation groups. Continuous variables were described as medians with interquartile ranges (25th, 75th percentile) or mean±SD, and categorical variables were described as frequencies. Continuous variables were compared with Wilcoxon rank-sum tests, whereas categorical variables were compared with 2 tests. Time to coronary angiography and revascularization procedures were described with Kaplan-Meier techniques and compared by the log-rank statistic.
For our natural experiment, we sought to examine the relationship between weekend presentation and outcomes. In these analyses, we adjusted for a broad range of patient and hospital characteristics, including age, female gender, body mass index, white race, hypertension, diabetes, hypercholesterolemia, prior myocardial infarction, prior percutaneous coronary intervention (PCI), prior coronary artery bypass grafting, prior congestive heart failure, prior stroke, smoking status, renal insufficiency, family history of coronary artery disease, presence of ST-segment depression, presence of transient ST-segment elevation, positive cardiac markers, signs of congestive heart failure on admission, admission heart rate, admission systolic blood pressure, insurance status, total number of hospital beds at the treating facility, geographic region, teaching status, and attending physician specialty (cardiologist or noncardiologist). All of these covariates were included in the regression analyses, regardless of their statistical significance. Because patients within a hospital were more likely to be similar than patients from different hospitals, all adjusted analyses were performed with generalized estimating equation logistic models16 to account for correlations among clustered responses (eg, within-hospital correlations). All ORs are reported with weekday presentation as the reference group.
To more directly characterize the impact of timing of cardiac catheterization on in-hospital outcomes, we also used day of hospital presentation as the basis for an instrumental variable analysis. We first confirmed that day of hospital presentation was a valid instrumental variable by verifying that (1) it correlated with timing of cardiac catheterization, (2) it was independent of other patient characteristics, and (3) it was independent of patient outcomes except as mediated by timing of cardiac catheterization. To test the first assumption, we ran generalized estimating equation models to predict early catheterization as a function of day of hospital presentation (eg, weekend versus weekday). To test the second assumption, we compared patient characteristics between patients presenting on the weekend and patients presenting during the work week. Finally, to test the last assumption, we regressed in-hospital outcomes on the weekend/weekday variable, after adjusting for early catheterization along with other patient and hospital characteristics.
After confirming that weekend presentation satisfied these conditions, we then performed a fully adjusted instrumental variable analysis using a 2-stage linear regression approach.17,18 The first stage predicted the performance of early catheterization as a function of weekend (versus weekday) presentation and other patient and hospital characteristics. The predicted probabilities of early catheterization from this first stage were used as input to the second stage, which regressed in-hospital mortality on these predicted probabilities, along with the other patient and hospital characteristics. The effect of early catheterization on in-hospital mortality was estimated by subtracting the mean predicted mortality when early catheterization=1 from the mean predicted mortality where early catheterization=0. This value is equal to the regression coefficient for the predicted probability of early catheterization and represents an upper-bound estimate of the impact of early catheterization on in-hospital outcome. Separate instrumental variable analyses were performed with 4 alternative definitions of early catheterization: <12, <24, <36, and <48 hours after presentation. A probability value of <0.05 was established as the level of statistical significance for all tests. The instrumental variable analyses were performed with the IVREG command in STATA version 8.0 (STATA Corporation), and all other analyses were performed with SAS software (version 8.2, SAS Institute).
Results
Study Population
Between January 1, 2001, and September 30, 2003, a total of 77 760 patients with NSTE-ACS were entered into the CRUSADE database at participating study centers. After the exclusion of 21 408 patients because they were either enrolled at hospitals without PCI or surgical facilities (n=19 960), transferred out of the CRUSADE hospital (generally for performance of cardiac catheterization or revascularization; n=1334), or had missing information regarding day of admission (n=114), the remaining 56 352 patients from 310 centers composed our study population. All of these patients were enrolled at centers with full catheterization, PCI, and surgical capabilities.
Of the total study population, 10 804 presented between 5 PM on Friday and 7 AM on Sunday and were included in the weekend group, whereas the remaining 45 548 presented at other times and comprised the weekday group. Baseline characteristics of the 2 groups were generally well matched (Table 1). The weekend group tended to have slightly higher-risk profiles, as reflected by marginally higher PURSUIT (Platelet IIb/IIIa in Unstable angina: Receptor Suppression Using Integrilin Therapy) risk scores and a greater proportion of patients with positive cardiac markers. Medical care, including use of evidence-based therapies (antiplatelet agents, antithrombotic therapy, and -blockers) within the first 24 hours of admission, was also similar for the weekend and weekday groups (Table 2).
Cardiac Catheterization and Revascularization
Although the weekday and weekend groups were similar in their overall rates of in-hospital cardiac catheterization (73.0% versus 71.2%), PCI (42.0% versus 40.3%), and bypass surgery (13.4% versus 12.9%), the timing of these interventions was substantially delayed for weekend patients (Table 2). Median time (25th, 75th percentile) to cardiac catheterization (among patients who underwent the procedures) was 23.4 hours (10.4 to 45.3 hours) for weekday patients compared with 46.3 hours (21.4 to 63.5 hours) for weekend patients (P<0.0001). Moreover, the percentage of patients undergoing cardiac catheterization at any time in the first 48 hours was lower for the weekend group than for the weekday group (Figure 1). Similar findings were observed for PCI and bypass surgery (Figure 2).
In-Hospital Outcomes
The results of our natural experiment comparing in-hospital outcomes for the weekday and weekend groups are summarized in Table 3. There were no significant differences in rates of death, myocardial infarction (or reinfarction for patients whose initial diagnosis was NSTEMI), cardiogenic shock, congestive heart failure, or the composite between the 2 groups, with crude ORs ranging from 0.96 for reinfarction or stroke to 1.07 for cardiogenic shock or congestive heart failure. Given the comparable baseline characteristics for the weekday and weekend groups, risk adjustment had little effect on the ORs or their CIs, and none of the differences in risk-adjusted outcome were statistically significant (Table 3). The only difference in hospital outcome between the 2 groups was length of stay (Figure 3). The mean length of stay was significantly longer for the weekend group than for the weekday group (6.1 versus 5.8 days, P<0.0001).
Stratified analyses according to baseline patient characteristics are summarized in Figure 4. There were no significant differences for in-hospital mortality between the weekday and weekend groups according to age (<65 versus 65 years), gender, diabetes mellitus, or type of presentation (unstable angina versus NSTEMI). To investigate whether our results were sensitive to the definition of weekend and weekday presentation, we reanalyzed our data according to several alternative definitions (Table 4). Regardless of the specific definition of weekend and weekday presentation considered, the crude and adjusted ORs for in-hospital mortality were virtually identical to our base case analysis.
Instrumental Variable Analysis
The estimated impact of early catheterization on in-hospital mortality from the instrumental variable analysis is summarized in Table 5. For each definition of early catheterization, there was a nonsignificant trend toward lower mortality among patients receiving early catheterization than among those receiving later catheterization or no catheterization. This effect was largest (absolute risk reduction 1.9%; 95% CI 6.7% lower to 2.9% higher) when early catheterization was defined as being within 12 hours of presentation.
