Myocardial Perfusion Imaging for Evaluation and Triage of Patients With Suspected Acute Cardiac Ischemia: A Randomized Controlled Trial

Context Observational studies of acute myocardial perfusion imaging in emergency department (ED) patients with chest pain have suggested high sensitivity and negative predictive value for acute cardiac ischemia, but use of this method has not been prospectively tested.

Objective To assess whether incorporating acute resting perfusion imaging into an ED evaluation strategy for patients with suspected acute ischemia but no initial electrocardiogram (ECG) changes diagnostic of acute ischemia improves clinical decision making for initial ED triage.

Design, Setting, and Patients Prospective, randomized controlled trial conducted at 7 academic medical centers and community hospitals between July 1997 and May 1999 among 2475 adult ED patients with chest pain or other symptoms suggestive of acute cardiac ischemia and with normal or nondiagnostic initial ECG results.

Intervention Patients were randomly assigned to receive either the usual ED evaluation strategy (n = 1260) or the usual strategy supplemented with results from acute resting myocardial perfusion imaging using single-photon emission computed tomography with injection of 20 to 30 mCi of Tc-99m sestamibi (n = 1215), interpreted in real time by local staff physicians and with results provided to the ED physician for incorporation into clinical decision making.

Main Outcome Measure Appropriateness of triage decision either to admit to hospital/observation or to discharge directly home from the ED.

Results Among patients with acute cardiac ischemia (ie, acute myocardial infarction [MI] or unstable angina; n = 329), there were no differences in ED triage decisions between those receiving standard evaluation and those whose evaluation was supplemented by a sestamibi scan. Among patients with acute MI (n = 56), 97% vs 96% were hospitalized (relative risk [RR], 1.00; 95% confidence interval [CI], 0.89-1.12), and among those with unstable angina (n = 273), 83% vs 81% were hospitalized (RR, 0.98; 95% CI, 0.87-1.10). However, among patients without acute cardiac ischemia (n = 2146), hospitalization was 52% with usual care vs 42% with sestamibi imaging (RR, 0.84; 95% CI, 0.77-0.92).

Conclusions Sestamibi perfusion imaging improves ED triage decision making for patients with symptoms suggestive of acute cardiac ischemia without obvious abnormalities on initial ECG. In this study, unnecessary hospitalizations were reduced among patients without acute ischemia, without reducing appropriate admission for patients with acute ischemia.

Each year in the United States, more than 6 million patients present to emergency departments (EDs) with chest pain or other symptoms suggestive of acute cardiac ischemia (ie, either acute myocardial infarction [MI] or unstable angina pectoris).¹ The majority of these patients are admitted to the hospital or to an observation unit, because the initial clinical examination, electrocardiogram (ECG) results, and initial cardiac enzyme levels are insufficient to eliminate the possibility of acute infarction or unstable angina.^1-4 Nevertheless, most patients without obvious ischemic ECG changes who are hospitalized or observed in special units ultimately prove not to have acute ischemia.^2,4-6 Moreover, among patients discharged directly from the ED, an important minority actually have acute ischemia, leading to unfavorable outcomes.^6-8

Since the predominant pathophysiology of acute cardiac ischemia is a reduction in coronary blood flow, myocardial perfusion imaging in the ED setting is conceptually attractive. The feasibility of imaging in this setting was established over 20 years ago using thallium 201,⁹ but its use in the real-time ED setting was impractical. The logistics of the use of newer technetium Tc 99m–based perfusion agents are more suitable for use in the ED.¹⁰ The technetium Tc 99m label is associated with generally higher-quality images, and there is minimal redistribution after initial injection. Thus, imaging may occur at a later time point after injection, with the resulting image reflecting myocardial blood flow at the time of injection. Several observational studies have demonstrated a very high negative predictive value of acute perfusion imaging using technetium Tc 99m–labeled isotopes in confirming the absence of acute infarction or subsequent cardiac events.^11-15 No previous study has prospectively tested the impact of an imaging vs a nonimaging strategy.¹⁶ Therefore, we conducted a randomized trial to determine whether incorporating myocardial perfusion imaging into an ED evaluation protocol for patients with symptoms suggestive of acute cardiac ischemia would improve clinical decision making.

