Submassive pulmonary embolism (PE) manifests a. Assessment of Cardiac Stress From Massive Pulmonary Embolism With 12-Lead ECG*

Transcription

1 Assessment of Cardiac Stress From Massive Pulmonary Embolism With 12-Lead ECG* Kurt R. Daniel, DO ; D. Mark Courtney, MD; and Jeffrey A. Kline, MD Background: Massive pulmonary embolism (PE) that causes severe pulmonary hypertension can produce specific ECG abnormalities. We hypothesized that an ECG scoring system would vary in proportion to the severity of pulmonary hypertension and would help to distinguish patients with massive PE from patients with smaller PE and those without PE. Methods: A 21-point ECG scoring system was derived (relative weights in parentheses): sinus tachycardia (2), incomplete right bundle branch block (2), complete right bundle branch block (3), T-wave inversion in leads V 1 through V 4 (0 to 12), S wave in lead I (0), Q wave in lead III (1), inverted T in lead III (1), and entire S 1 Q 3 T 3 complex (2). ECGs obtained within 48 h prior to pulmonary arteriography were located for 60 patients (26 positive for PE, 34 negative for PE) and for 25 patients with fatal PE. Results: Interobserver agreement (11 readers) for ECG score was good (Spearman r 0.74). The ECG score showed significant positive relationship to systolic pulmonary arterial pressure (spap) in patients with PE (r 0.387, p < 0.001), whereas no significant relationship was seen in patients without PE (r 0.08, p 0.122). When patients were grouped by severity of pulmonary hypertension (low, moderate, severe), only patients with severe pulmonary hypertension from PE had a significantly higher ECG score (mean, ). At a cutoff of 10 points, the ECG score was 23.5% (95% confidence interval [CI], 16 to 31%) sensitive and 97.7% (95% CI, 96 to 99%) specific for the recognition of severe pulmonary hypertension (spap > 50 mm Hg) secondary to PE. In 25 patients with fatal PE, the ECG score was Conclusions: The derived ECG score increases with severity of pulmonary hypertension from PE, and a score > 10 is highly suggestive of severe pulmonary hypertension from PE. (CHEST 2001; 120: ) Key words: clinical; decision making; decision support systems; ECG; heart arrest; pulmonary embolism; pulmonary heart disease Abbreviations: CI confidence interval; ED emergency department; PE pulmonary embolism; ROC receiver operating characteristic; RV right ventricular; spap systolic pulmonary arterial pressure; V /Q ventilation/ perfusion Submassive pulmonary embolism (PE) manifests a wide spectrum of nonspecific changes on surface 12-lead ECG, ranging from no abnormality to multiple disturbances of rate, rhythm, and conduction pattern. 1,2 As a result, acceptable ECG criteria have been difficult to define that can exclude or diagnose *From the Oklahoma State University College of Osteopathic Medicine (Dr. Daniel), Tulsa, OK; and the Department of Emergency Medicine (Drs. Courtney and Kline), Carolinas Medical Center, Charlotte, NC. Dr. Daniel is currently at the Bowman-Gray School of Medicine, Dept of Internal Medicine, Wake Forest University, Winston-Salem, NC. Manuscript received October 13, 2000; revision accepted February 1, Correspondence to: Jeffrey Kline, MD, Assistant Director of Research, Department of Emergency Medicine, Carolinas Medical Center, PO Box 32861, Charlotte, NC 28232; Jkline@ carolinas.org PE with reasonable certainty. Instead, the ECG is frequently relegated to the role of screening for other cardiac abnormalities in patients with PE, 3 or to the whim of academic curiosity about whether certain textbook findings of PE are present (eg, the McGinn-White S 1 Q 3 T 3 pattern). However, right ventricular (RV) strain manifested on the ECG may have more than trivial importance, inasmuch as this evidence might predict the severity of pulmonary hypertension, 4 which appears to correspond to the presence of RV dysfunction with PE. 5 At present, to our knowledge, no published evidence has indicated how primary-care physicians can use the ECG to stratify the severity of PE. The purpose of the present report was to develop a structured ECG score based on ECG criteria that have previously been associated with RV strain from 474 Clinical Investigations

