We performed a systematic review and meta-analysis of all clinical trials comparing routine invasive strategy with selective invasive strategy in patients with non–ST-segment elevation acute coronary syndrome. We did not find either strategy to offer an advantage over the other, even when we excluded trials that did not use coronary stents and glycoprotein IIb/IIIa inhibitors.
Every year, hospitals in the United States discharge more than 1 million patients who have received a diagnosis of non—ST-segment elevation (NSTE) acute coronary syndrome (ACS).1 Clinicians use 1 of 2 strategies to manage patients with NSTE-ACS: routine invasive strategy or selective invasive strategy. All patients managed with routine invasive strategy undergo coronary angiography, whereas selective invasive strategy employs coronary angiography only for those patients at high risk of having a subsequent cardiovascular event.2 Both strategies require that angiographic findings be used to guide any decision regarding revascularization.
Together, the American College of Cardiology and the American Heart Association have established guidelines to help clinicians manage patients with NSTE-ACS. The guidelines recommend using routine invasive strategy for patients with NSTE-ACS who have refractory ischemia; elevated serum cardiac enzyme levels; new ST-segment depression; or other high-risk features, as indicated by the patient’s history, admission characteristics, hospital course, or diagnostic testing.2 Results of clinical trials that compared these strategies were conflicting.3-12 We conducted this systematic review with the goal of determining whether one strategy produced better cardiovascular outcomes than the other in patients with NSTE-ACS.
Subjects and methods
We conducted an electronic search of all articles in the PubMed, EMBASE, and Cochrane Central Register of Controlled Trials databases as of September 18, 2007, for all relevant, randomized clinical trials published in English-language, peer-reviewed journals. The reference lists of all identified articles, along with the investigators’ personal files, were searched manually to determine whether there were any studies missing from the electronic search. We identified 4996 potentially relevant articles.
The meta-analysis included only those trials that randomized patients with NSTE-ACS to receive either routine invasive strategy or selective invasive strategy and reported death or nonfatal myocardial infarction (MI) as outcomes. Trials that enrolled patients who had stable angina, ST-segment elevation MI, or cardiogenic shock were excluded, along with trials in which all patients underwent coronary angiography. Our outcomes of interest comprised a composite of death or nonfatal MI, nonfatal MI alone, and death alone. We identified 10 trials that met our inclusion criteria.
Trial results were pooled according to the clinical outcomes reported at the maximum follow-up period. We also pooled the outcomes from all trials according to the patient’s status on discharge from the hospital, at 1-year follow-up, and at any follow-up point after 1 year. Any trial that did not report outcomes for a specified time interval was excluded from the analysis covering that time interval. We used a random effects model for meta-analysis in classical and Bayesian statistical frameworks.
To evaluate publication bias, we used the estimate of the effect size for each outcome measure from each trial against the inverse of the standard error to generate a funnel plot. The funnel plot was visually inspected for asymmetry and formally evaluated using the regression test by Egger and associates13 and Duval and Tweedie’s trim-and-fill method.14
An examination of the 4996 potentially relevant articles produced by our search identified 10 trials that met our inclusion criteria. These 10 trials enrolled a cumulative 10,648 patients.3-12 Of the 10 trials, 6 reported maximum follow-up results after 200059,12; the TIMI (Thrombolysis in Myocardial Infarction) IIIB trial was the earliest study and published its results in 1995.3 Most of the studies were conducted in North America3-5,10 or Europe,7-9,11,12 except for a small study of patients from South Africa and South Asia.6 The largest study was the FRISC II (Framingham and Fast Revascularization during Instability in Coronary Artery Disease) trial, which enrolled 2457 patients.9
All 10 studies required that patients demonstrate typical and consistent symptoms of unstable angina. There were differences between the trials regarding additional inclusion criteria, baseline characteristics of the patients, management protocols, and end point definitions for nonfatal MI (Table). The weighted mean average age of the patients (N = 10,648) was 62 years; 71% were men, 31% had a history of MI, and 16% had a diagnosis of diabetes mellitus. On presentation, 60% of patients demonstrated elevated cardiac enzymes, and 38.5% had ST-segment depression. The average follow-up period was 25 months (median, 16.5 months; range, 6-60 months). The routine invasive strategy group had a weighted average time of 48 hours from randomization to angiography.