Discussion
Although recent consensus guidelines recommend a strategy of early invasive management for patients with high-risk acute coronary syndromes,2 few data exist on the optimal timing of cardiac catheterization and revascularization for such patients. In the United States, contemporary approaches may range from immediate angiography and revascularization (similar to the standard approach for ST-elevation MI) to medical stabilization with antiplatelet and antithrombotic therapy followed by angiography and revascularization in the next several days. In the present study, we used the related techniques of natural experimentation and instrumental variable estimation to provide insight into this important question from a large-scale, unselected patient population derived from the CRUSADE quality improvement initiative.
By stratifying patients according to date and time of initial presentation, we identified 2 patient subgroups that were comparable with respect to all observed characteristics except timing of invasive therapy, which was delayed by 24 hours among patients who presented on the weekend compared with those who presented during the work week. Despite this substantial delay, we were unable to identify any differences in adverse outcomes, including death or recurrent myocardial infarction, between the weekend and weekday subgroups. These findings were consistent across a range of subgroups, unaffected by conventional risk-adjustment, and insensitive to alternative definitions of "weekend presentation." Given the large sample size and broad patient population included in the CRUSADE initiative, these findings provide strong evidence that in-hospital outcomes would not be improved substantially by adopting a treatment strategy on weekends similar to our current approach during the work week, with reductions in median delays in time to cardiac catheterization from 48 to 24 hours.
The only previous study to rigorously address the question of optimal timing of revascularization therapy for patients with NSTE-ACS was the Intracoronary Stenting with Antithrombotic Regimen Cooling-Off (ISAR-COOL) trial.8 In that study, 410 patients with UA or NSTEMI received aggressive antithrombotic therapy with aspirin, clopidogrel, heparin, and tirofiban and were randomized to early intervention (within 6 hours) or delayed intervention (between 3 and 5 days). The principal finding of the study was that the early intervention strategy was associated with a 49% reduction in the risk of death or myocardial infarction at 30 days (95% CI 1% to 74%), all of which occurred during the pretreatment period. The investigators concluded that the duration of antithrombotic pretreatment should be as brief as possible to minimize the risk of ischemic complications in this patient population.
At first blush, the results of the present weekend versus weekday natural experiment would appear to be discordant with those of the ISAR-COOL trial. There are several major differences between the 2 studies, however, that may explain this apparent discrepancy. First, by design, the ISAR-COOL trial proscribed revascularization in the delayed group unless the patient experienced an ischemic complication, severe refractory angina, or hemodynamic instability. In the present study, both weekend and weekday patients could undergo intervention at any time according to the clinical judgment of their treating physicians. Indeed, even among weekend patients, 9% underwent angiography within 6 hours of presentation, and 17% underwent angiography within 24 hours. It is likely that these patients were selected because of ongoing ischemia or other characteristics that would suggest enhanced benefit from an early revascularization strategy.19
Second, the between-group differences in time to intervention were substantially greater in ISAR-COOL (median 2.6 versus 86 hours) than in the present study (median 23.4 versus 46.3 hours). It is possible that significant differences in outcome would have been observed in the present study had the extent of delay been similar. However, given its much larger sample size, our observational study had >80% power to detect a relative difference in mortality of 18% and 13% in death or myocardial infarction. Moreover, in the United States, it is unlikely that NSTE-ACS patients, in particular, those with evidence of recurrent or ongoing ischemia, would wait up to 5 days to undergo cardiac catheterization (as occurred in ISAR-COOL).
Finally, even though randomized clinical trials and natural experiments are alternative techniques for eliminating unmeasured confounding, these 2 study designs address fundamentally different questions. Randomized trials are designed to address a specific clinical question for an individual patient. In contrast, the present natural experiment estimates the average effect of a delay in performance of invasive therapy over the marginal range in treatment delays defined by the stratification variable.14 In the case of our natural experiment, the question we have addressed is, "What would be the effect of a policy that increased (or reduced) average invasive treatment delays by 24 hours for all patients with non-ST elevation acute coronary syndromes" Our finding that modest incremental treatment delays in the weekend group were not associated with an increase in adverse outcomes thus suggests that for the sizeable proportion of NSTE-ACS patients admitted to US hospitals on weekends, in-hospital outcomes would not be improved substantially by simply adopting a treatment strategy on weekends that is similar to our current approach during the work week.
Whether more dramatic reductions in existing delays in timing of invasive therapy (such as those achieved in ISAR-COOL) would result in meaningful improvements in in-hospital outcomes for our broad population is more directly addressed by our instrumental variable analysis but cannot be resolved definitively by the present study. Although our instrumental variable analysis did not demonstrate a significant reduction in in-hospital mortality associated with earlier cardiac catheterization, the CIs surrounding the absolute mortality reductions were relatively wide, particularly for catheterization within 12 hours of hospital admission. In fact, the central estimate of our instrumental variable analysis suggests that catheterization within 12 hours may be associated with an absolute 1.9% reduction in mortality and an even larger benefit if one considers the full 95% CI. This uncertainty reflects an inherent limitation of instrumental variables analyses in that they generally sacrifice statistical power in exchange for limiting confounding. Nonetheless, when taken together with the ISAR-COOL results, our findings provide additional support to the concept that if early catheterization is to achieve its maximal benefit for patients with NSTE-ACS, greater benefits may be associated with performance of catheterization within 12 hours of presentation.
Study Limitations
This study has several limitations. First, participation in CRUSADE is voluntary and thus may not be fully representative of the universe of NSTE-ACS patients. Nonetheless, the large number of participating centers and high rates of adverse outcomes compared with recent randomized clinical trials suggests that our population is relatively unselected, and thus, our findings should be applicable to the vast majority of such patients.
Second, we were only able to compare in-hospital outcomes. It is unknown whether longer-term mortality or late congestive heart failure was increased in the weekend group. Such an effect seems unlikely, however, because overall rates of revascularization were similar by hospital discharge, and there were no trends toward increased complications in the weekend group in the present study. Moreover, among patients managed aggressively (similar to the present study population), the majority of adverse events emerge by 7 days,3 close to the average length of stay for the present study. Thus, it is unlikely that a large number of adverse events would have occurred in a differential fashion beyond the scope of our assessment.
