Author + information
- Received October 29, 1998
- Revision received February 24, 1999
- Accepted March 26, 1999
- Published online July 1, 1999.
- Judd E Hollander, MD∗,* (, )
- M.Ranu Muttreja, MD†,
- Margaret R Dalesandro, PhD‡ and
- Frances S Shofer, PhD∗
- ↵*Reprint requests and correspondence: Dr. Judd E. Hollander, Department of Emergency Medicine, Hospital of the University of Pennsylvania, Ground Floor, Ravdin Building, 3400 Spruce Street, Philadelphia, Pennsylvania 19104-4283
We compared the predictive properties of P-selectin to creatine kinase, MB fraction (CK-MB) for detecting acute myocardial infarction (AMI), acute coronary syndromes (ACS) and serious cardiac events upon emergency department (ED) arrival.
Practioners detecting early diagnosis of ACS have focused on cardiac markers of myocardial injury. Plaque rupture/platelet aggregation precedes myocardial ischemia. Therefore, markers of platelet aggregation may detect ACS earlier than cardiac markers.
Consecutive patients with potential ACS presenting to an urban university ED were identified by research assistants who screened all ED patients between November 12, 1997 and January 31, 1998. Whole blood was drawn at presentation and 1 h later and rapidly stained and fixed for membrane P-selectin assay and plasma was separated for soluble P-selectin assay. Creatine kinase, MB fraction values were determined using standard immunoassay techniques. Clinical history and hospital course were followed daily. Outcomes were AMI, ACS (AMI and unstable angina) and serious cardiac events. Receiver operator characteristic curves were derived for CK-MB, and soluble and membrane-bound P-selectin to determine the optimal cutoff values. Predictive properties were calculated with 95% confidence intervals.
A total of 263 patients were enrolled. They had a mean age of 56.5 ± 14 years; 52% were male. There were 22 patients with AMI; 87 patients with ACS and 54 patients with serious cardiac events. Creatine kinase, MB fraction had a higher specificity for detection of AMI, ACS and serious cardiac events than both soluble and membrane-bound P-selectin. At the time of ED presentation, the specificity of CK-MB, and soluble and membrane-bound P-selectin for AMI was 91% versus 76% versus 71%; for ACS, 95% versus 79% versus 71%, and for serious cardiac events, 91% versus 76% versus 72% (p < 0.05). The sensitivities for AMI were 50% versus 45% versus 32%; for ACS, 26% versus 35% versus 30%, and for serious cardiac events, 29% versus 35% versus 36%.
Although theoretically attractive, the use of soluble and membrane-bound P-selectin for risk stratification of chest pain patients at the time of ED presentation does not appear to have any advantages over the use of CK-MB.
There are more than five million emergency department (ED) visits for evaluation of symptoms of acute coronary syndromes (ACS) annually in the United States (1). Because the accuracy of the initial working diagnosis is low and the reliability of early exclusion of ACS is uncertain, most patients with symptoms suggestive of ACS are admitted to expensive monitored settings or observation units for further evaluation, placing a heavy burden on health care economics. As a result, several strategies to delineate the patients’ initial risk have been developed and tested. Although several clinical and computer algorithms have been able to stratify patients into low and high risk groups for infarction or cardiovascular complications, these algorithms have not identified a group of patients at such low risk that they can be safely and immediately released from the ED (2–8). Consequently, recent research has focused on the use of adjunctive technologies to stratify patients into high and low risk groups for infarction and other cardiovascular complications. These strategies include serial electrocardiograms, continuous ST-segment monitoring (9,10), additional electrocardiographic leads (11–13), marker proteins of myocardial injury, such as creatine kinase, MB fraction (CK-MB) (14–20), myoglobin (21–25)and the cardiac troponins (26–30)and the use of short-term observation units (31–34). Despite intensive investigation, these strategies, which largely focus on identification of patients with myocardial injury, have not identified patients at such low risk that they can be safely released from the ED. An alternative strategy, identification of patients with ACS before the commencement of myocardial injury, is appealing. In theory, detection of platelet activation could identify high risk patients before the occurrence of myocardial injury.