The routine invasive cohort for the 10 studies cumulatively reported a composite outcome of death or nonfatal MI for 847 patients, an outcome of death for 438 patients, and the outcome of nonfatal MI for 490 patients. The selective invasive strategy cohort for the 10 studies cumulatively documented 928 patients with a composite outcome of death or nonfatal MI, 463 deaths, and 569 patients with nonfatal MI. Pooling the relative risks (RRs) for the combined outcome of death or nonfatal MI did not find that either strategy offered greater benefit than the other (RR, 0.90; 95% confidence interval [CI], 0.74-1.08). Similarly, pooling the RRs for the individual outcomes of death (RR, 0.95; 95% CI, 0.80-1.14) and nonfatal MI (RR, 0.86; 95% CI, 0.68-1.08) did not find an advantage to using one strategy over the other (Figure). When we pooled outcomes reported at the time of discharge, selective invasive strategy appeared to offer a modest advantage, but the advantage was not significant and disappeared after discharge. Routine invasive strategy appeared modestly beneficial at 1 year and beyond 1 year of follow-up, but this also was not statistically significant.
For sensitivity analysis, we performed the above meta-analyses in the Bayesian statistical framework and found similar results (Figure). We conducted additional sensitivity analyses by removing 1 clinical trial at a time from the analysis and recalculating the summary RR for each outcome. Removing the ICTUS (Invasive versus Conservative Treatment in Unstable Coronary Syndromes) trial from the analysis resulted in a significant change to the RR for the composite outcome (RR, 0.84; 95% CI, 0.74-0.97), and nonfatal MI (RR, 0.77; 95% CI, 0.68-0.88), favoring routine invasive strategy. It did not significantly affect the RR for the outcome of death (RR, 0.93; 95% CI, 0.76-1.14). The individual exclusion of other trials had no effect on the outcomes of the meta-analyses. To evaluate the effects of advances in coronary intervention on treating patients with NSTE-ACS, we conducted additional meta-analyses excluding the 3 trials that did not use glycoprotein IIb/IIIa inhibitors or coronary stents.3,4,10 Our results did not change for the outcomes of death (RR, 0.88; 95% CI, 0.72-1.07), nonfatal MI (RR, 0.82; 95% CI, 0.57-1.17), or the composite outcome of death or nonfatal MI (RR, 0.84; 95% CI, 0.63-1.10).
To assess possible publication bias, we generated a funnel plot, which was asymmetrical. The trim-and-fill method showed that 1 study needed to be imputed to make this plot symmetrical; adding the imputed study, however, did not change the results of this meta-analysis. Similarly, Egger’s regression method did not find statistically significant publication bias (intercept, -0.55; 95% CI, -3.1 to 2.0; P = .63).
The meta-analysis we conducted did not find sufficient evidence to support using routine invasive strategy to manage patients with NSTE-ACS. Results remained unchanged whether the statistical framework of the meta-analysis was classical or Bayesian. The lack of evidence favoring one strategy over another suggests that the strategies are equivalent, but such an assumption may be false. It is clear from the 95% CI that the data cannot rule out whether either strategy offers a clinically significant advantage.
Our study updates the results of 4 recent meta-analyses, all of which found a significant advantage for using routine invasive strategy over selective invasive strategy.15-18 Our results differ from these meta-analyses either because of the inclusion of new trials or the inclusion of longer follow-up data reported after the 4 meta-analyses were published. In addition, 2 of the 4 meta-analyses15,16 included the ISAR-COOL (Intracoronary Stenting with Antithrombotic Regimen Cooling-off) 19 trial, which we excluded because almost all the patients in ISAR-COOL underwent angiography within 5 days of randomization; therefore, it did not have a selective invasive strategy arm, as defined for our meta-analysis.
This meta-analysis has important clinical and research implications. We did not find either of the 2 strategies to be superior, suggesting that other factors may need to play a greater role in choosing a treatment strategy for a particular patient. These factors might include applicability of the evidence to the specific patient, feasibility of employing the intervention in a particular setting, risk-benefit ratio for the patient, and the patient’s values and preferences.20 This meta-analysis also provides grounds for future clinical trials to identify the subset of patients who are most likely to benefit from a routine invasive strategy.
Pooled results of the available clinical trial data for patients with NSTE-ACS who received treatment with either a routine invasive strategy or selective invasive strategy did not identify any advantage of one approach over the other, although the direction of effect seems to favor routine invasive strategy. When choosing between the 2 strategies for a particular patient, clinicians must consider patient preferences, local expertise, and characteristics of the health care system.
The authors have no relationship with any commercial entity that might represent a conflict of interest with the content of this article.