Third, our definition of postadmission myocardial infarction was based entirely on clinical signs and symptoms as detected during clinical care rather than routine surveillance of cardiac isoenzymes. Thus, it is possible that the present study could have failed to detect some episodes of asymptomatic cardiac myonecrosis. Of note, the rates of in-hospital death and myocardial infarction in the present study were, if anything, higher than those in most recent clinical trials of NSTE-ACS patients.3–5 Moreover, in the absence of differences in in-hospital mortality or other overt complications, the clinical significance of such asymptomatic events remains somewhat controversial.20,21
Finally, a critical assumption underlying this analysis is the validity of weekend presentation as an instrumental variable for evaluating delays in invasive therapy for patients with NSTE-ACS. As an instrumental variable, we assumed that weekend presentation would be associated with delayed invasive treatment but would not be correlated with any other differences in patient characteristics or treatment patterns (either observed or unobserved) that directly affect outcomes, an assumption whose validity can be explored but not proven definitively. Indeed, it is possible that weekend presentation is associated with other factors such as hospital staffing and unobserved practice patterns that contributed in part to our findings. Nonetheless, the fact that weekday and weekend patients were similar with respect to baseline clinical characteristics and acute medical therapies and that weekend status was not independently associated with in-hospital mortality after controlling for timing of invasive therapy supports the validity of this assumption.
Conclusions
In this observational study, we found that weekend versus weekday presentation represents a potentially useful instrumental variable for investigating the timing of invasive therapy for patients with NSTE-ACS. Although weekend presentation was associated with a delay of nearly 24 hours in the performance of cardiac catheterization and revascularization, in the context of contemporary medical therapy, there is no evidence that these delays are associated with an increase in adverse outcomes. These findings thus suggest that a policy of more liberal access to interventional therapy on weekends would not substantially improve population-wide outcomes compared with current US standards. Whether more dramatic reductions in across-the-board delays might result in meaningful improvements in in-hospital mortality cannot be definitively resolved by the present study, however, and is deserving of further examination in either a larger patient population or with longer follow-up.
Acknowledgments
CRUSADE is a National Quality Improvement Initiative of the Duke Clinical Research Institute. CRUSADE is funded by Schering-Plough Corporation. Bristol-Myers Squibb/Sanofi-Aventis Pharmaceuticals Partnership provides additional funding support. Millennium Pharmaceuticals, Inc, also funded the research presented here.
Disclosure
Dr Peterson received research grants from Bristol-Myers Squibb/Sanofi Pharmaceuticals, Schering-Plough, and Millennium Pharmaceuticals. Dr Roe has received research grants and honoraria from Millennium Pharmaceuticals, Schering-Plough, and Bristol-Myers Squibb/Sanofi-Aventis Pharmaceutical Partnership. Dr Ohman has received research grants from Millennium Pharmaceuticals, SPRI, Sanofi-Aventis, Berelex, and Bristol-Myers Squibb. Dr Cannon received research grant support from Bristol-Myers Squibb, Merck, Sanofi-Aventis, AstraZeneca, and Merck-Schering Plough, serves on the advisory boards for AstraZeneca, Bristol-Myers Squibb, GlaxoSmithKline, Merck, Merck/Schering-Plough Partnership, Pfizer, Sanofi-Aventis, Schering Plough, and Vertex, received lecture fees from AstraZeneca, Bristol-Myers Squibb, Guliford Pharmaceuticals, Merck, Millennium Pharmaceuticals, Pfizer, Sanofi-Aventis, and Schering Plough, and received honoraria from BestMed, I3 Magnifi, and NCME. Dr Normand is a member of the Massachusetts Cardiac Care Quality Commission and leads a data-coordinating center that examines the quality of PCI care. Dr Pollack serves on the Speakers Bureaus of Sanofi-Aventis, Schering-Plough, Millennium Pharmaceuticals, Genentech, and Pfizer, serves as a consultant for Sanofi-Aventis, Millennium Pharmaceuticals, Genentech, PDL, and The Medicines Company, received research support from Sanofi-Aventis, and owns a significant amount of stock in Pyng Medical.
Footnotes
Guest Editor for this article was Harvey D. White, DSc.
References
Yusuf S, Reddy S, unpuu S, Anand S. Global burden of cardiovascular diseases: part I: general considerations, the epidemiologic transition, risk factors, and impact of urbanization. Circulation. 2001; 104: 2746–2753.
Braunwald E, Antman E, Beasley J, Califf R, Cheitlin M, Hochman J, Jones R, Kereiakes D, Kupersmith J, Levin T, Pepine C, Schaeffer J, Smith E, Steward D, Theroux P, Alpert J, Eagle K, Faxon D, Fuster V, Gardner T, Gregoratos G, Russell R, Smith S. ACC/AHA guidelines for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on the Management of Patients with Unstable Angina). J Am Coll Cardiol. 2000; 36: 970–972.
Cannon CP, Weintraub WS, Demopoulos LA, Vicari R, Frey MJ, Lakkis N, Neumann FJ, Robertson DH, DeLucca PT, DiBattiste PM, Gibson CM, Braunwald E; TACTICS (Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy)–Thrombolysis in Myocardial Infarction 18 Investigators. Comparison of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban. N Engl J Med. 2001; 344: 1879–1887.
Fox KA, Poole-Wilson PA, Henderson RA, Clayton TC, Chamberlain DA, Shaw TR, Wheatley DJ, Pocock SJ; Randomized Intervention Trial of unstable Angina Investigators. Interventional versus conservative treatment for patients with unstable angina or non-ST-elevation myocardial infarction: the British Heart Foundation RITA 3 randomised trial: Randomized Intervention Trial of unstable Angina. Lancet. 2002; 360: 743–751.
Wallentin L, Lagerqvist B, Husted S, Kontny F, Stahle E, Swahn E. Outcome at 1 year after an invasive compared with non-invasive strategy in unstable coronary-artery disease: the FRISC II invasive randomized trial. Lancet. 2000; 356: 9–16.
McCullough PA, Gibson CM, Dibattiste PM, Demopoulos LA, Murphy SA, Weintraub WS, Neumann FJ, Khanal S, Cannon CP; TACTICS-TIMI-18 Investigators. Timing of angiography and revascularization in acute coronary syndromes: an analysis of the TACTICS-TIMI-18 trial. J Interv Cardiol. 2004; 17: 1–5.
Montalescot G, Dabbous OH, Lim MJ, Flather MD, Mehta RH; Global Registry of Acute Coronary Events Investigators. Relation of timing of cardiac catheterization to outcomes in patients with non-ST-segment elevation myocardial infarction or unstable angina pectoris enrolled in the Multinational Global Registry of Acute Coronary Events. Am J Cardiol. 2005; 95: 1397–1403.
Neumann FJ, Kastrati A, Pogatsa-Murray G, Mehilli J, Bollwein H, Bestehorn HP, Schmitt C, Seyfarth M, Dirschinger J, Schomig A. Evaluation of prolonged antithrombotic pretreatment ("cooling off" strategy) before intervention in patients with unstable coronary syndromes: a randomized controlled trial. JAMA. 2003; 290: 1593–1599.
American Heart Association, Heart Disease and Stroke Statistics: 2004 Update. Dallas, Tex: American Heart Association; 2003.
