Incidence and predictors of myocardial infarction among patients with atrial fibrillation

Patient group

Between January 1998 and January 1999, we prospectively identified and followed all patients with AF who presented to the emergency department (ED) of the Beth Israel Deaconess Medical Center. Demographic and clinical data were collected prospectively by a trained research nurse using standardized case report forms. Additional information collected included treatments performed in the ED (antiarrhythmic therapy, cardioversion), patient disposition (admit, discharge, indication) and clinical course once admitted.

During the study period, 255 patients presented to the ED with a primary diagnosis of new onset, recurrent (i.e., paroxysmal) or chronic AF (Fig. 1). Sixty-five patients were discharged directly from the ED, and 190 patients were admitted to the hospital. Of these 190 patients, 109 (57%) underwent a standard “rule-out MI” protocol at the discretion of the primary treating physician. These patients formed the primary analytic cohort for this study.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Diagram of consecutive patients who presented to the ED with AF over a one-year period.

Classification of symptoms

Symptoms, including chest pain, dyspnea and palpitations, were recorded as described by the patient to the ED physician. Because of difficulties in defining and enforcing standard definitions, we did not attempt to further specify the quality or severity of these symptoms.

Electrocardiographic classification

All 12-lead electrocardiograms (ECGs) were independently analyzed by two cardiologists (M.E.J. and P.Z.). The extent of ST segment elevation or depression was measured (to the nearest millimeter) in the leads with maximal depression or elevation. ST segment changes were measured at the maximal point of depression or elevation beginning at 80 ms after the QRS complex. To maximize the sensitivity of the ECG for the detection of myocardial ischemia, we did not attempt to differentiate potentially nonspecific ST segment depressions or depressions associated with left ventricular hypertrophy from more “ischemic-appearing” ST segment depressions. We prospectively defined major ST segment changes as any ST segment elevation or depression >2 mm. Minor ST segment changes were defined as ST segment depression ≤2 mm. Left ventricular hypertrophy was defined according to the criteria of Romhilt and Estes (4).

Determination of MI

The standard rule-out MI protocol included three serial creatine phosphokinase, MB isoenzyme (CPK-MB) determinations during a minimal 24-h admission to a medical floor with full disclosure telemetery. A patient was considered to have experienced a MI if any CPK-MB measurement was greater than two times the upper limit of normal (i.e., >20 IU/ml) (5).

Statistical analysis

Continuous data are described as the mean value ± SD, and categoric data are described as frequencies. Confidence limits were calculated using exact techniques. The association between the diagnosis of MI and specific clinical or electrophysiologic characteristics was evaluated using the t test for continuous variables or the Fisher exact test for categoric variables. A p value <0.05 was considered statistically significant.

Incidence of MI

This study represents a comprehensive evaluation of all patients with AF who presented to a large urban hospital over a one-year period. In the subgroup of patients who underwent the rule-out MI protocol, we found that the incidence of MI at the time of admission was relatively low (6 [5.5%] of 109 patients). If those patients who were discharged or did not undergo systematic monitoring of cardiac isoenzymes, but had a benign subsequent clinical course, were also considered to have “ruled-out” for MI, the incidence of clinically significant MI in our AF group was only 2.3% (6 of 255 patients).

The relatively low incidence of MI in this patient group is not surprising. Previous studies have reported the incidence of MI in patients with new-onset AF to be 11% (6). In our population-based series, the incidence of documented MI was even lower, in part because only 69% of patients who underwent evaluation for MI had either chest pain or some evidence of myocardial ischemia on their admission ECG. This reflects the common practice in our institution (and many others) of admitting patients with AF to monitor for possible MI, even in the absence of any supporting signs or symptoms. Our study does not support this practice, however. Among the 31% of patients without chest pain or ST segment deviation on their admission ECG, there were no confirmed MIs.

Predictors of MI

Despite the low incidence of MI in our group, we identified several objective clinical and ECG features whose absence could be helpful in reliably identifying a subgroup of patients at extremely low risk of MI. These features include new-onset AF, the presence of chest pain and the presence of ST segment depression or elevation on the admission ECG. Although each of these factors was sensitive for the occurrence of MI, their value as potential discriminating factors for deciding which patients to admit and rule out MI and which to consider for less intensive management was limited by their low specificities (50% to 65%), and thus low positive predictive values (11% to 14%). In particular, the value of chest pain as a predictor of the occurrence of MI in this patient group was poor, with a positive predictive value of only 14% (95% CI 5% to 29%). This finding is consistent with previous studies of AF in which chest pain was a relatively common but nonspecific finding (7,8).

The most sensitive and specific predictor of MI was the presence of major ST segment deviation—defined as >2 mm of ST segment depression or any ST segment elevation—on the presenting ECG. In our patient group, this single factor was an excellent discriminator of patients with and without MI, with a positive predictive value of 86% (95% CI 42% to 99%) and a negative predictive value of 100% (95% CI 96% to 100%). This finding suggests that the vast majority of patients admitted to the hospital for evaluation of a possible MI might be managed equally well in a less intensive setting—either on an unmonitored medical unit or in the outpatient setting. In fact, use of this criterion would have reduced the number of MI evaluations in our group by 94%, with no loss of sensitivity. Even the adoption of less stringent ECG criteria, such as the presence of any degree of ST segment deviation, would have identified all of the patients with an acute MI in our group, while allowing for 46% to be managed more conservatively.

Although evaluation for a possible MI was cited frequently by the primary physician as an indication for hospital admission in our AF group, it is uncertain whether better risk stratification would substantially alter this practice and the high rate of hospital admission for these patients. In fact, most patients in our rule-out MI group had more than one indication for hospital admission, and evaluation for ongoing myocardial ischemia or infarction was cited as the sole indication for admission in only 21%. Nonetheless, many of the other indications for admission, including rate control, cardioversion and initiation of anticoagulation, can be safely performed on an outpatient basis with appropriate patient education, close follow-up and ancillary support (9). In contrast, ongoing myocardial ischemia or infarction is generally an “absolute” indication for hospital admission and close observation (10). Thus, we believe that more precise identification of patients with AF who are at risk for MI could substantially reduce the rate of hospital admission without affecting overall outcomes or quality of care.

Study limitations

Our study has several important limitations. The major limitation of our study is that the diagnostic thresholds we identified for prediction of MI among patients with AF were selected on a post-hoc basis. Accordingly, our findings should be viewed as hypothesis-generating and will require validation in an independent data set before their widespread implementation in clinical practice. Second, given the limited number of MIs in our patient group, the confidence limits for our estimates of sensitivity (and positive predictive value) are wide. Nonetheless, the major role of our diagnostic criteria is in identifying patients at low risk for MI who do not necessarily require hospital admission if otherwise stable. The value of these criteria is thus related mainly to their negative predictive values, for which the confidence limits are substantially more narrow. Finally, the fact that our group was drawn from a single, urban, academic medical center may also limit the generalizability of our findings to other settings and patient groups.

Conclusions

Despite the common practice of admitting patients with AF to the hospital to “rule out MI,” the incidence of MI in such patients without other evidence of myocardial ischemia is low. The presence of major ST segment abnormalities on the presenting ECG was both a sensitive and specific predictor of MI in our patients. Use of this criterion (or other related factors) must be prospectively validated, but may avoid unnecessary testing and hospital admissions and result in important cost savings without loss of quality of care.