A soluble form of P-selectin recently identified in the plasma of normal individuals is an indicator of platelet activation. It is elevated in the plasma of patients with acute myocardial infarction (AMI) (35), unstable angina (36)and acute lung injury (37). This molecule appears to represent both a proteolytic fragment of membrane-bound P-selectin and a soluble splice variant lacking the transmembrane domain of P-selectin (38). Preliminary data suggest that flow cytometric determination of transmembrane P-selectin expression as a marker of platelet activation may differentiate stable from unstable coronary artery disease (39). The purpose of the present study was to evaluate the predictive properties of soluble and membrane-bound P-selectin for the ED risk stratification of patients with ACS.
We performed a prospective observational clinical study to evaluate the sensitivity and specificity of P-selectin for the diagnosis of ACS (myocardial infarction and unstable angina) in patients who presented to the ED with chest pain symptoms consistent with myocardial ischemia.
This study was conducted in the ED of Philadelphia’s Hospital of the University of Pennsylvania, an urban tertiary care center with an annual patient census of approximately 47,000 visits.
Consecutive patients greater than 18 years of age who presented to the ED between November 10, 1997 and January 31, 1998 with a chief complaint of chest pain of less than 6 h duration consistent with acute myocardial ischemia were eligible for inclusion. During the study period, dedicated research assistants were present in the ED 24 h per day, 7 days per week to identify potential study patients.
A full-time board-certified/board-eligible emergency physician treated patients who consented to enrollment alone or in conjunction with house staff. Patients were excluded if they had symptoms that were clearly not cardiac in origin (e.g., chest pain secondary to herpes zoster or recent trauma). All study subjects provided written informed consent. The study was approved by the University of Pennsylvania Committee on Research Involving Human Subjects.
Patients had a structured history and physical examination performed at the time of ED presentation. Patients had study blood specimens obtained at the time of presentation and 1 h later. These times were chosen because serial sampling at these early times might influence decision making and triage. Patient disposition was determined by the ED attending physician, blinded to study sample results.
Admitted patients were followed daily throughout their hospital course by one of the investigators. Patients were contacted 30 days after presentation for follow-up.
Platelet P-selectin flow cytometry assay
Membrane P-selectin expression on platelets was determined by two-color flow cytometry of fluorophor-conjugated antibody stained and fixed whole blood samples using the Centocor P-Selectin FLOW Assay (Malvern, Pennsylvania). For each patient at each time point, a 4.5-ml blood sample was drawn into an Diatube-H (C.T.A.D.) Vacutainer evacuated anticoagulant tube (Becton Dickinson, ref. 367015, Investigational Use Only), through a 19-ga or larger needle. The sample was immediately diluted 1:10 in Tyrodes’ buffer by transferring 50 μl of whole blood to an Eppendorf tube containing 450 μl of diluent. The tube was inverted gently to mix. Forty microliters of reagents A (buffer only), B (fluorescein isothiocyanate [FITC]-conjugated isotype control plus phycoerythrin (PE)-labeled pan-platelet monoclonal antibody) and C (FITC-conjugated P-selectin monoclonal antibody S12 plus PE-labeled pan-platelet monoclonal antibody) was added to the bottom of appropriately labeled amber Eppendorf tubes. Sixty microliters of diluted whole blood were added to each of the reagent tubes. Blood was incubated in the reagent tubes at 18 to 25°C for 15 min. After staining, an equal volume of freshly made 2% paraformaldehyde solution was added to the bottom of each amber reaction tube to fix the cells (final concentration of paraformaldehyde concentration, 1%). Samples were fixed for 30 to 120 min at room temperature. Samples were stored refrigerated for up to 72 h before flow cytometry without loss of fluorescence intensity. At the start of each flow cytometric analysis, the performance of the Becton-Dickinson FACScan and FACSSCalibur flow cytometers were standardized by running flow cytometer control beads including Quantum 24 beads, Full Spectrum Microbeads and Calibrite Beads. Any changes in the mean channel fluorescence observed with the control beads resulted in restandardization of the instrument to assure that there was uniformity during the course of the study in the detection of the percent of platelets that were positive for P-selectin. In addition, samples were prepared using matched FITC- and PE-labeled antibodies to check the compensation settings used. For each clinical sample, the unstained sample was used to check for autofluorescence of the patient’s platelets. The stained samples were gated on the PE-positive population (red) that was platelets. The P-selectin–specific FTIC-conjugated (green) antibody fluorescence was used to determine the percent of platelets in the clinical sample that were positive for expression of P-selectin. Histogram overlays of the isotype control and P-selectin–specific antibody stained samples were created and a statistical marker was set at 1 ± 0.1% gated events for the isotype control antibody. The resulting percent positivity for the P-selectin–specific antibody samples using that same marker is reported as P-selectin percent positive. The integrity of the reagents was determined by the use of the matched lot of the Centocor Diagnostics P-selectin compensation kit along with the complementary Centocor Diagnostics P-selectin flow cytometry test kit. As part of the compensation test, the intensity of the FITC-labeled P-selectin was determined by binding to platelets activated with adenosine diphosphate (ADP). The integrity of the PE-labeled pan-platelet monoclonal antibody was determined by binding to platelets in diluted whole blood.