Hoekstra JW, Pollack CV Jr, Roe MT, Peterson ED, Brindis R, Harrington RA, Christenson RH, Smith SC, Ohman EM, Gibler WB. Improving the care of patients with non-ST-elevation acute coronary syndromes in the emergency department: the CRUSADE initiative. Acad Emerg Med. 2002; 11: 1146–1155.
Braitman LE, Rosebaum PR. Rare outcomes, common treatments: analytic strategies using propensity scores. Ann Intern Med. 2002; 137: 693–696.
Creswell J. Research Design: Qualitative and Quantitative Approaches. London, UK: SAGE Publications; 1994.
Rosenbaum PR. Discussing hidden bias in observational studies. Ann Intern Med. 1991; 115: 901–905.
Gennetian L, Bos J, Morris P. Using instrumental variables to learn more from social policy experiments. MDRC Working Papers on Research Methodology.
McClellan M, McNeil B, Newhouse J. Does more intensive treatment of acute myocardial infarction in the elderly reduce morality Analysis using instrumental variables. JAMA. 1994; 272: 859–866.
Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986; 73: 13–22.
Greenland S. An introduction to instrumental variables for epidemiologists. Int J Epidemiol. 2000; 29: 772–720.
Eittner SL, Frank RG, Kessler RC. The impact of psychiatric disorders on labor market outcomes. Ind Labor Relations Rev. 1997; 51: 64–81.
Cannon CP. Evidence-based risk stratification to target therapies in acute coronary syndromes. Circulation. 2002; 106: 1588–1591.
The Clopidogrel in Unstable Angina to Prevent Recurrent Events Trial Investigators. Effects of clopidogrel in addition to aspirin in patients with acute coronary syndromes without ST-segment elevation. N Engl J Med. 2001; 345: 494–502.
Anderson KM, Califf RM, Stone GW, Neumann FJ, Montalescot G, Miller DP, Ferguson JJ III, Willerson JT, Weisman HF, Topol EJ. Long-term mortality benefit with abciximab in patients undergoing percutaneous coronary intervention. J Am Coll Cardiol. 2001; 37: 2059–2065.(Jason W. Ryan, MD, MPH; E)
Duke Clinical Research Institute and Division of Cardiology, Duke University Medical Center, Durham, NC (E.D.P., A.Y.C., M.T.R., E.M.O., P.B.B., E.R.D.)
Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, Mass (C.P.C.)
University of Oklahoma Health Sciences Center, Oklahoma City, Okla (J.F.S.)
Department of Health Care Policy, Harvard Medical School (S.-L.N.)
Pennsylvania Hospital, University of Pennsylvania School of Medicine, Philadelphia, Pa (C.V.P.).
Abstract
Background— Recent studies indicate that a routine invasive approach for patients with unstable angina (UA) and non–ST-segment elevation myocardial infarction (NSTEMI) yields improved outcomes compared with a conservative approach, but the optimal timing of this approach remains open to debate.
Methods and Results— We used day of hospital presentation as an instrumental variable to study the impact of timing of cardiac catheterization and revascularization therapy on acute outcomes (death, reinfarction, stroke, cardiogenic shock, or congestive heart failure) among patients with UA and NSTEMI. Between January 2001 and September 2003, 56 352 patients with UA or NSTEMI were treated at 310 US hospitals participating in the CRUSADE national quality improvement initiative. Weekend patients were defined as those who presented to the hospital between 5 PM on Friday and 7 AM on Sunday. All other patients were classified as weekday. Weekday patients were similar to weekend patients in terms of demographics, clinical characteristics, and the use of medical therapies in the first 24 hours. Although overall rates of cardiac catheterization and revascularization were similar for the 2 groups, median time to catheterization was significantly longer for weekend than for weekday patients (46.3 versus 23.4 hours, P<0.0001). This delay was not associated with increased in-hospital adverse events, including death (weekend 4.4% versus weekday 4.1%, P=0.23), recurrent MI (2.9% versus 3.0%, P=0.36), or their combination (6.6% versus 6.6%, P=0.86). These findings were not affected by risk adjustment or use of alternative definitions of weekend versus weekday presentation. When weekend presentation was used as the basis for an instrumental variable analysis, we found that catheterization within the first 12 hours of presentation was associated with a nonsignificant trend toward reduced in-hospital mortality (absolute risk reduction 1.9%; 95% CI 6.7% lower to 2.9% higher; P=0.43) that decreased with longer treatment delays.
Conclusions— Although weekend presentation is associated with a delay in invasive management among patients with UA and NSTEMI, in the context of contemporary medical therapy, this does not increase adverse events. Weekend presentation appears to fulfill accepted criteria as an instrumental variable for studying the optimal timing of invasive management for acute coronary syndrome patients. Using weekend status as an instrumental variable, we found no significant benefit to early catheterization, although we could not exclude an important risk reduction, particularly for catheterization within 12 hours of presentation.
Key Words: coronary disease myocardial infarction angioplasty catheterization revascularization
Introduction
The syndromes of unstable angina (UA) and non–ST-segment elevation myocardial infarction (NSTEMI) account for much of the morbidity and mortality of cardiovascular disease.1 Until recently, there was considerable debate regarding whether these patients should be treated conservatively or undergo routine invasive assessment and revascularization therapy. With the completion of several landmark trials, there is now general consensus regarding the value of an early invasive strategy, particularly in higher-risk patient subsets.2–5
Nonetheless, there remains considerable uncertainty about the optimal timing of cardiac catheterization and revascularization for UA/NSTEMI patients. The definition of "early" intervention differed from as brief as 24 to 48 hours after presentation in the Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy–Thrombolysis in Myocardial Infarction 18 (TACTICS-TIMI 18) trial3 to as long as 3 to 5 days in the Randomized Intervention Trial of Unstable Angina-3 (RITA-3) and the Fast Revascularization During Instability in Coronary Artery Disease-2 (FRISC-2) trial.4,5 Moreover, post hoc analysis of the TACTICS-TIMI 18 trial failed to identify a benefit of very early intervention (<48 hours) versus later intervention,6 whereas data from the Global Registry of Acute Coronary Events (GRACE) have suggested that very early intervention is associated with worse clinical outcomes.7
Recently, a single small trial directly compared very early intervention (within hours of presentation) versus delayed intervention (5 days) in high-risk UA/NSTEMI patients.8 In that study, despite use of aggressive anticoagulation and antiplatelet therapy, very early treatment was superior to a prolonged cooling off strategy. If this finding were confirmed in other studies, it would suggest that high-risk non–ST-segment elevation acute coronary syndrome (NSTE-ACS) patients should be managed with immediate coronary angiography and revascularization when appropriate, similar to the current primary angioplasty treatment paradigm for patients with ST-segment elevation MI. Given that admissions for NSTE-ACS currently outnumber admissions for STEMI by nearly a 4:1 margin,9 widespread adoption of such a policy would have important implications for hospital staffing, workflow, and regionalization of cardiovascular services.
The Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the ACC/AHA guidelines (CRUSADE) quality improvement initiative is a voluntary nationwide program designed to examine processes and outcomes of care for patients hospitalized with UA/NSTEMI.10 We used the CRUSADE database to study the optimal timing of cardiac catheterization and coronary revascularization in these patients by means of 2 related nonexperimental, quasi-randomized study designs: a natural experiment and an instrumental variable analysis.
Methods
Patient Population
The patient population for the present study was derived from the CRUSADE initiative. As described previously, CRUSADE is a voluntary, industry-supported, nationwide quality improvement program that enrolls unselected high-risk patients hospitalized with UA or NSTEMI.10 To be eligible, patients must arrive at a participating facility either via the emergency department or by transfer within 24 hours of onset of symptoms. In addition, their initial evaluation must reveal 1 or more high-risk features, including ST-segment depression on ECG, transient ST-segment elevation, or positive cardiac biomarkers. The study protocol was approved by the institutional review boards at each institution.
Data Collection
For each patient, data were collected based on chart review by trained personnel using standardized data collection forms and definitions. Data collected included baseline clinical characteristics, laboratory results, use of acute medications (within 24 hours of hospital arrival), use and timing of invasive cardiac procedures (coronary angiography, percutaneous coronary revascularization, or bypass surgery), and clinical outcomes through hospital discharge. Contraindications to specific therapies were also recorded.
All adverse events were investigator-reported. Postadmission myocardial infarction was defined as clinical signs and symptoms of a new infarction confirmed by new ECG changes or (re)elevation of cardiac enzymes. Stroke was defined as a new focal neurological defect that lasted >24 hours. Congestive heart failure was defined as exertional dyspnea, orthopnea, rales over more than one third of lung fields, elevated jugular venous pressure, or pulmonary congestion on chest radiograph thought to be related to cardiac dysfunction. Cardiogenic shock was defined as systolic blood pressure <90 mm Hg for more than 1 hour believed to be secondary to cardiac dysfunction.
General Analytic Approach
The usual approach to using observational (ie, nonrandomized) data to assess the impact of treatment delays on outcomes for patients with UA/NSTEMI would involve classifying patients according to the timing of invasive therapy and comparing outcomes according to this classification with use of standard risk-adjustment techniques, such as logistic regression or propensity scores.11 Although commonly employed, this approach inevitably suffers from residual confounding, in particular because standard risk-adjustment techniques can only account for measured confounders.12,13
Several alternative approaches that have been termed "quasi-randomized" take advantage of natural variation in practice patterns and allow researchers to control for both measured and unmeasured confounding in observational research.14 One such method is the "natural experiment." In this approach, patients are segregated and compared according to an observed characteristic that is related to the overall care pattern of interest but is otherwise unrelated to any underlying patient characteristics.
Another related alternative to standard risk-adjustment is instrumental variable analysis. An instrumental variable is an observable factor that is associated with a specific treatment pattern but is otherwise unrelated to patient characteristics and does not directly affect the outcome of interest.15 By utilizing the instrumental variable (rather than the treatment itself) as a covariate in a 2-stage regression analysis, one can take advantage of the generalizability and statistical power afforded by large data sets while overcoming many of the intrinsic limitations of observational studies.
Natural experiments and instrumental variable analysis have the advantage of allowing the investigator to eliminate both observed and unobserved confounding as an explanation for a study’s findings, similar to the benefits of randomization in an experimental study.15 Although these quasi-experimental approaches are attractive in theory, identification of an appropriate natural experiment or instrumental variable is often challenging. In the case of timing of invasive therapy for patients with NSTE-ACS, however, day of hospital admission is an attractive candidate for both approaches. On the basis of preliminary clinical observations from our center, we hypothesized that patients admitted on the weekend would be less likely to undergo very early catheterization and revascularization than patients admitted during the work week. If patient characteristics and other treatment patterns did not differ substantially on the basis of weekday versus weekend presentation, this characteristic could serve as the basis for a natural experiment comparing the impact of differential treatment delays on patient outcomes. Furthermore, weekend status could also serve as an instrumental variable for more direct examination of the impact of timing of cardiac catheterization and revascularization on outcomes among patients with NSTE-ACS.
Weekend Status
For both our natural experiment and instrumental variable analyses, we defined weekend patients as those who presented to the admitting hospital between 5 PM. Friday and 7 AM Sunday. All other patients were considered weekday patients. These definitions were chosen so as to maximize the number of weekend patients presenting >24 hours from Monday morning, at which time we expected catheterization laboratory facilities would be fully operational. Alternative definitions of weekend and weekday were considered in sensitivity analyses.
Statistical Analysis
Baseline characteristics, treatment patterns (medication use within 24 hours of presentation and use and timing of coronary angiography and revascularization procedures), and in-hospital outcomes were compared between the weekend and weekday presentation groups. Continuous variables were described as medians with interquartile ranges (25th, 75th percentile) or mean±SD, and categorical variables were described as frequencies. Continuous variables were compared with Wilcoxon rank-sum tests, whereas categorical variables were compared with 2 tests. Time to coronary angiography and revascularization procedures were described with Kaplan-Meier techniques and compared by the log-rank statistic.
For our natural experiment, we sought to examine the relationship between weekend presentation and outcomes. In these analyses, we adjusted for a broad range of patient and hospital characteristics, including age, female gender, body mass index, white race, hypertension, diabetes, hypercholesterolemia, prior myocardial infarction, prior percutaneous coronary intervention (PCI), prior coronary artery bypass grafting, prior congestive heart failure, prior stroke, smoking status, renal insufficiency, family history of coronary artery disease, presence of ST-segment depression, presence of transient ST-segment elevation, positive cardiac markers, signs of congestive heart failure on admission, admission heart rate, admission systolic blood pressure, insurance status, total number of hospital beds at the treating facility, geographic region, teaching status, and attending physician specialty (cardiologist or noncardiologist). All of these covariates were included in the regression analyses, regardless of their statistical significance. Because patients within a hospital were more likely to be similar than patients from different hospitals, all adjusted analyses were performed with generalized estimating equation logistic models16 to account for correlations among clustered responses (eg, within-hospital correlations). All ORs are reported with weekday presentation as the reference group.
To more directly characterize the impact of timing of cardiac catheterization on in-hospital outcomes, we also used day of hospital presentation as the basis for an instrumental variable analysis. We first confirmed that day of hospital presentation was a valid instrumental variable by verifying that (1) it correlated with timing of cardiac catheterization, (2) it was independent of other patient characteristics, and (3) it was independent of patient outcomes except as mediated by timing of cardiac catheterization. To test the first assumption, we ran generalized estimating equation models to predict early catheterization as a function of day of hospital presentation (eg, weekend versus weekday). To test the second assumption, we compared patient characteristics between patients presenting on the weekend and patients presenting during the work week. Finally, to test the last assumption, we regressed in-hospital outcomes on the weekend/weekday variable, after adjusting for early catheterization along with other patient and hospital characteristics.
