The value of the total leukocyte count for the diagnosis of myocardial infarction has long been recognized [13]. Elevation of the total leukocyte count is probably caused by the increase in cortisol levels during an acute stress reaction [14]. The potential value of the leukocyte differential has probably been obscured by the great variation in the traditional manual leukocyte differential, which has a coefficient of variation of 20% to 30% [15]. Widely available, fast, and accurate automated hematology analyzers have greatly decreased the analytical variation for leukocyte differentials [16]. Although the leukocyte differential is influenced by corticosteroid use and other stressful illnesses, an initial study of patients presenting with chest pain to the emergency department suggested that many of these other causes could be identified by simple clinical assessment. In our initial study, we also found that myocardial infarction, but not unstable angina, often altered the leukocyte differential by elevating the neutrophil count and decreasing the lymphocyte count. These changes thereby caused relative lymphocytopenia and granulocytosis, frequently without an elevation of the total leukocyte count (unpublished observations).

Because the automated leukocyte differential is widely available, fast, and inexpensive, it offers an attractive combination of incremental value for the diagnosis of infarction with little incremental cost. We therefore conducted a prospective, blinded study to assess the potential incremental value of the leukocyte differential as an early diagnostic tool. We examined whether relative lymphocytopenia, abnormal total creatine kinase levels, or abnormal rapid creatine kinase-MB measurements were useful for the early diagnosis of acute myocardial infarction in patients presenting to the emergency department with chest pain.

Methods

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Patients

The study group of 384 patients was selected from 511 consecutive patients aged 20 years or older who presented to the emergency department with anterior or left lateral chest pain between 8 March and 31 August 1992 and from whom initial blood samples were obtained (Figure 1). Eleven patients who presented twice and three patients who presented three times during this period were included separately for each visit. Exclusion criteria were the following: infection in the previous week detected by history or initial clinical evaluation (34 patients); exogenous glucocorticoid use within the previous month (25 patients); documentation of malignancy in the previous 5 years (20 patients); transfer from other hospitals after diagnosis (17 patients); major trauma, major gastrointestinal bleeding, surgery, dialysis, or resuscitation in the previous week (8 patients); and insufficient data to exclude myocardial infarction (17 of 20 patients dismissed from the emergency department who did not return after 24 hours for blood tests as part of the protocol and 6 patients who were admitted but had only one determination of the creatine kinase level because of low clinical suspicion). Of the 127 patients who were excluded, 4 patients had evidence of infarction (see below).

View larger version (19K): [in this window] [in a new window]   Figure 1. Flow diagram showing patient selection and the prevalence of ST-segment elevation and positive serum markers (defined as both a relative lymphocytopenia and an elevated rapid creatine kinase-MB level) in patients with and patients without myocardial infarction. In one patient with infarction and no ST-segment elevation, the rapid creatine kinase-MB level was not measured.

The study protocol was approved by the Mayo Clinic institutional review board. In the protocol, the leukocyte differential was identified only as "test X." Only five of the investigators and one representative of the institutional review board knew the identity of "test X." Patients who were not admitted were asked to return the next day as part of the protocol for the measurement of creatine kinase levels and leukocyte counts, but only 3 of 20 such patients returned. To determine the value of blood samples and electrocardiograms for the early diagnosis of acute myocardial infarction, we used only the initial measurements.

Definition of Myocardial Infarction

A final diagnosis of myocardial infarction was made if one of the following two criteria was met: 1) standard creatine kinase-MB isoenzyme level greater than 9.6 ng/mL within 48 hours [73 patients] or 2) sudden unexplained death within 72 hours of presentation (1 patient). The absence of myocardial infarction was defined by the absence of these criteria. In five patients, the final clinical (discharge) diagnosis differed from that defined above. Clinical records for these patients were reviewed by two experienced cardiologists blinded to the results of "test X." One patient had a previously defined myopathy. In the other four patients, creatine kinase-MB levels were mildly elevated, and no further evidence of infarction was observed. These patients were considered by the panel to not have infarction and were categorized as such for the analysis.

Initial electrocardiograms were reviewed independently by one cardiologist (who was blinded to the patient's history, clinical diagnosis, and blood test results) for the presence of diagnostic ST-segment elevation, defined as an elevation of more than 0.1 mV in two electrically contiguous leads. Electrocardiograms showing ST-segment elevation that was not seen on previous electrocardiograms from our institution or ST elevation in patients for whom previous electrocardiograms were not available were considered "positive." Electrocardiograms showing no ST-segment elevation or an elevation that was unchanged from previous electrocardiograms were considered "negative."