Soluble P-selectin enzyme-linked immunosorbent assay (EIA) assay
Within 60 min from the time it was drawn, the remainder of the 4.5-ml blood sample was centrifuged at 1,800 gat room temperature for 15 min to make platelet-poor plasma. The clear platelet-poor plasma was separated from the underlying cell pellet, aliquotted and frozen at ≤70°C. The levels of soluble P-selectin were determined in the plasma using the Centocor Soluble P-Selectin EIA. Platelet-poor plasma was prediluted 1:20 with assay buffer. Fifty microliters of prediluted samples were run on each EIA plate in duplicate along with triplicate 50-μl samples of each of five standards of known soluble P-selectin value from 0 to 45 ng/ml (multiplied by the dilution factor of 20, the standard curve covers soluble P-selectin values up to 900 ng/ml). Plasma control samples with known values of soluble P-selectin considered high (500 ng/ml) and low (100 ng/ml) were run in triplicate on each plate. If the control samples did not fall within a predetermined range of values, the assay was considered invalid and was repeated.
Briefly, standards, control or study plasma samples, diluent, biotinylated P-selectin–specific monoclonal antibody W40 and horseradish peroxidase–conjugated P-selectin–specific monoclonal antibody S12 were added to a streptavidin-coated microplate. The plate was incubated at room temperature for 3 h. After the plate was washed, 3,3′,5,5′-tetramethylbenzidine substrate solution was added and incubated at room temperature for 30 min. Stop solution was added to stop color development. The plate was read on a microplate reader using optical density 450 to 650. A standard curve was created using a four-parameter fit; mean absorbances of the unknowns were used to determine their P-selectin levels. Samples containing greater than 850 ng/ml were prediluted 1:100 and were reevaluated in the Centocor Soluble P-Selectin EIA.
Creatine kinase, MB fraction assay
Creatine kinase, MB fraction was analyzed on the Dade Stratus immunoassay analyzer (Dade International, Miami, Florida) utilizing a rapid two-site sandwich monoclonal assay. This assay measures immunologic activity rather than enzymatic activity. The within-run coefficient of variation ranges from 2.5% to 3.8% and the between-run coefficient of variation ranges between 3.0% and 3.7%. The manufacturer defined the reference range for CK-MB as less than 5 ng/dl.
Electrocardiograms (ECGs) were interpreted blinded to the patients’ serum results and clinical outcome. Each ECG was classified into one of seven categories using a closed question format (40): 1) normal (no electrocardiographic evidence of ischemia); 2) nonspecific (accepted deviation from the norm with the lowest likelihood of ischemia such as an abnormal T-wave axis in lead III or sinus tachycardia); 3) early repolarization variant (elevated takeoff of the ST segment at the J point of the QRS complex greater than 1 mm relative to the isoelectric line, downward concavity of the ST segment and symmetrically peaked T waves often of large amplitude); 4) abnormal but nondiagnostic of myocardial ischemia (prolonged QRS, corrected QT intervals, bundle branch blocks, left ventricular hypertrophy with strain); 5) ischemia or prior infarction known to be old (ST depression more than 0.1 mV measured 80 ms from the J point, inverted T waves more than 0.3 mV or Q waves at least 30 ms in duration); 6) ischemia or prior infarction not known to be old, or 7) suggestive of AMI (ST elevation greater than 0.1 mV measured 80 ms from the J point in two or more contiguous leads, with or without reciprocal ST depressions).