After confirming that weekend presentation satisfied these conditions, we then performed a fully adjusted instrumental variable analysis using a 2-stage linear regression approach.17,18 The first stage predicted the performance of early catheterization as a function of weekend (versus weekday) presentation and other patient and hospital characteristics. The predicted probabilities of early catheterization from this first stage were used as input to the second stage, which regressed in-hospital mortality on these predicted probabilities, along with the other patient and hospital characteristics. The effect of early catheterization on in-hospital mortality was estimated by subtracting the mean predicted mortality when early catheterization=1 from the mean predicted mortality where early catheterization=0. This value is equal to the regression coefficient for the predicted probability of early catheterization and represents an upper-bound estimate of the impact of early catheterization on in-hospital outcome. Separate instrumental variable analyses were performed with 4 alternative definitions of early catheterization: <12, <24, <36, and <48 hours after presentation. A probability value of <0.05 was established as the level of statistical significance for all tests. The instrumental variable analyses were performed with the IVREG command in STATA version 8.0 (STATA Corporation), and all other analyses were performed with SAS software (version 8.2, SAS Institute).
Results
Study Population
Between January 1, 2001, and September 30, 2003, a total of 77 760 patients with NSTE-ACS were entered into the CRUSADE database at participating study centers. After the exclusion of 21 408 patients because they were either enrolled at hospitals without PCI or surgical facilities (n=19 960), transferred out of the CRUSADE hospital (generally for performance of cardiac catheterization or revascularization; n=1334), or had missing information regarding day of admission (n=114), the remaining 56 352 patients from 310 centers composed our study population. All of these patients were enrolled at centers with full catheterization, PCI, and surgical capabilities.
Of the total study population, 10 804 presented between 5 PM on Friday and 7 AM on Sunday and were included in the weekend group, whereas the remaining 45 548 presented at other times and comprised the weekday group. Baseline characteristics of the 2 groups were generally well matched (Table 1). The weekend group tended to have slightly higher-risk profiles, as reflected by marginally higher PURSUIT (Platelet IIb/IIIa in Unstable angina: Receptor Suppression Using Integrilin Therapy) risk scores and a greater proportion of patients with positive cardiac markers. Medical care, including use of evidence-based therapies (antiplatelet agents, antithrombotic therapy, and -blockers) within the first 24 hours of admission, was also similar for the weekend and weekday groups (Table 2).
Cardiac Catheterization and Revascularization
Although the weekday and weekend groups were similar in their overall rates of in-hospital cardiac catheterization (73.0% versus 71.2%), PCI (42.0% versus 40.3%), and bypass surgery (13.4% versus 12.9%), the timing of these interventions was substantially delayed for weekend patients (Table 2). Median time (25th, 75th percentile) to cardiac catheterization (among patients who underwent the procedures) was 23.4 hours (10.4 to 45.3 hours) for weekday patients compared with 46.3 hours (21.4 to 63.5 hours) for weekend patients (P<0.0001). Moreover, the percentage of patients undergoing cardiac catheterization at any time in the first 48 hours was lower for the weekend group than for the weekday group (Figure 1). Similar findings were observed for PCI and bypass surgery (Figure 2).
In-Hospital Outcomes
The results of our natural experiment comparing in-hospital outcomes for the weekday and weekend groups are summarized in Table 3. There were no significant differences in rates of death, myocardial infarction (or reinfarction for patients whose initial diagnosis was NSTEMI), cardiogenic shock, congestive heart failure, or the composite between the 2 groups, with crude ORs ranging from 0.96 for reinfarction or stroke to 1.07 for cardiogenic shock or congestive heart failure. Given the comparable baseline characteristics for the weekday and weekend groups, risk adjustment had little effect on the ORs or their CIs, and none of the differences in risk-adjusted outcome were statistically significant (Table 3). The only difference in hospital outcome between the 2 groups was length of stay (Figure 3). The mean length of stay was significantly longer for the weekend group than for the weekday group (6.1 versus 5.8 days, P<0.0001).
Stratified analyses according to baseline patient characteristics are summarized in Figure 4. There were no significant differences for in-hospital mortality between the weekday and weekend groups according to age (<65 versus 65 years), gender, diabetes mellitus, or type of presentation (unstable angina versus NSTEMI). To investigate whether our results were sensitive to the definition of weekend and weekday presentation, we reanalyzed our data according to several alternative definitions (Table 4). Regardless of the specific definition of weekend and weekday presentation considered, the crude and adjusted ORs for in-hospital mortality were virtually identical to our base case analysis.
Instrumental Variable Analysis
The estimated impact of early catheterization on in-hospital mortality from the instrumental variable analysis is summarized in Table 5. For each definition of early catheterization, there was a nonsignificant trend toward lower mortality among patients receiving early catheterization than among those receiving later catheterization or no catheterization. This effect was largest (absolute risk reduction 1.9%; 95% CI 6.7% lower to 2.9% higher) when early catheterization was defined as being within 12 hours of presentation.
Discussion
Although recent consensus guidelines recommend a strategy of early invasive management for patients with high-risk acute coronary syndromes,2 few data exist on the optimal timing of cardiac catheterization and revascularization for such patients. In the United States, contemporary approaches may range from immediate angiography and revascularization (similar to the standard approach for ST-elevation MI) to medical stabilization with antiplatelet and antithrombotic therapy followed by angiography and revascularization in the next several days. In the present study, we used the related techniques of natural experimentation and instrumental variable estimation to provide insight into this important question from a large-scale, unselected patient population derived from the CRUSADE quality improvement initiative.
By stratifying patients according to date and time of initial presentation, we identified 2 patient subgroups that were comparable with respect to all observed characteristics except timing of invasive therapy, which was delayed by 24 hours among patients who presented on the weekend compared with those who presented during the work week. Despite this substantial delay, we were unable to identify any differences in adverse outcomes, including death or recurrent myocardial infarction, between the weekend and weekday subgroups. These findings were consistent across a range of subgroups, unaffected by conventional risk-adjustment, and insensitive to alternative definitions of "weekend presentation." Given the large sample size and broad patient population included in the CRUSADE initiative, these findings provide strong evidence that in-hospital outcomes would not be improved substantially by adopting a treatment strategy on weekends similar to our current approach during the work week, with reductions in median delays in time to cardiac catheterization from 48 to 24 hours.
The only previous study to rigorously address the question of optimal timing of revascularization therapy for patients with NSTE-ACS was the Intracoronary Stenting with Antithrombotic Regimen Cooling-Off (ISAR-COOL) trial.8 In that study, 410 patients with UA or NSTEMI received aggressive antithrombotic therapy with aspirin, clopidogrel, heparin, and tirofiban and were randomized to early intervention (within 6 hours) or delayed intervention (between 3 and 5 days). The principal finding of the study was that the early intervention strategy was associated with a 49% reduction in the risk of death or myocardial infarction at 30 days (95% CI 1% to 74%), all of which occurred during the pretreatment period. The investigators concluded that the duration of antithrombotic pretreatment should be as brief as possible to minimize the risk of ischemic complications in this patient population.