This trial was designed to compare the clinical effectiveness of 2 ED evaluation strategies for patients suspected of having acute cardiac ischemia but without diagnostic initial ECG changes of acute ischemia: the usual care strategy (the standard clinical evaluation strategy used in each hospital's ED), and the scan strategy (the same standard clinical evaluation strategy but also incorporating results derived from immediate resting single-photon emission computed tomography [SPECT] myocardial perfusion imaging with technetium Tc 99m sestamibi). The study protocol was approved by the institutional review boards of all participating centers.

Between July 1997 and May 1999, all potentially eligible patients (defined as any initial clinical suspicion of acute ischemia) were screened at the EDs of 7 diverse hospitals during daytime hours. Eligible ED patients who consented were immediately randomized, using simple unrestricted randomization, by means of a central computerized telephone randomization system. Screened patients who were ineligible, and eligible patients who did not consent to participate, were entered anonymously into a registry for comparison with randomized patients.

For patients randomized to the scan strategy, injection of 20 to 30 mCi of technetium Tc 99m sestamibi (DuPont Pharmaceuticals, Wilmington, Del) was performed in the ED, generally by a nuclear medicine technologist, as soon as possible after randomization. SPECT imaging was performed beginning 30 to 60 minutes later in the nuclear cardiology laboratory, with images acquired in the gated SPECT mode for simultaneous evaluation of resting perfusion and systolic left ventricular function.¹⁷ The images were interpreted on site in real time by the appropriate staff physicians at each site, with results reported immediately to the responsible physician. Based on all available results, a triage decision was made by the physician to hospitalize the patient (to an observation unit, ward, telemetry unit, or cardiac care unit [CCU]) or to discharge the patient directly home. The decision-making physician was not directed by protocol to admit or discharge the patient based on certain scan results; rather, imaging results were given to the physicians who then incorporated the data into the information base they used to make a clinical decision regarding their patient's disposition.

Included were all patients 30 years of age or older (unless there was a history of recent cocaine use, in which case they were required to be 18 years or older) who had chest pain or any other symptoms suggestive of acute cardiac ischemia (eg, shortness of breath) and, according to the evaluating ED physician, an ECG either normal or nondiagnostic for acute ischemia or infarction. To be included, patients' symptoms were required to be ongoing or to have resolved no longer than 3 hours prior to consent. Excluded were patients with a history of MI, since such patients will most often have an abnormal resting perfusion pattern. Written informed consent was obtained from all patients.

Gated SPECT images were obtained using acquisition and processing parameters standard at each study site. Images were classified as normal (no obvious perfusion abnormalities with normal regional and global ventricular function), abnormal (definite perfusion abnormality and/or regional or global function), or equivocal (images that did not clearly fit into the other categories). Physicians were advised to consider equivocal results as mildly abnormal, based on a previous study showing that patients with equivocal scans in this setting have a slightly higher event rate than patients with normal scans.¹²

For characterization of final diagnosis (acute cardiac ischemia or not), all patients had follow-up ECGs, measurement of cardiac enzyme levels, and protocol-specified follow-up stress testing with perfusion or echocardiographic imaging. For patients initially admitted, this was usually accomplished during the observation or hospitalization period. For all ED patients initially discharged directly home, a return visit to the study site 24 to 36 hours later was made for follow-up biomarker measurements, ECGs, and stress testing. The confirmed diagnosis was assigned by the site principal investigator based on all available data, including enzymatic, ECG, and stress testing data, as well as cardiac catheterization data when available; principal investigators were blinded to the randomization assignment and to initial scan result for patients randomized to the scan strategy. For all patients whose final diagnosis was MI, for the majority of patients whose final diagnosis was unstable angina, and for an equal and randomly chosen group of patients whose final diagnosis was not acute ischemia, medical records were reviewed at the coordinating center by an independent investigator who was blinded to the original confirmed diagnosis assignment. This independent assignment of confirmed diagnosis was concordant in 98% of cases. Patients were contacted at 30 days following initial ED presentation to determine vital status and the occurrence of other cardiac events and procedures; follow-up was 99% complete.