The final diagnosis of myocardial infarction was based on the clinical presentation, serial ECGs and serial in-hospital biochemical marker analysis in the hospital clinical laboratory using World Health Organization criteria (41). The biochemical profile diagnostic of AMI consists of a typical rise and fall of total CK above twice normal values within 24 h of presentation and CK-MB higher than twice normal values.
Unstable angina was classified according to the AHCPR risk stratification scheme (42). Because the definition of unstable angina in the guidelines is based on the accurate distinction of unstable angina from nonischemic chest pain syndromes, we required demonstration of underlying coronary artery disease (>70% stenosis), exercise-induced ischemia (>1.5 mm ST segment depression or reversible defects on nuclear images or echocardiography) or CK-MB detection above the upper limits of normal (5 ng/dl) but below the threshold for AMI (5 to 10 ng/dl) to be classified as angina.
Chest pain syndromes were considered to be nonischemic if the patients received diagnostic testing (stress testing or cardiac catheterization) without detection of underlying disease. Patients who did not receive an evaluation for underlying coronary artery disease were classified as possibly ischemic, since ischemia was not “ruled out.”
The composite end point of serious cardiovascular events was defined as death within 30 days from a cardiac etiology, AMI arrhythmias requiring intervention, congestive heart failure, hypotension requiring treatment, respiratory failure requiring intubation, angioplasty, coronary artery bypass surgery or cardiac arrest.
Data were entered into a Microsoft Access 95 database (Microsoft, Redmond, Washington) and were imported into SAS 6.12 (SAS Institute, Cary, North Carolina) for statistical analysis. Continuous data are presented as means with standard deviations. Categorical data are presented as the percent frequency occurrence. Because patients with the final diagnosis of possibly ischemic had similar profiles and outcomes to the patients classified as nonischemic, and the decision to forego an evaluation in the possibly ischemic group was based upon a low clinical index of suspicion, these groups were combined for analysis. Receiver operator characteristic (ROC) curves were constructed to determine the best cutoff values for P-selectin for each of the main outcomes (AMI, ACS and serious cardiac events). Area under the curve was calculated for each test and outcome and a z test was used to evaluate the differences between two ROC curves (True Epistat 3.0, Tracy Gustafson, Epistat Σ Services). Using the “best” cutoff value for CK-MB, and soluble and membrane-bound P-selectin, the sensitivity, specificity and positive and negative predictive values of these tests to predict AMI, ACS and serious cardiac events were calculated with 95% confidence intervals (43). Predictive properties were also provided for the combination of soluble and membrane-bound P-selectin (the P-selectin index) and the combination of the P-selectin index and CK-MB, requiring elevation in one or more of these values.
During the 10-week study period, 293 patients with potential ACS presented to the ED. Thirty patients were excluded from participation in the study: 25 refused participation, 4 were unable to have blood samples obtained in a timely manner and 1 patient was missed by research assistants. As a result, the final study group was 263 patients. Comparison of patients included and excluded from the study showed that they were similar with respect to age, chief complaint, chest pain characteristics, cardiac risk factors, ED diagnostic impression, triage disposition and initial ECG.
Study patients had a mean (± SD) age of 56.5 ± 14.2 years; 52% were male. Patients presented a median of 3 h (IQR, 60 min to 240 min) after symptom onset and the total duration of symptoms was a median of 60 min (interquartile range, 25 min to 180 min). The most common chief complaints were chest pain (76%) and shortness of breath (11%). Forty-four percent of patients had multiple symptomatic episodes within 48 h of presentation and symptoms occurred at rest in 74%. Chest pain characteristics, associated symptoms and medical history are shown in Table 1.
