Correspondence
Chest-Pain Observation Units
N Engl J Med 1999; 340:1596-1597May 20, 1999
- Article
To the Editor:
Farkouh et al. (Dec. 24 issue)1 have shown that patients who present to the emergency room with chest pain and who have an “intermediate risk” of cardiovascular events can be “safely” cared for with the use of a chest-pain observation unit (CPU). The study is timely because of the exorbitant costs associated with a diagnosis recorded as “rule out myocardial infarction.” The study is also important because it alerts clinicians to the ways in which the results of small, randomized trials can easily be misinterpreted.
The study compared primary end points (nonfatal myocardial infarction, death, congestive heart failure, stroke, and out-of-hospital cardiac arrest) in a group of patients assigned to admission to a CPU and a group assigned to routine hospital admission. At six months, primary end points had occurred in 8.5 percent of the hospital-admission group and in 6.6 percent of the CPU group (P=0.94). The authors conclude that “a CPU based in the emergency department can identify patients with intermediate-risk unstable angina who can be safely discharged to their homes and can save resources.”
With the increased surveillance of the patients assigned to routine hospital admission, it is not surprising that specific cardiovascular events were documented more frequently in this group than in the group of patients assigned to the CPU. In fact, the diagnosis of in-hospital myocardial infarction was made in 13 patients (6.1 percent) in the hospital-admission group and in 5 (2.4 percent) in the CPU group. If we assume that the diagnosis (and therefore treatment) of myocardial infarction may have been either missed or delayed in the CPU group, the critical question is, what was the consequence? If we look just at differences in mortality between the two strategies, assignment to the CPU was associated with three deaths at six months (overall mortality rate, 1.4 percent), whereas there were no deaths associated with assignment to routine hospital admission. Therefore, the authors' conclusion that the two strategies are equivalent in terms of safety is not legitimate.
1 ReferencesEdward O. McFalls, M.D., Ph.D.
Veterans Affairs Medical Center, Minneapolis, MN 554171
Farkouh ME ,Smars PA ,Reeder GS , et al. A clinical trial of a chest-pain observation unit for patients with unstable angina. N Engl J Med 1998;339:1882-1888
Full Text | Web of Science | Medline
To the Editor:
Farkouh et al. present interesting data on the safety and efficacy of a CPU for the triage of patients with acute chest pain. However, the interpretation of their findings and, more important, the application of their findings to clinical practice are hampered by the paucity of details about the treatment their patients received.
Although 50 percent of the patients had normal electrocardiographic findings at randomization, the combined event rate was 7.5 percent at six months. There were 14 cases of myocardial infarction, 4 cases of congestive heart failure, and no deaths in the hospital-admission group, as compared with 7 cases of myocardial infarction, 4 cases of congestive heart failure, and 3 deaths in the CPU group. The trend toward more myocardial infarctions among the admitted patients may be a result of chance, as the statistical analysis would suggest, but the trend also raises the possibility of differential surveillance during the initial hospitalization or perhaps an increase in small myocardial infarctions after revascularization procedures were performed. The difference in the number of deaths is also likely to be a statistical anomaly, but it would be important to know whether there were differences in treatment that might be associated with a difference in survival.
The trends for these outcomes accentuate the importance of knowing what medical treatment and other interventions were provided in the two groups of patients. We are told that the patients in the CPU were treated according to a protocol that included aspirin, heparin (if clinically warranted), and noninvasive stress testing. On the other hand, patients admitted to the hospital were not treated in a standard fashion. How many of these patients received aspirin or heparin? How many underwent noninvasive testing, coronary angiography, or revascularization? If the two groups were treated differently, one might expect different short-term, and long-term, outcomes. In addition, it would be interesting to know what treatment was received by the patients who were randomly assigned to the CPU and subsequently admitted to the hospital. Did they all undergo invasive evaluation? Although the authors' results are promising, they cannot be accurately used to determine the best mode of evaluation for patients with acute chest pain, and the best place to provide care, until all the interventions have been fully described.
William S. Getchell, M.D., M.P.H.
Greg Larsen, M.D.
Portland Veterans Affairs Medical Center, Portland, OR 97201Author/Editor Response
The authors reply:
To the Editor: Dr. McFalls raises a question about the safety of the CPU strategy, citing the six-month mortality rate of 1.4 percent in the CPU group, as compared with a rate of 0 percent in the hospital-admission group, a difference that was not significant. We agree that the sample size in our trial provided limited power to detect differences of 1 to 2 percent in overall mortality at six months. However, there was adequate power to detect the difference between 0 percent and 4.5 percent mortality or larger differences (type II error rate, 0.2) and to detect clinically relevant differences in the primary end points of the study. An 80 percent power to detect the difference between 0 percent and 1.4 percent would have required a sample of roughly 700 patients per group. Though small differences may be important, additional adverse outcomes need to be considered.
A more important consideration with respect to safety is the concern that patients sent home directly from the CPU after a short observation period might have subsequent adverse events. The patients in the CPU group who died, and indeed all patients with adverse events in this group, were admitted to the hospital from the CPU. No patient who was discharged home from the CPU after the observation protocol had an adverse event during the six-month follow-up period. We believe that the use of a treadmill exercise test (with a Duke University score of more than 4 considered to be a passing score)1 at the end of the observation period, followed by an outpatient cardiac evaluation within 72 hours after discharge, adequately safeguards against missing an early event. Other investigators have also shown that this approach provides valid prognostic information.1,2 Thus, we believe that the CPU strategy did not place the patients at increased risk.
Although the secondary outcomes of our study will be reported separately, we believe that the points raised by Drs. Getchell and Larsen should be addressed. Data on all encounters with physicians, stress tests, and invasive studies were collected for six months after randomization. In general, we found no significant differences in the rates of angiography (about 30 percent overall), coronary revascularization (about 15 percent overall), or subsequent cardiac hospitalization between the CPU group and the hospital-admission group. Patients who were randomly assigned to the CPU group and who subsequently required hospitalization accounted for the vast majority of studies carried out in the overall CPU group. The use of aspirin was part of the routine care provided in the emergency room. An examination of the excess myocardial infarctions in the hospital-admission group revealed no clear temporal relation between these events and the invasive cardiac procedures that were performed.
2 ReferencesPeter A. Smars, M.D.
Mayo Clinic, Rochester, MN 55905Michael E. Farkouh, M.D.
Mount Sinai Medical Center, New York, NY 10029Guy S. Reeder, M.D.
Mayo Clinic, Rochester, MN 559051
Mark DB ,Shaw L ,Harrell FE Jr , et al. Prognostic value of a treadmill exercise score in outpatients with suspected coronary artery disease. N Engl J Med 1991;325:849-853
Full Text | Web of Science | Medline2
Mark DB ,Hlatky MA ,Harrell FE Jr ,Lee KL ,Califf RM ,Pryor DB . Exercise treadmill score for predicting prognosis in coronary artery disease. Ann Intern Med 1987;106:793-800
Web of Science | Medline