The primary end point of the trial was the appropriateness of initial ED triage decision, assuming that patients with acute infarction or unstable angina should have been hospitalized (to a CCU, telemetry unit, ward, or chest pain unit), and that patients without acute ischemia (ie, those with negative serial enzyme test results, no evolutionary ECG changes on serial testing, and a negative follow-up stress test result) ideally should not have been hospitalized or admitted for observation.

The study was designed to have 80% power to detect a reduction (10% to 3%) in ED discharges to home for patients ultimately diagnosed with acute cardiac ischemia, and to have 80% power to detect a reduction in unnecessary admissions for patients without acute ischemia from a 45% unnecessary admission rate to 39%, ie, an absolute risk reduction of 6% and a relative risk reduction of 15%. A data and safety monitoring board reviewed results at an interim time point, to ensure that triage was not unfavorably affected by the intervention.

Patient characteristics were compared using the t test for continuous variables, and the Pearson χ² test or Fisher exact test for dichotomous variables. P<.05 was used to determine statistical significance. Trial data were analyzed based on an intent-to-treat strategy. The effect of imaging on ED triage was analyzed by computing the relative risk (RR) and the 95% confidence intervals (CIs) for hospitalization vs discharge to home, using the Cochran-Mantel-Haenszel method to adjust for hospital differences. Odds ratios for the effect of imaging on admission were computed for different subgroups using logistic regression analyses, including terms for the treatment effect, subgroup effect, the interaction between treatment and subgroup, and terms for hospital effects to adjust for possible hospital differences for factors that were not directly hospital related (ie, hospital volume). Possible interactions between patient and site subgroups and the effect of the scan strategy on admission rate were analyzed using Wald χ² testing from results of logistic regression analysis. With the exception of the site stratifications, all P values were adjusted for site effects. To facilitate interpretation of these results, the adjusted odds ratios from these analyses were converted to RRs using the method of Zhang and Yu,¹⁸ in which the strata-specific admission rates for the nonacute cardiac ischemia usual care groups were used as the control prevalence. All analyses were performed using SAS v8.0 (SAS Institute Inc, Cary, NC).

During 20 months of recruitment, 7955 patients were screened for inclusion (Figure 1), on the basis of presenting symptoms suspicious for acute ischemia. Of these, 2908 consecutive patients met all study eligibility criteria and were eligible for consent, of whom 2475 patients (85%) consented to participate and were randomized to 1 of the 2 ED evaluation strategies. Of the 5047 protocol- or consent-ineligible patients, exclusion was most often because of either an ECG diagnostic for acute ischemia or infarction (34%) or a history of prior infarction (27%). Other less common reasons for ineligibility included: resolution of symptoms more than 3 hours prior to presentation (10%), patient became unstable in the ED (4%), evaluating physician or primary physician refused enrollment (3%), patient previously enrolled in the study (2%), and patient unable to sign informed consent appropriately (2%). Among eligible patients, those who were protocol-eligible but did not sign informed consent were older than those consenting (mean [SD] age, 57 [16] vs 53 [14] years; P = .001) and were slightly more often women (55% vs 49%; P = .03).

Among randomized patients, 1260 were randomized to usual care and 1215 were randomized to the strategy incorporating sestamibi perfusion imaging. Baseline characteristics for the groups are shown in Table 1. Other than the slight difference in proportion of women between groups, there were no significant clinical or demographic differences at baseline.

Overall confirmed diagnoses of acute cardiac ischemia were made in 13% of patients: 2% had acute MI, while 11% had unstable angina. Diagnoses other than acute cardiac ischemia were confirmed for 87% overall of patients in each group (Table 1).

Patients randomized to the scan strategy had a lower rate of hospital admission than patients randomized to usual care (47.5% vs 56.1%, respectively; RR, 0.87; 95% CI, 0.81-0.93; P<.001), and a higher rate of direct discharge home from the ED (Table 2). The increase in direct discharges home from the ED in the scan strategy group was predominantly due to a reduction in admission to a telemetry unit or ward, or to a chest pain unit. Despite fewer admissions in the group randomized to the imaging strategy, there were no outcome differences 30 days after ED presentation between patients receiving usual ED care and those receiving sestamibi imaging (Table 3). Thus, similar outcomes were observed with fewer hospital admissions in the group randomized to the imaging strategy.