The initial ECGs were classified as follows: normal, 17%; nonspecific, 33%; early repolarization, 5%; abnormal but not diagnostic, 21%; ischemia known to be old, 10%; ischemia not known to be old, 10% and suggestive of AMI, 4%.
Two hundred thirty patients were admitted to the hospital (cardiac care unit, 10%; cardiac intermediate care unit, 19%; non–intensive care unit monitored bed, 57%, unmonitored bed, 2%), 6 patients (2%) underwent prolonged ED observation and 27 patients (10%) were discharged to home. The study institution did not have an ED observation unit. There were 22 patients with AMI (8%), 65 patients classified as unstable angina (25%), 113 patients (43%) with a final diagnosis of nonischemic chest pain syndrome and 59 patients (22%) classified as possibly ischemic (they did not receive a sufficient workup to “rule in” or “rule out” coronary artery disease).
There were serious cardiac events in 54 patients, including the 22 patients with AMI. Cardiovascular complications that occurred after ED arrival included the following: congestive heart failure, 8 patients (3.1%); ventricular tachycardia/fibrillation, 3 patients (1.2%); supraventricular tachycardia, 8 patients (3.1%); bradydysrhythmias, 3 patients (1.2%); hypotension requiring pressors, 3 patients (1.2%), and permanent pacemaker placement in 4 patients (1.6%). There were three deaths during index hospitalization.
Forty-five patients had cardiac catheterization performed, 32 had coronary artery disease. Of these, 15 patients had angioplasty and 10 patients had stent placement during index hospitalization. One patient had coronary artery bypass surgery. Seventy-three patients had stress testing performed. Fifteen showed reversible ischemia.
Thirty-day follow-up was completed in 251 of the 263 patients (95%). There were 12 deaths (5%), 1 nonfatal AMI, 4 patients who had angioplasty and 3 patients who had stent placement after index hospitalization.
Receiver operator characteristic curves for membrane-bound P-selectin, soluble P-selectin and CK-MB at the time of ED presentation are shown for each of the outcomes in Figures 1 to 3. ⇓⇓⇓Figure 1shows the ROC curves for the diagnosis of AMI. The area under the curve for CK-MB is 0.721; membrane-bound P-selectin is 0.531 and soluble P-selectin is 0.646 (p < 0.0001 for CK-MB vs. membrane-bound P-selectin; p = 0.086 for CK-MB vs. soluble P-selectin). Figure 2shows the ROC curves for the diagnosis of ACS (the 87 patients with either AMI or unstable angina). The area under the curve for CK-MB is 0.670; membrane-bound P-selectin is 0.555 and soluble P-selectin is 0.583 (p = 0.003 for CK-MB vs. membrane-bound P-selectin; p = 0.01 for CK-MB vs. soluble P-selectin). Figure 3shows the ROC curves for serious cardiac events. The area under the curve for CK-MB is 0.587, for membrane-bound P-selectin is 0.554 and for soluble P-selectin is 0.539 (p = 0.233 for CK-MB vs. membrane-bound P-selectin; p = 0.236 for CK-MB vs. soluble P-selectin).
Using the optimal cutoff values for each assay, as derived from the ROC curves, the sensitivity, specificity and positive and negative predictive values along with 95% confidence intervals are shown in Tables 2 to 4. ⇓⇓⇓Creatine kinase, MB fraction was more specific than soluble or membrane-bound P-selectin for the detection of AMI, ACS and serious cardiac events. The sensitivities of the three assays were similar.
For the detection of AMI at the time of ED presentation, the use of a P-selectin index (requiring elevation in either soluble or membrane-bound P-selectin) improved the sensitivity at the cost of decreased specificity relative to either P-selectin assay alone (Table 2). It did not attain greater sensitivity than CK-MB. The specificity of the P-selectin index was less than CK-MB for detection of AMI. The addition of CK-MB to the P-selectin index did not improve the predictive properties compared with the P-selectin index alone.