2 PE, and to test this scoring system with multiple emergency physicians using ECGs from patients who underwent pulmonary angiography, or who died from acute PE. We sought to test two hypotheses: (1) that a structured ECG score would vary positively with the pulmonary arterial pressure in patients with PE, and (2) that a structured ECG score obtained from patients with severe pulmonary arterial hypertension induced by PE would be higher than patients without pulmonary hypertension, and higher than patients without PE. Materials and Methods To derive the ECG scoring system, electronic databases (MEDLINE, EMBASE) were searched using the exploded key words pulmonary embolism and electrocardiography from 1966 to September Original full-length clinical studies were reviewed and were examined for reports that included the frequency of ECG findings associated with PE, together with data to assess the severity of PE based on results of pulmonary angiography, echocardiography, or autopsy. 1 4,6 9 Studies were then limited to those reports in which the authors defined patients with massive or severe PE from submassive PE. 1,3,4,7 Data from these published studies were aggregated and used to rank the frequency of seven major abnormalities associated with massive PE in the following order: (1) T-wave inversion in leads V 1 through V 4, (2) incomplete or complete right bundle branch block, (3) sinus tachycardia, (4) the S 1 Q 3 T 3 pattern, (5) right-axis deviation, (6) left-axis deviation, and (7) P-pulmonale. No study was found that directly compared the frequency of these ECG findings between patients with PE proven by pulmonary angiography and patients without PE. The study aim was to devise a scoring system that could be completed by primary-care physicians with reasonable accuracy within 2 min. A pilot study was therefore performed with three emergency physicians who interpreted 11 ECGs from patients with PE to examine for speed and accuracy in interpreting these criteria. Based on this pilot study, the ECG criteria were narrowed to four abnormalities (Fig 1). We eliminated several findings commonly associated with PE because the physicians all had difficulty interpreting the criteria, including P-pulmonale, and right-axis or left-axis deviation. ECGs were copied without the computer interpretation for the reader to see. The relative contribution of the four abnormalities to the overall score is shown in parentheses: sinus tachycardia (2); incomplete right bundle branch block (2); complete right bundle branch block (3); T-wave inversion, graded by magnitude (V 1 [0 to 2], V 2 [1 to 3], V 3 [1 to 3], V 1 through V 4 all inverted 2 mm [4]); and components of the S 1 Q 3 T 3 complex (S wave in lead I [0], Q wave in lead III [1], inverted T wave in lead III [1], and the entire S 1 Q 3 T 3 complex [2]). The maximum score was 21 points. The relative weight of each criterion was set on the basis of the expected importance of each abnormality in the distinction of patients with massive PE from patients with submassive PE or no PE. 1,4 To obtain ECGs from patients with suspected and confirmed PE, medical records from two teaching hospitals (Carolinas Medical Center, Charlotte, NC, and Tulsa Regional Medical Center, Tulsa, OK) were searched for two groups of patients. Group A comprised patients who underwent pulmonary arteriography for suspected PE from 1997 to 1999 and had pulmonary arterial pressures measured; group B comprised patients not in group A who died from acute PE. The study protocol was reviewed and approved by the Institutional Review Board at both hospitals prior to data collection. Patients in group A were included only if a legible ECG was obtained within 48 h prior to pulmonary angiogram. Cases were excluded if the clinical impression and the interpretation of the pulmonary angiogram indicated chronic PE. When multiple ECGs were recorded, the tracing that was temporally closest to the pulmonary angiogram was used. Patients in group B died from acute PE based on one (or more) of the following criteria: (1) massive PE was established as the cause of death by autopsy; or (2) pulmonary angiography showed clear evidence of massive PE within 1 week prior to death; or (3) a ventilation/perfusion (V /Q ) scan was performed within 1 week prior to death, and the final interpretation was high probability for PE, and a formal echocardiographic study was also performed which demonstrated evidence of RV strain (enlarged right ventricle, abnormal septal motion, moderate or severe tricuspid regurgitation, or RV hypokinesis). An ECG had to be obtained after the initial symptoms and signs of PE. If multiple ECGs were obtained prior to death, two ECGs were copied: the one most proximate to the time of symptom onset (eg, the ECG obtained in the emergency department [ED]), and the last ECG obtained prior to death. Clinical data were abstracted from the chart, the ECGs were each given