The median time from ED presentation to admission or discharge home for patients in the usual care group was 4.7 hours (interquartile range, 3.4-6.4 hours), and for those randomized to the scan strategy, 5.3 hours (interquartile range, 4.0-7.0 hours) (P<.001).

The impact of ED sestamibi imaging on hospitalization for patients with acute cardiac ischemia is shown in Table 4. Among all patients with acute ischemia, there was no difference in the appropriate hospitalization rate of approximately 85%. For patients with acute infarction, there was also no difference in the hospitalization rate, appropriately high at 96% overall: one such patient in each group was inappropriately sent home from the ED. Follow-up testing showed the inappropriately discharged patient in the usual care group to have an inferior wall infarction with an inferior wall motion abnormality. The inappropriately discharged patient in the sestamibi group had a normal sestamibi scan in the ED, as well as normal results of echocardiography and follow-up stress perfusion testing, but had positive enzyme test results on follow-up.

Among those with unstable angina, there also was no difference in the appropriate admission rate between the 2 groups, nor were there differences in specific in-hospital locations for admitted ED patients (Table 4).

Among the 2146 patients whose final confirmed diagnosis was not acute cardiac ischemia, 52% of patients randomized to usual care were hospitalized, who, retrospectively, could be classified as unnecessary admissions (Table 5). In the sestamibi scan group, this unnecessary admission rate was reduced to 42%, a 10% absolute reduction, and a 20% relative change (RR, 0.84, 95% CI, 0.77-0.92, P<.001). Results were similar when the location of admission triage from the ED to hospital was subcategorized as CCU, telemetry ward, or chest pain unit.

The reduction in hospitalization of patients without acute ischemia who had sestamibi scans occurred at 6 of the 7 study hospitals, the exception being 1 site with a very low baseline admission rate that followed a prolonged evaluation process.

As shown in Table 6, perfusion imaging increased direct ED discharges to home in patients without acute ischemia in both men and women, in both younger and older patients, and in both those with and without risk factors for coronary disease. The effect of incorporating perfusion imaging into the evaluation strategy on increasing direct ED discharges to home was similar at sites with an established chest pain center protocol compared with those sites without such a protocol, and was also similar in patients whose initial evaluation included a troponin I value compared with those without an initial value. There were trends toward greater effect of imaging for patients without a history of coronary disease and in those with chest pain as their primary presenting symptom.

Among the 1215 patients randomized to the scan strategy, the imaging results were related to the risk of an adverse outcome. Among patients with normal, equivocal, or abnormal scan results, the risk of acute MI was 0.6%, 0.8%, and 10.3%, respectively (RR for equivocal or abnormal vs normal scan, 6.61; 95% CI, 2.47-17.67; P<.001). A similar relationship was observed between scan findings and any cardiovascular event at 30 days, defined as acute MI, death, or revascularization. Among patients with normal, equivocal, and abnormal scan results, the risk of a cardiac event defined as such was 3.0%, 6.1%, and 20.5%, respectively (RR for equivocal or abnormal vs normal scan, 3.83; 95% CI, 2.36-6.21; P<.001).

This trial indicates that incorporating results from acute resting perfusion imaging with technetium Tc 99m sestamibi into an ED evaluation strategy for patients with suspected acute cardiac ischemia improves triage decision making. Specifically, unnecessary hospitalization of patients without acute cardiac ischemia was reduced, without affecting appropriate hospitalization of patients with acute ischemia. Accordingly, the reduction in hospital admission related to the use of perfusion imaging was associated with unchanged 30-day clinical outcomes.