For detection of ACS and serious cardiac events, the P-selectin index had a higher sensitivity and lower specificity than CK-MB. The positive predictive value of the P-selectin index was less than CK-MB and the negative predictive value was similar (Tables 3 and 4). The addition of CK-MB to the P-selectin index did not improve the predictive properties compared with the P-selectin index alone.
The addition of serial sampling 1 h after presentation did not significantly affect the sensitivity, specificity and positive and negative predictive values of either soluble or membrane-bound P-selectin. Patients who presented early were similar to patients who presented later.
Acute Q-wave myocardial infarction, non–Q-wave myocardial infarction and unstable angina represent the continuum of ACS. Whereas Q-wave AMI results from an occlusive thrombus (44), non–Q-wave myocardial infarction and unstable angina result from subtotal thrombotic obstruction of the coronary arteries (45). Early identification of patients with acute ischemic syndromes has largely focused on markers of myocardial cell injury, such as CK-MB, myoglobin and the cardiac troponins (14–30). Although many of these markers have independent predictive value for cardiovascular complications, they have not been widely used to identify patients at such low risk that they may be discharged from the ED.
Recent treatment algorithms for patients with ACS focus on platelet aggregation and thrombosis. After injury to the vessel wall, atherosclerotic plaques rupture, and platelet activation, adherence and aggregation occur (46). Acute coronary syndromes involve a complex cascade of interactions between platelets and the vascular endothelium. Logically, identification of patients with ACS by detection of platelet activation could lead to diagnosis before electrocardiographic or serologic evidence of myocardial ischemia. A reliable marker of platelet activation could maximize the benefits of early treatment through early diagnosis of myocardial ischemia.
Theoretical advantages of P-selectin
P-selectin, a 140-kDa integral membrane glycoprotein, mediates early leukocyte rolling during the acute inflammatory response. It is stored in the alpha granules of platelets and Weibel-Palade bodies of endothelium (47,48). P-selectin is rapidly expressed at the cell surface and cleaved into the circulation when these granules fuse with the cell surface after activation by agonists (such as thrombin and ADP). The soluble form of P-selectin represents both a proteolytic fragment of membrane-bound P-selectin cleaved from the surface of circulating activated platelets (49)and also a soluble splice variant lacking the transmembrane domain of P-selectin (38).
Among the markers of platelet activation, P-selectin has significant advantages over the other markers for which assays exist. The most obvious advantage of P-selectin is that it can be measured in either of two forms, on the platelet surface, or in platelet-poor plasma. Simultaneous measurement of both membrane-bound and soluble P-selectin (P-selectin profile) is designed to increase the diagnostic window for this analyte around a platelet-activating event. P-selectin molecules on the surface of platelets are increased from 200 to 300 on an unactivated platelet to 8,000 to 12,000 on a fully activated platelet within minutes of exposure to platelet agonists such as thrombin, tartrate-resistant acid phosphatase, collagen, ADP and shear. Membrane P-selectin is detected by two-color flow cytometry using fluorescently tagged P-selectin–specific and pan-platelet antibodies. The soluble form of P-selectin is readily assessed in a sandwich EIA format.
Among the other proteins released in association with platelet activation, beta-thromboglobulin (βTG) and platelet factor 4 (PF4) have been used for the early diagnosis of some thrombotic events. P-selectin has several advantages over these proteins. First, plasma P-selectin has a longer half-life than BTG and is not subject to ex vivo release caused by the method of phlebotomy, as is BTG (50). Although PF4 is only elevated in acute clinical states, its diagnostic utility is limited by its clearance within 5 min of release through binding to circulating glycosaminoglycans (such as heparin) as well as those associated with the vascular endothelium (51). This results in a very narrow diagnostic window for PF4 and makes it not useful in patients on heparin. Serotonin is released by activated platelets but is assayed by a rather time-consuming radioimmunoassay. Thromboxane A2 and B2 are produced by the metabolism of arachidonic acid and therefore cannot be measured accurately in the presence of aspirin-containing medications, which inhibit arachidonic acid metabolism. In contrast, P-selectin is stable in plasma, and less sensitive to blood collection artifacts than these other markers of platelet activation. In addition, the presence of heparin or traditional doses of aspirin does not affect P-selectin (52,53).