an identifier numeral, and clinical data from each subject were entered into a database (Excel 97; Microsoft; Seattle, WA). The ECG for each case was labeled with the identifier numeral, and ECG tracings were distributed to the ECG readers with a reading sheet (Fig 1) attached to each ECG. Eleven physicians were selected to serve as ECG readers. These individuals were chosen to represent a diverse spectrum of primary-care emergency medicine practitioners from two academic emergency medicine residencies, including seven attending emergency physicians with an average of 6 years of postresidency experience and four postgraduate year 3 emergency medicine residents. The readers were given instructions to complete the reading of all 60 cases within 90 min. A written cover letter instructed the readers that each ECG comes from a patient in whom you suspect pulmonary embolism. The readers were blinded to all other clinical data, including the diagnosis of each subject. A scoring sheet was completed for each ECG. Score sheets had a check box for each criterion, but did not include the relative values of the ECG findings. Data from each score sheet were entered into the computer spreadsheet for calculation of the ECG scores and for statistical analysis. ECG scores were computed using the derived scoring system as detailed above; no adjustment was made to the scoring system after the data from readers were obtained. Statistical Analysis The specific aim was to test if the ECG score would increase in proportion to the pulmonary artery pressure in patients with PE. Because the ECG scores in group A were not normally distributed, (Kolmogorov-Smirnov distance, 0.183; p 0.001; Sigma- Stat v2.0; Jandel Scientific; San Rafael, CA), data were compared with nonparametric statistical tests. Accordingly, a Spearman rank correlation coefficient, r, is reported for the regression plot of ECG score as a function of pulmonary arterial pressure. Interobserver agreement was also examined using r values. Cases were classified into three classes of severity based on the systolic pulmonary arterial pressure (spap): low, 0 to 30 mm Hg; moderate, 31 to 49 mm Hg; and severe, 50 mm Hg. ECG scores between PE-negative and PE-positive categories were compared using Mann-Whitney rank sum test or analysis of variance on ranks with Dunn s post hoc comparison when more than two groups were compared; p 0.05 was considered significant. Area under the curve for the receiver operating character- CHEST / 120 / 2/ AUGUST,

3 Figure 1. The ECG scoring system used by 11 emergency physicians to grade 60 ECGs. The actual score sheet used by the physicians did not show the relative weights of each component. The relative weight of each component was based on data in previous publications and a pilot study done by the authors. SI QIII TIII S 1 Q 3 T 3 ; Max maximum. istic (ROC) was calculated by the trapezoidal rule, and SEM was calculated using the method of Hanley and McNeil. 10 For standard diagnostic indexes, 95% confidence intervals (CIs) were calculated using the Confidence Interval Analysis program. 11 The ECG scores from group B were found to be normally distributed; therefore, a paired t test was used to compare the first ECG score to the last ECG score obtained prior to death. Results For group A, 60 cases were identified. Fifty-eight patients underwent a V /Q scan performed prior to the pulmonary angiography. Prior V /Q scans were read as low probability for PE in 10 patients, intermediate probability in 38 patients, indeterminate probability in 7 patients, and high probability in 3 patients. The mean SD time after symptom onset to the time of the ECG used for scoring was h. Six patients had symptoms that began 1 week prior to the first ECG, but complained of worsening symptoms within the previous 72 h. Pulmonary angiogram findings were read as positive for PE in 26 patients and negative for PE in 34 patients 476 Clinical Investigations