Laboratory Measurements

Peripheral blood leukocyte counts were estimated with a widely available automated hematology analyzer, Coulter STK-R (Hialeah, Florida) [16]. This instrument uses approximately 10 000 cells per differential to produce, within minutes, a precise three-part leukocyte differential consisting of granulocytes, lymphocytes, and monocytes. The normal ranges for leukocyte concentrations were defined as the central 95th percentile for 150 healthy adults and have been used since 1984 [16]: 45.8% to 73.7% for granulocytes and 20.3% to 46.7% for lymphocytes.

The total creatine kinase activity was determined on the Kodak Ektachem E700XR (Eastman Kodak Company, Rochester, New York) [17]. The upper limits of normal were determined to be 176 U/L for women and 336 U/L for men on the basis of values from 160 healthy adults. Creatine kinase-MB levels were measured on the Magic Lite system (CIBA Corning Diagnostics, Medfield, Massachusetts) using a two-site chemiluminescent immunoassay method; the cut-off value for the detection of myocardial infarction was 9.7 ng/mL [18]. After this initial clinical testing, serum specimens were frozen at –70°C until they were thawed for measurements of cortisol and rapid creatine kinase-MB levels. Rapid creatine kinase—MB measurements were done on the Stratus II system (Baxter Diagnostics, Inc., Miami, Florida) using a two-site fluorometric immunoassay [19]. The upper limit of normal was chosen to be 4.7 ng/mL, which is three standard deviations greater than the mean value for 73 patients with cardiac disorders but no current myocardial infarction [20]. Cortisol levels were tested by RIA (Diagnostics Products Corporation, Los Angeles, California). The upper limit of normal was 690 nmol/L (25 µg/dL) at 0800 h [20]. The amount of the sample was insufficient for the determination of the total creatine kinase level in one patient without an infarction, for the determination of rapid creatine kinase—MB levels in 9 patients (1 with infarction and 8 without infarction), and for the measurement of cortisol levels in 8 patients without infarction. Laboratory variables were measured without knowledge of clinical or other laboratory results.

Statistical Analysis

Top Methods Statistical Analysis Results Discussion Author & Article Info References

We used the Wilcoxon rank-sum test to assess whether patients with and patients without infarction differed significantly for individual variables. We determined the sensitivity and specificity for individual variables using the abnormal thresholds of greater than 10.4 x 10⁹/L (in women) or 10.9 x 10⁹/L (in men) for total leukocyte counts and of greater than 689.75 nmol/L (25 µg/dL) for serum cortisol levels, as well as the cut-off values described above for total creatine kinase levels, rapid creatine kinase-MB levels, granulocytosis, and lymphocytopenia. Because exogenous estrogen replacement therapy increases cortisol levels, cortisol values from 29 women (4 with infarctions) who were receiving estrogens were excluded from the analysis of cortisol results.

Sex, age, abnormal total creatine kinase levels, abnormal rapid creatine kinase-MB levels, and relative lymphocytopenia were entered into a multivariate logistic regression analysis to predict the presence of infarction in the 345 patients without electrocardiographic ST-segment elevation and for whom complete blood test data were available (10 patients with incomplete data were excluded). We examined these variables because they are potentially available early in the patient's evaluation. We excluded patients with ST-segment elevation from this analysis because of the known high, positive predictive value of this electrocardiographic finding [1], which was confirmed in our study.

Results

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Initial Blood Sample in Patients with and Patients without Infarction

Sixty-nine patients in the study group (18% of 384) had a myocardial infarction (Figure 1). Sixty-eight percent of patients with infarction were male (mean age, 64 years; range, 36 to 94 years) compared with 59% of the patients without infarction (mean age, 63 years; range, 31 to 93 years). Patients with infarction varied in the time from the onset of symptoms to presentation: Forty-eight percent presented in less than 6 hours, 25% between 6 and 12 hours, and 27% in more than 12 hours. Patients with infarction had higher (P < 0.001) total creatine kinase and creatine kinase-MB levels than did patients without infarction (Table 1). Although the total lymphocyte counts were similar in the two groups, the relative lymphocyte count was lower and the total and relative granulocyte counts were higher in patients with infarction. The relative lymphocytopenia in patients with infarction often occurred without a total leukocytosis (Figure 2). As expected, the relative lymphocyte and granulocyte percentages were strongly and inversely related (r = –0.98;P < 0.001); when one proportion increased, the other decreased. Patients with infarction had higher cortisol levels than patients without infarction.