At first blush, the results of the present weekend versus weekday natural experiment would appear to be discordant with those of the ISAR-COOL trial. There are several major differences between the 2 studies, however, that may explain this apparent discrepancy. First, by design, the ISAR-COOL trial proscribed revascularization in the delayed group unless the patient experienced an ischemic complication, severe refractory angina, or hemodynamic instability. In the present study, both weekend and weekday patients could undergo intervention at any time according to the clinical judgment of their treating physicians. Indeed, even among weekend patients, 9% underwent angiography within 6 hours of presentation, and 17% underwent angiography within 24 hours. It is likely that these patients were selected because of ongoing ischemia or other characteristics that would suggest enhanced benefit from an early revascularization strategy.19
Second, the between-group differences in time to intervention were substantially greater in ISAR-COOL (median 2.6 versus 86 hours) than in the present study (median 23.4 versus 46.3 hours). It is possible that significant differences in outcome would have been observed in the present study had the extent of delay been similar. However, given its much larger sample size, our observational study had >80% power to detect a relative difference in mortality of 18% and 13% in death or myocardial infarction. Moreover, in the United States, it is unlikely that NSTE-ACS patients, in particular, those with evidence of recurrent or ongoing ischemia, would wait up to 5 days to undergo cardiac catheterization (as occurred in ISAR-COOL).
Finally, even though randomized clinical trials and natural experiments are alternative techniques for eliminating unmeasured confounding, these 2 study designs address fundamentally different questions. Randomized trials are designed to address a specific clinical question for an individual patient. In contrast, the present natural experiment estimates the average effect of a delay in performance of invasive therapy over the marginal range in treatment delays defined by the stratification variable.14 In the case of our natural experiment, the question we have addressed is, "What would be the effect of a policy that increased (or reduced) average invasive treatment delays by 24 hours for all patients with non-ST elevation acute coronary syndromes" Our finding that modest incremental treatment delays in the weekend group were not associated with an increase in adverse outcomes thus suggests that for the sizeable proportion of NSTE-ACS patients admitted to US hospitals on weekends, in-hospital outcomes would not be improved substantially by simply adopting a treatment strategy on weekends that is similar to our current approach during the work week.
Whether more dramatic reductions in existing delays in timing of invasive therapy (such as those achieved in ISAR-COOL) would result in meaningful improvements in in-hospital outcomes for our broad population is more directly addressed by our instrumental variable analysis but cannot be resolved definitively by the present study. Although our instrumental variable analysis did not demonstrate a significant reduction in in-hospital mortality associated with earlier cardiac catheterization, the CIs surrounding the absolute mortality reductions were relatively wide, particularly for catheterization within 12 hours of hospital admission. In fact, the central estimate of our instrumental variable analysis suggests that catheterization within 12 hours may be associated with an absolute 1.9% reduction in mortality and an even larger benefit if one considers the full 95% CI. This uncertainty reflects an inherent limitation of instrumental variables analyses in that they generally sacrifice statistical power in exchange for limiting confounding. Nonetheless, when taken together with the ISAR-COOL results, our findings provide additional support to the concept that if early catheterization is to achieve its maximal benefit for patients with NSTE-ACS, greater benefits may be associated with performance of catheterization within 12 hours of presentation.
Study Limitations
This study has several limitations. First, participation in CRUSADE is voluntary and thus may not be fully representative of the universe of NSTE-ACS patients. Nonetheless, the large number of participating centers and high rates of adverse outcomes compared with recent randomized clinical trials suggests that our population is relatively unselected, and thus, our findings should be applicable to the vast majority of such patients.
Second, we were only able to compare in-hospital outcomes. It is unknown whether longer-term mortality or late congestive heart failure was increased in the weekend group. Such an effect seems unlikely, however, because overall rates of revascularization were similar by hospital discharge, and there were no trends toward increased complications in the weekend group in the present study. Moreover, among patients managed aggressively (similar to the present study population), the majority of adverse events emerge by 7 days,3 close to the average length of stay for the present study. Thus, it is unlikely that a large number of adverse events would have occurred in a differential fashion beyond the scope of our assessment.
Third, our definition of postadmission myocardial infarction was based entirely on clinical signs and symptoms as detected during clinical care rather than routine surveillance of cardiac isoenzymes. Thus, it is possible that the present study could have failed to detect some episodes of asymptomatic cardiac myonecrosis. Of note, the rates of in-hospital death and myocardial infarction in the present study were, if anything, higher than those in most recent clinical trials of NSTE-ACS patients.3–5 Moreover, in the absence of differences in in-hospital mortality or other overt complications, the clinical significance of such asymptomatic events remains somewhat controversial.20,21
Finally, a critical assumption underlying this analysis is the validity of weekend presentation as an instrumental variable for evaluating delays in invasive therapy for patients with NSTE-ACS. As an instrumental variable, we assumed that weekend presentation would be associated with delayed invasive treatment but would not be correlated with any other differences in patient characteristics or treatment patterns (either observed or unobserved) that directly affect outcomes, an assumption whose validity can be explored but not proven definitively. Indeed, it is possible that weekend presentation is associated with other factors such as hospital staffing and unobserved practice patterns that contributed in part to our findings. Nonetheless, the fact that weekday and weekend patients were similar with respect to baseline clinical characteristics and acute medical therapies and that weekend status was not independently associated with in-hospital mortality after controlling for timing of invasive therapy supports the validity of this assumption.
Conclusions
In this observational study, we found that weekend versus weekday presentation represents a potentially useful instrumental variable for investigating the timing of invasive therapy for patients with NSTE-ACS. Although weekend presentation was associated with a delay of nearly 24 hours in the performance of cardiac catheterization and revascularization, in the context of contemporary medical therapy, there is no evidence that these delays are associated with an increase in adverse outcomes. These findings thus suggest that a policy of more liberal access to interventional therapy on weekends would not substantially improve population-wide outcomes compared with current US standards. Whether more dramatic reductions in across-the-board delays might result in meaningful improvements in in-hospital mortality cannot be definitively resolved by the present study, however, and is deserving of further examination in either a larger patient population or with longer follow-up.
Acknowledgments
CRUSADE is a National Quality Improvement Initiative of the Duke Clinical Research Institute. CRUSADE is funded by Schering-Plough Corporation. Bristol-Myers Squibb/Sanofi-Aventis Pharmaceuticals Partnership provides additional funding support. Millennium Pharmaceuticals, Inc, also funded the research presented here.