In a soluble P-selectin EIA developed by Centocor Diagnostics, a mean of 100 ng/ml ± 27.7 was reported for 200 apparently healthy control subjects. Soluble P-selectin levels up to 800 ng/ml have been recorded using this EIA for patients suffering from thrombotic disorders. Ikeda et al. found that plasma concentrations of soluble P-selectin were higher in nine patients with AMI than in a control cohort of 10 healthy volunteers. These samples were obtained during the first 3 days of hospitalization (36). In a study of 12 patients with unstable angina, 11 patients with stable exertional angina and 15 healthy volunteers, Ikeda et al. found that plasma P-selectin levels were elevated in the cohort of patients with unstable angina at 1 and 3 h but not 5 h after symptom onset (35). Although the patients with exertional angina developed ST-segment depression during treadmill exercise testing, their P-selectin values were not elevated relative to normal control subjects. These data suggested that patients with ACS might have elevations in circulating soluble P-selectin shortly after the onset of symptoms. Shimomura et al. evaluated soluble P-selectin levels in 16 patients with AMI, 15 patients with angina and 13 patients with chest pain syndromes but without coronary artery disease (54). Samples were obtained at 7 amthe morning after admission. Patients with AMI were found to have increased soluble P-selectin relative to the other groups, although nine of 16 AMI patients had P-selectin levels below the highest values for both the control and angina groups. These data suggest that P-selectin may not adequately discriminate between patients with and without AMI.
Risk stratification with P-selectin
The present study examined the use of both soluble P-selectin and transmembrane P-selectin for identification of patients with AMI, ACS and serious cardiac events. We found that neither soluble nor transmembrane P-selectin was useful for early identification or risk stratification of patients who present to the ED with acute chest pain syndromes. On the basis of receiver operator curve analysis of the optimal cutoff for each assay, the sensitivity and specificity for identification of patients with AMI, ACS and serious cardiac events were less than those for CK-MB. Use of the P-selectin index, although theoretically attractive, did not appear to offer additional benefits. The sensitivity of the P-selectin index for AMI was not greater than that of CK-MB and the specificity was less than that of CK-MB. For prediction of ACS and serious cardiac events, the sensitivity of the P-selectin index was greater than that of CK-MB but the specificity was considerably less. This resulted in a positive predictive value of only 39% for ACS and 24% for serious cardiac events. Clearly, the low positive predictive values of these assays prohibit initiation of any therapeutic interventions with potentially harmful side effects (glycoprotein IIb/IIIa inhibitors or fibrinolytic agents).
Potential explanations and limitations
There are several potential explanations for our results. Our results differ from prior studies, which mostly compared normal control subjects with patients with AMI. The soluble P-selectin values in the normal subjects were below those found in our subjects who presented to the ED with chest pain syndromes but were “ruled out” for ACS. It may be that our population included a higher prevalence of patients with hypertension, diabetes and underlying vascular disease than the healthy control subjects used in prior studies. Our patient population might have higher baseline levels of platelet activation and higher mean P-selectin values than those seen in healthy control subjects without ongoing chest pain syndromes. In addition, platelet activation is a dynamic process. Prior studies (36,54), in general, obtained serum samples later in the course of disease. Our samples were obtained at the time of ED presentation, which was approximately 3 h after symptom onset and before any treatment interventions. It is possible that later samples or posttreatment samples may have higher levels of P-selectin.
It is important to address the possibility of selection and misclassification bias in any study that evaluates early identification and risk stratification of patients with ACS. To reduce the likelihood of selection bias, we had trained research assistants present in the ED 24 h per day, 7 days per week during the study period. In conjunction with the ED staff, they screened all ED patients to identify those being evaluated for potential ACS. Statistical comparison of the 29 patients excluded from the study with those who consented for study enrollment found that they were similar.