Although several observational studies^11-15,19 supported the incorporation of perfusion imaging into an ED evaluation strategy, these studies examined the prognostic value of imaging results but did not examine the impact of imaging on ED triage decision making. Thus, this "effectiveness" trial was designed to determine how imaging would affect ED triage decisions when incorporated into usual care at a diverse group of hospitals, including academic medical centers, urban public teaching hospitals, and moderate to large community hospitals. The inclusion criteria were designed to capture all low- to moderate-risk patients with possible acute cardiac ischemia in whom the triage decision regarding admission or direct discharge to home is often difficult.^1-3 At each site, the SPECT perfusion images were read on site and in real time by the usual interpreting physician.

The rates of acute infarction (2%) and unstable angina (12%) among study patients are lower than those observed in previous studies evaluating imaging or other technologies in this setting.^12-15 This is likely a result of the broader inclusion criteria of this trial, allowing inclusion of patients with any symptoms suggestive of acute ischemia. Other studies with higher proportions of patients who prove to have acute infarction have used narrower inclusion criteria, studying only patients with angina-like chest pain and nondiagnostic ECGs,¹² or patients for whom the decision to hospitalize had already been made.¹⁵ The inclusion criteria in this trial were designed so that trial results would be applicable to the broad array of ED patients presenting with possible acute ischemia.

Few ED diagnostic technologies or evaluation strategies for acute ischemia have been rigorously tested in controlled clinical trials.¹ Predictive instruments for acute cardiac ischemia or infarction have improved triage effectiveness in some trials^20-22 but not others.²³ Controlled trials have also addressed the use of an "accelerated diagnostic protocol" or a chest pain evaluation unit as a means to create uniform, efficient evaluations for such patients, and have shown shorter lengths of stay, fewer standard hospital admissions, and lower resource utilization with similar outcomes.^24,25

In these studies of accelerated protocols or observation units, the majority of patients were ultimately diagnosed as not having acute ischemia.^24,25 The results of the present trial suggest that incorporating perfusion imaging into such ED evaluation strategies should afford the possibility of determining the presence or absence of acute ischemia even earlier in the ED evaluation process, facilitating an earlier triage decision and diminishing the need for observation in a chest pain unit or a similar pathway. In this regard, Kontos et al¹⁴ reported that resting perfusion imaging with sestamibi was 92% sensitive for detecting acute infarction in studies performed early in the ED, whereas initial measurement of troponin I levels in the ED had only 39% sensitivity. In that study, the sensitivity of troponin I levels became similar to that of sestamibi imaging within the first 24 hours,^14,26 consistent with its known release kinetics after myocardial injury.²⁷ Others reported that optimal sensitivity and prognostic value of levels of troponins T or I are not achieved until 18 hours after symptom onset.^28,29

The finding that 42% of patients ultimately found to be without acute ischemia in the scan group were admitted to the hospital suggests that even further improvement in ED triage decision-making is possible. Such improvement may result from further confidence in imaging results with increasing familiarity, or may require a combination of imaging results with other probability estimates. Future studies should address which populations benefit most from incorporating imaging into the evaluation process, and whether there is a threshold of probability, as might be provided by a predictive instrument, above or below which imaging or any other technology is not needed. Patients who had a perfusion image performed but were still admitted may also have had false-positive images, potentially resulting from diaphragmatic or breast attenuation, or motion artifact.

Whether the data from this trial can be applied to smaller or more rural hospitals is not clear, as such institutions were not represented in this trial. However, the trial results are potentially generalizable to hospitals with on-site imaging facilities, since hospitals with little or no experience in using perfusion imaging in the ED demonstrated a similarly favorable triage effect as those with some experience (Table 5). The safety of increasing direct discharges to home from the ED by incorporation of a resting perfusion imaging strategy as seen in the current study presumes incorporation of a follow-up evaluation after ED discharge. In the current study, follow-up stress testing was directed per protocol. This study was conducted during expanded daytime hours, and thus whether the data generalize to use during nighttime hours cannot be determined with certainty.

In summary, incorporating acute resting sestamibi myocardial perfusion imaging into an ED evaluation strategy for patients with symptoms suggestive of acute cardiac ischemia reduced unnecessary hospitalizations among patients without acute ischemia (ie, improved the specificity of the admitting decision), without reducing appropriate admission of patients with acute ischemia (maintained sensitivity), thereby improving the overall clinical effectiveness of the ED triage process.