Disclosure
Dr Peterson received research grants from Bristol-Myers Squibb/Sanofi Pharmaceuticals, Schering-Plough, and Millennium Pharmaceuticals. Dr Roe has received research grants and honoraria from Millennium Pharmaceuticals, Schering-Plough, and Bristol-Myers Squibb/Sanofi-Aventis Pharmaceutical Partnership. Dr Ohman has received research grants from Millennium Pharmaceuticals, SPRI, Sanofi-Aventis, Berelex, and Bristol-Myers Squibb. Dr Cannon received research grant support from Bristol-Myers Squibb, Merck, Sanofi-Aventis, AstraZeneca, and Merck-Schering Plough, serves on the advisory boards for AstraZeneca, Bristol-Myers Squibb, GlaxoSmithKline, Merck, Merck/Schering-Plough Partnership, Pfizer, Sanofi-Aventis, Schering Plough, and Vertex, received lecture fees from AstraZeneca, Bristol-Myers Squibb, Guliford Pharmaceuticals, Merck, Millennium Pharmaceuticals, Pfizer, Sanofi-Aventis, and Schering Plough, and received honoraria from BestMed, I3 Magnifi, and NCME. Dr Normand is a member of the Massachusetts Cardiac Care Quality Commission and leads a data-coordinating center that examines the quality of PCI care. Dr Pollack serves on the Speakers Bureaus of Sanofi-Aventis, Schering-Plough, Millennium Pharmaceuticals, Genentech, and Pfizer, serves as a consultant for Sanofi-Aventis, Millennium Pharmaceuticals, Genentech, PDL, and The Medicines Company, received research support from Sanofi-Aventis, and owns a significant amount of stock in Pyng Medical.
Footnotes
Guest Editor for this article was Harvey D. White, DSc.
References
Yusuf S, Reddy S, unpuu S, Anand S. Global burden of cardiovascular diseases: part I: general considerations, the epidemiologic transition, risk factors, and impact of urbanization. Circulation. 2001; 104: 2746–2753.
Braunwald E, Antman E, Beasley J, Califf R, Cheitlin M, Hochman J, Jones R, Kereiakes D, Kupersmith J, Levin T, Pepine C, Schaeffer J, Smith E, Steward D, Theroux P, Alpert J, Eagle K, Faxon D, Fuster V, Gardner T, Gregoratos G, Russell R, Smith S. ACC/AHA guidelines for the management of patients with unstable angina and non-ST-segment elevation myocardial infarction: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on the Management of Patients with Unstable Angina). J Am Coll Cardiol. 2000; 36: 970–972.
Cannon CP, Weintraub WS, Demopoulos LA, Vicari R, Frey MJ, Lakkis N, Neumann FJ, Robertson DH, DeLucca PT, DiBattiste PM, Gibson CM, Braunwald E; TACTICS (Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy)–Thrombolysis in Myocardial Infarction 18 Investigators. Comparison of early invasive and conservative strategies in patients with unstable coronary syndromes treated with the glycoprotein IIb/IIIa inhibitor tirofiban. N Engl J Med. 2001; 344: 1879–1887.
Fox KA, Poole-Wilson PA, Henderson RA, Clayton TC, Chamberlain DA, Shaw TR, Wheatley DJ, Pocock SJ; Randomized Intervention Trial of unstable Angina Investigators. Interventional versus conservative treatment for patients with unstable angina or non-ST-elevation myocardial infarction: the British Heart Foundation RITA 3 randomised trial: Randomized Intervention Trial of unstable Angina. Lancet. 2002; 360: 743–751.
Wallentin L, Lagerqvist B, Husted S, Kontny F, Stahle E, Swahn E. Outcome at 1 year after an invasive compared with non-invasive strategy in unstable coronary-artery disease: the FRISC II invasive randomized trial. Lancet. 2000; 356: 9–16.
McCullough PA, Gibson CM, Dibattiste PM, Demopoulos LA, Murphy SA, Weintraub WS, Neumann FJ, Khanal S, Cannon CP; TACTICS-TIMI-18 Investigators. Timing of angiography and revascularization in acute coronary syndromes: an analysis of the TACTICS-TIMI-18 trial. J Interv Cardiol. 2004; 17: 1–5.
Montalescot G, Dabbous OH, Lim MJ, Flather MD, Mehta RH; Global Registry of Acute Coronary Events Investigators. Relation of timing of cardiac catheterization to outcomes in patients with non-ST-segment elevation myocardial infarction or unstable angina pectoris enrolled in the Multinational Global Registry of Acute Coronary Events. Am J Cardiol. 2005; 95: 1397–1403.
Neumann FJ, Kastrati A, Pogatsa-Murray G, Mehilli J, Bollwein H, Bestehorn HP, Schmitt C, Seyfarth M, Dirschinger J, Schomig A. Evaluation of prolonged antithrombotic pretreatment ("cooling off" strategy) before intervention in patients with unstable coronary syndromes: a randomized controlled trial. JAMA. 2003; 290: 1593–1599.
American Heart Association, Heart Disease and Stroke Statistics: 2004 Update. Dallas, Tex: American Heart Association; 2003.
Hoekstra JW, Pollack CV Jr, Roe MT, Peterson ED, Brindis R, Harrington RA, Christenson RH, Smith SC, Ohman EM, Gibler WB. Improving the care of patients with non-ST-elevation acute coronary syndromes in the emergency department: the CRUSADE initiative. Acad Emerg Med. 2002; 11: 1146–1155.
Braitman LE, Rosebaum PR. Rare outcomes, common treatments: analytic strategies using propensity scores. Ann Intern Med. 2002; 137: 693–696.
Creswell J. Research Design: Qualitative and Quantitative Approaches. London, UK: SAGE Publications; 1994.
Rosenbaum PR. Discussing hidden bias in observational studies. Ann Intern Med. 1991; 115: 901–905.
Gennetian L, Bos J, Morris P. Using instrumental variables to learn more from social policy experiments. MDRC Working Papers on Research Methodology.
McClellan M, McNeil B, Newhouse J. Does more intensive treatment of acute myocardial infarction in the elderly reduce morality Analysis using instrumental variables. JAMA. 1994; 272: 859–866.
Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986; 73: 13–22.
Greenland S. An introduction to instrumental variables for epidemiologists. Int J Epidemiol. 2000; 29: 772–720.
Eittner SL, Frank RG, Kessler RC. The impact of psychiatric disorders on labor market outcomes. Ind Labor Relations Rev. 1997; 51: 64–81.
Cannon CP. Evidence-based risk stratification to target therapies in acute coronary syndromes. Circulation. 2002; 106: 1588–1591.
The Clopidogrel in Unstable Angina to Prevent Recurrent Events Trial Investigators. Effects of clopidogrel in addition to aspirin in patients with acute coronary syndromes without ST-segment elevation. N Engl J Med. 2001; 345: 494–502.
Anderson KM, Califf RM, Stone GW, Neumann FJ, Montalescot G, Miller DP, Ferguson JJ III, Willerson JT, Weisman HF, Topol EJ. Long-term mortality benefit with abciximab in patients undergoing percutaneous coronary intervention. J Am Coll Cardiol. 2001; 37: 2059–2065.(Jason W. Ryan, MD, MPH; E)