View larger version (21K): [in this window] [in a new window]   Figure 2. Peripheral blood leukocyte counts. Top. Sixty-nine patients with infarction and positive () or negative (\#9679;) electrocardiograms. Bottom. Three hundred fifteen patients without infarction (\#9679;). The dashed line represents the relative lymphocyte (lymph) concentration of 20.3%. The dotted line represents the upper limit of normal for total leukocyte count, 10.9 x 109/L.

Diagnostic Accuracy of Markers for Myocardial Infarction

The initial electrocardiogram was positive in 27 of the patients with myocardial infarction (sensitivity of 39%) and in only 2 of the 315 patients without myocardial infarction (specificity of 99%) (Table 2). The positive predictive value for the electrocardiogram was far higher than the value for any of the other individual markers. Cortisol was the most sensitive marker (sensitivity of 68%) but also the least specific (specificity of 86%); its positive predictive value was therefore only 52%. Rapid creatine kinase-MB had a better sensitivity than did the electrocardiogram (sensitivity of 56%), but its lower specificity (93%) resulted in a positive predictive value of only 66%. The performance of the relative lymphocyte percentage was similar to that of the rapid creatine kinase-MB level, with a sensitivity of 58%, a specificity of 91%, and a positive predictive value of 58%. The performance of the relative granulocyte percentage, with its inverse relation to the relative lymphocyte percentage, was also similar to that of the rapid creatine kinase-MB level. In some patients without infarction, either the rapid creatine kinase-MB level or the relative lymphocyte percentage were markedly abnormal Figure 3, but only 1 patient without infarction had abnormal values by both markers. When abnormal values for both the rapid creatine kinase-MB level and the relative lymphocyte percentage were required for the diagnosis of infarction, this combination had a sensitivity of 44% and a specificity of 99.7%. This combination had a sensitivity of 21% for the 33 patients with infarction who presented within 6 hours of symptom onset, a sensitivity of 65% for the 17 patients who presented between 6 and 12 hours after symptom onset, and a sensitivity of 63% for the 19 patients who presented 12 hours after symptom onset. Exclusion of the 17 repeat visits of the 14 patients who presented more than once had little effect on the results; the major effect was only a 1% to 2% increase in the sensitivity of the individual markers.

View larger version (22K): [in this window] [in a new window]   Figure 3. Relative lymphocyte percentages and rapid creatine kinase-MB levels. Top. Patients with acute myocardial infarction and positive (squares) or negative (circles) electrocardiograms. Bottom. Patients without infarction (circles). Both values were available in 307 patients without infarction and in 68 with infarction. The dashed lines represent the lower limit of normal for the relative lymphocyte percentage and the upper limit of normal for rapid creatine kinase-MB (CK-MB) levels.

Multivariate Logistic Regression Analysis for Patients without ST-Segment Elevation

On the basis of the previously described thresholds and according to a multivariable logistic regression, rapid creatine kinase-MB levels (P = 0.001), relative lymphocyte percentages (P = 0.001), and total creatine kinase levels (P = 0.015) in the 345 patients without ST-segment elevation were independent predictors of myocardial infarction. Because the relative granulocyte percentage was highly correlated with the relative lymphocyte percentage, it yielded similar results when it was substituted in the model for the relative lymphocyte percentage. Subsequent analysis focused on the relative lymphocyte percentage because it had slightly better specificity and positive predictive value on the basis of a univariate analysis (Table 2).

Application of Serum Markers in All Study Patients

Patients were divided into four subgroups on the basis of results for measurements of both rapid creatine kinase-MB levels and relative lymphocyte percentages (available for 375 patients) (Figures 3 and 4). Two hundred eighty-one patients had negative results for both tests; only 7.5% had infarctions (Figure 4). In 36 patients, the relative lymphocyte percentage was abnormal, but the rapid creatine kinase-MB level was normal; 25% of these patients had infarction. In 27 patients, the relative lymphocyte percentage was normal and the rapid creatine kinase-MB level was abnormal; 30% of these patients had infarction. Most importantly, 31 patients had abnormal results by both relative lymphocyte percentages and rapid creatine kinase-MB levels; 97% of these patients had infarction.