Misclassification bias of important end points such as AMI, unstable angina and serious cardiac events was reduced through daily follow-up on all study patients and adherence to rigid definitions. All patients with a final diagnosis of ACS had demonstration of underlying coronary artery disease, positive findings on provocative testing or myocardial marker evidence of AMI. However, it is possible that some patients may have had false positive or false negative stress tests, resulting in some misclassification of unstable angina patients.
Many different processes and various clinical situations readily influence platelets as sentinel markers of vessel wall integrity. Their responsiveness can be measured by changes in the expression of membrane glycoproteins including P-selectin. It is interesting to note that Cannon et al. observed that the platelets of patients with ACS have increased reactivity to agonists as shown by flow cytometric measurement of P-selectin expression when compared with apparently healthy control subjects (55). This hypersensitivity to agonists among patients with ACS means that extraordinary care must be taken to eliminate any causes of ex vivo platelet activation. The effects of ex vivo platelet activation would be greater on platelets from the subpopulation of patients whose chest pain had its etiology in coronary artery disease, because these patients have a lower threshold for platelet activation. Opportunities for platelet activation had to be minimized during the blood collection and platelet and plasma assay preparation steps. Some of the factors that are known to affect the P-selectin expression of blood samples are the anticoagulant, shear caused by needle gauge during venipuncture, residence time of unfixed blood in a glass vacutainer tube and agitation or temperature changes of blood after it is drawn and before fixation. These variables were carefully controlled in this study. The anticoagulant of choice was CTAD (Diatube H, Becton/Dickinson, Franklin Lakes, New Jersey, investigational use only) which results in a lower percent of activated platelets than acid-citrate-dexrose, disodium citrate, ethylenediaminetetraacetic acid, and heparin. Venipuncture was performed with a 19-ga needle to minimize the shear produced. The staining and fixation of platelets was initiated within 1 h of the blood draw in accord with experimental data showing the possibility of in vitro platelet activation in blood sitting for greater than 60 min before the start of staining and fixation. Vacutainers were carefully and gently inverted to mix anticoagulant with blood. All samples were stored in insulated carriers to assure constant temperature. Because soluble P-selectin is in part produced by the proteolytic cleavage of membrane-bound P-selectin, any variable that causes the expression of P-selectin on the platelet surface can also contribute to the plasma P-selectin levels. Therefore, processing of plasma was carried out with the same rigor as that observed for preparation of platelets for flow cytometric determination of membrane-bound P-selectin.
Our study of consecutive ED patients with potential ACS found that neither soluble nor membrane-bound P-selectin was useful for risk stratification of patients for AMI, unstable angina or serious cardiac events.
The authors thank Jason Cohen, Henry Lin, Kristen Robinson, Lane Ertugay and the Emergency Department Academic Associates and Staff for their dedication to the study and assistance with patient enrollment.
☆ This study was supported by a grant from Centocor Diagnostics, Inc., Malvern, Pennsylvania.
- acute coronary syndromes
- adenosine diphosphate
- acute myocardial infarction
- creatine kinase, MB fraction
- emergency department
- enzyme-linked immunosorbent assay
- fluorescein isothiocyanate
- platelet factor 4
- receiver operator characteristic
- Received October 29, 1998.
- Revision received February 24, 1999.
- Accepted March 26, 1999.
- American College of Cardiology
- Baxt W.G
- Mair J,
- Artner-Dworzak E,
- Lechleitner P,
- et al.
- Ikeda H,
- Takajo Y,
- Ichiki K,
- et al.
- Ushiyama S,
- Laue T.M,
- Moore K.L,
- et al.
- Tunstall-Pedoe H,
- Kuulasmaa K,
- Amouyel P,
- et al.
- Braunwald E,
- Mark D.B,
- Jones R.H,
- et al.
- Fleiss J.L
- Coller B.S
- Neumann F.J,
- Gawaz M,
- Dickfield T,
- et al.
- ↵Cannon CP, Ault K, Mitchell J, et al. P-selectin in patients post acute coronary syndromes treated with Sibrafiban, an oral IIb/IIIa antagonist: results from TIMI 12. Circulation 1997;96 Suppl 932:I-169.