View larger version (21K): [in this window] [in a new window]   Figure 4. Combinations of results for relative lymphocyte percentages and rapid creatine kinase-MB levels. Patient subgroups were determined by relative lymphocyte percentages (lymph) and rapid creatine kinase-MB (CK-MB) levels; + indicates abnormal results; –indicates normal results. The number of patients with each combination of relative lymphocyte percentages and creatine kinase-MB levels (top) and the prevalence of acute myocardial infarction (bottom) are shown.

The incremental value of the relative lymphocyte percentage is shown by the findings in these subgroups. Of the 58 patients with abnormal rapid creatine kinase-MB levels, 38 (66%) had infarction (Table 2). Of these 58 patients, 27 had a normal relative lymphocyte percentage; only 8 (30%) of these 27 patients had infarction. The remaining 31 patients had a decreased relative lymphocyte percentage; 30 of these 31 patients (97%) had infarction. Of the 30 patients with infarction and both abnormal relative lymphocyte percentages and rapid creatine kinase-MB levels, 17 had negative electrocardiograms.

If the criterion for the diagnosis of infarction was the presence of both abnormal markers or ST-segment elevation, or both, then the sensitivity for the diagnosis of infarction increased from 39% to 65% (44 of 68 infarctions; 95% CI, 52% to 76%). Furthermore, the specificity remained high at 99%, and the positive predictive value for infarction remained high at 94% (44 of 47 patients with positive test results; CI, 82% to 98%).

Discussion

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Our main finding was that the combination of a relative lymphocytopenia and increased rapid creatine kinase-MB level was an accurate early marker of myocardial infarction. Small increases in rapid creatine kinase-MB levels were much more likely to predict infarction when they were accompanied by stress-induced changes in leukocyte counts. The combination also provided incremental value because it led to the identification of 17 of the patients with infarction who lacked ST-segment elevation. Although potentially life-saving therapies are available for patients who will have myocardial infarction, many patients do not receive them because they do not have ST-segment elevation [4]. Patients who present without these diagnostic electrocardiographic changes have a high in-hospital mortality [4]. If our data are confirmed by further studies, patients without diagnostic electrocardiographic changes who have both increased rapid creatine kinase-MB levels and a relative lymphocytopenia may obtain better triage and may be candidates for future trials of therapeutic interventions.

The increase in the total leukocyte count that occurs with myocardial infarction, including a shift in the leukocyte differential, has been known for decades [13] and may be caused by an increase in cortisol levels [14, 21]. The conventional manual differential has significant analytical variation [15] that, combined with the low specificity caused by the leukocytosis of other illness, has caused a decreased interest in changes in the leukocyte count that are associated with myocardial infarction. We used a widely available automated hematology analyzer to markedly decrease the analytical variation of the leukocyte differential and excluded patients with easily identified illness. However, the 91% specificity for a relative lymphocytopenia was similar to that of the biochemical markers, and the positive predictive value was only 58%. On the basis of a single measurement, this value is inadequate for the diagnosis of infarction.

The modest sensitivity, high specificity, and high positive predictive value for electrocardiographic ST-segment elevation measured in our study are consistent with findings of previous studies. Rouan and colleagues [3] reported a sensitivity of 45% for electrocardiographic ST-segment elevation in a large series of 7115 patients. Karlson and colleagues [2] reported an identical sensitivity in a population-based series. Several investigators have reported a high positive predictive value for ST-segment elevation [1]. Other acute electrocardiographic changes, such as ST-segment depression or left bundle-branch block, have a much lower positive predictive value [1, 2].

Unfortunately, in our study and other studies [22], individual peripheral blood markers lacked the extremely high specificity needed to produce clinically useful high positive predictive values. This is because of the low prevalence of infarction for patients who present with chest pain. For example, on the basis of the prevalence of infarction in our study group (18%), rapid creatine kinase-MB levels would have had a positive predictive value of greater than 90% only if the specificity had been greater than 98.7%. Even the promising new marker, the creatine kinase-MB subform, has a specificity of only 93.9% [23]. Only ST-segment elevation has had sufficient specificity to provide the high positive predictive value required to treat patients with expensive and potentially hazardous interventions.

Our study suggests that a combination of peripheral blood markers, increased rapid creatine kinase-MB levels, and relative lymphocytopenia may have sufficient specificity, even in patients without diagnostic electrocardiographic changes, to obtain high positive predictive values. The combination of an increase in the rapid creatine kinase-MB level and stress-induced changes in the leukocyte differential (a relative lymphocytopenia) had a specificity of 99.7% and a positive predictive value of 97%. Because this combination also led to the identification of some of the patients with infarctions who lacked ST-segment elevation, it may help identify more candidates for trials of early therapy. Further studies are needed to determine whether some of these patients have left-circumflex occlusion that has been systematically under-represented in trials that require ST-segment elevation for entry.

The mechanism for the changes in leukocyte counts probably involves interactions among the nervous, endocrine, and immune systems, which differ from the electrical and cell-damage mechanisms for electrocardiographic and biochemical markers of infarction. The myocardium has afferent nerves to inform the central nervous system of severe ischemia and impending myocardial damage [24]. The hypothalamus responds by increasing corticotropin-releasing hormone levels to increase adrenocorticotropic hormone levels, which then cause pulsatile increases in cortisol levels [25, 26]. Cortisol regulates peripheral blood leukocyte counts [27, 28], and increases in cortisol levels induce granulocytosis and relative lymphocytopenia. The changes in the leukocyte count, which may integrate the pulsatile changes in cortisol levels, are inexpensively and accurately estimated by automated hematology analyzers. Cortisol increases peripheral blood granulocyte counts in part by decreasing migration into tissues, including a damaged myocardium; this may limit leukocyte destruction of the myocardium [29-32].

Although cortisol was the most sensitive of the individual markers, several factors probably caused its low specificity and poor positive predictive value. Elevated cortisol levels tend to decrease into the upper range of normal soon after infarction occurs [12]. Because serum cortisol levels have a short half-life and increase in a series of pulses, single measurements often do not reflect recent secretion [25]. Other stressful clinical conditions can also increase cortisol levels. However, because increases in cortisol levels precede the changes in leukocyte counts, the rapid assay of cortisol may be useful in patients presenting immediately after the onset of symptoms.

The combination of a relative lymphocytopenia and the creatine kinase-MB level may be a general way to use a marker of severe stress to identify whom, of many patients with a slight increase in a biochemical marker, is having an infarction. For example, a "stress" marker may not be elevated with unstable angina, small amounts of myocardial damage, or noncardiac conditions that cause a slight increase in a biochemical marker, including assay variation. Because the stress marker reflects a different mechanism, it can improve even a good biochemical marker. For example, changes in the leukocyte differential, an inexpensive stress marker, could potentially be combined with slight increases in troponin T [11], C-reactive protein [33], or creatine kinase-MB subform levels [23] to discriminate which biochemical change is caused by unstable angina and which is caused by an impending myocardial infarction.

Our study had several limitations. We did not consider the clinical variables that have been shown to improve the initial evaluation of patients with chest pain [34]. In the proper clinical application of our results, the character and duration of pain, previous clinical history of ischemic heart disease, and physical examination should be considered. We could not assess the possible treatment implications of duration of pain because we did not collect this information from patients without infarction. In addition, patients must be clinically screened for conditions associated with stress or changes in leukocyte counts, such as surgery, infections, and exogenous steroid use; our findings only apply when such patients have been excluded. Although investigators of an earlier study [35] did not find any age or sex differences in relative leukocyte counts, data are inadequate to assess possible racial differences. The sensitivity of the combination of relative lymphocytopenia and increased rapid creatine kinase-MB levels is too low to exclude infarction and justify discharge from the emergency department. Although we screened more than 500 patients, our estimates of sensitivity and positive predictive value are subject to considerable error because of the modest number of infarctions. Application of our results in settings with a lower prevalence of infarction will decrease the positive predictive value; for example, a prevalence of 9% would lower the positive predictive value to 94%. Creatine kinase-MB results may be assay-specific [8]. Our results must be confirmed in more patients, and outcomes other than myocardial infarction should be examined.

Our study suggests that an accurate leukocyte differential is as valuable for the early diagnosis of infarction as the more widely used biochemical markers. Because the differential is usually obtained for patients with chest pain, it offers the attractive combination of incremental value at little incremental cost. However, because of the low prevalence of infarction in patients who present with chest pain, any current individual marker, with the exception of electrocardiographic ST-segment elevation, is not specific enough to ensure a high positive predictive value. The presence of both a relative lymphocytopenia and an elevated rapid creatine kinase—MB level had greater positive predictive value than either individual marker alone and may lead to the identification of patients with impending myocardial infarction, including some who do not have ST-segment elevation. Similar combinations of stress and biochemical markers may enhance the early identification of patients with myocardial infarction.