4 (probability of PE, 43.3%; 95% CI, 30 to 57%). In the 26 patients with PE, prior V /Q scan findings were read as low probability in 4 patients, moderate probability in 12 patients, indeterminate probability in 5 patients, and high probability in 3 patients. Two of the 26 patients with PE underwent no V /Q scanning prior to pulmonary angiography. Subjects with PE did not differ from subjects without PE in age (mean, years) or in the frequency of preexisting cardiopulmonary disease. No patient with PE had a documented history of myocardial infarction, whereas one patient without PE had a history of myocardial infarction. Three patients with PE had a history of COPD, and four patients without PE had a history of COPD. Average spap for subjects with PE was mm Hg vs mm Hg for subjects without PE (p 0.05, unpaired t test). For group B, 25 patients (14 female patients) were found to meet inclusion criteria (all from Carolinas Medical Center). The initial symptoms and signs of PE included one or more of syncope or seizure, dyspnea, hypoxemia, tachypnea, tachycardia, or overt cardiogenic shock. Documentation indicated strong antemortem suspicion or diagnosis of PE in 24 of 25 cases. Eighteen patients had massive bilateral PE demonstrated on autopsy, 4 patients had massive bilateral pulmonary arterial obstruction found on pulmonary angiography, and 3 patients had high-probability V /Q scan finding plus echocardiographic evidence of RV dysfunction prior to death. The mean age in group B was years. Paired ECGs (one at the time of initial symptoms, and the last ECG recorded prior to death) were available for 18 patients in group B. The mean time after symptom onset to the first ECG was h. The mean time interval between these 18 ECGs was h (median, 26.6 h; range, 2 to 192 h). The mean time interval between the last ECG obtained and death was h (median, 15.5 h; range, 45 min to 240 h). Sixteen patients in group B initially presented with symptoms of PE to the ED that raised enough suspicion to warrant testing for PE. Thirteen of the 16 ED patients died within 24 h of presentation to the ED, 4 of whom experienced cardiac arrest while being transferred to have a V /Q scan performed. The ECG scores were first examined for the degree of interobserver variability among the 11 readers. In group A, 11 readers produced 60 scores, for a total of 660 scores. The mean of all 660 scores was (median score, 3; first to third interquartile range, 0 to 5). Assessment of interrater reliability was determined for all 11 readers by calculating r values for all possible reader pairs (55 unique r-value calculations). With this comparison, r values calculated for all possible pairs of observers ranged from 0.55 to 0.89 with an average r of 0.74 for all 11 observers, indicating good overall agreement. 12 Table 1 shows that only two components of the ECG score were found with a significantly higher frequency in patients with PE and that no ECG component was markedly more frequent in patients with PE compared to patients without PE. To test the first hypothesis (that the ECG score would vary positively with the severity of pulmonary arterial pressure in patients with PE), group A was divided into patients without PE (n 34) and with PE (n 26). The ECG score was then plotted as a function of either the systolic or the mean pulmonary arterial pressure for patients without PE; no significant positive relationship was observed (r 0.08, p 0.122). Figure 2, top, demonstrates that the ECG score does not vary significantly with the spap. A similar plot was observed for the ECG score vs the mean pulmonary arterial pressure in patients without PE (r 0.195, p 0.281; data not shown). However, when the ECG score was plotted as a function of the spap for patients with PE (Fig 2, bottom), a significant positive relationship was observed between the ECG score and the spap (r 0.387, p 0.001). A similar finding was observed for the regression of the ECG score vs the mean pulmonary arterial pressure in patients with PE (r 0.236, Table 1 Frequency of ECG Findings by 11 Emergency Medicine Physicians Who Were Not Aware of Final Diagnosis, But Were Told That PE Was Suspected* Characteristics PE PE Tachycardia Incomplete right bundle branch block Complete right bundle branch block 5 4 T-wave inversion in all leads V 1 through V T-wave inversion in lead V 1 1 mm mm mm T-wave inversion in lead V 2 1mm mm mm 4 6 T-wave inversion in lead V 3 1mm mm mm 5 6 S wave in lead I Q wave in lead III Inverted T wave in lead III All of S 1 Q 3 T *Data are presented as %. Sixty ECGs were reviewed, 34 from patients without PE (PE ) and 26 from patients with PE (PE ). Significant by 95% CI for difference. CHEST / 120 / 2/ AUGUST,

5 p 0.001) from the other 48 patients. Among patients without PE, Figure 3, top, shows that the ECG score did not show a stepwise increase with the severity of systolic pulmonary hypertension. In contrast, among patients with PE, the ECG score did significantly increase in relation to the severity of pulmonary hypertension in subjects with PE. Figure 3, bottom, shows that the ECG score increased in a statistically significant stepwise fashion. As an extension of the second hypothesis, we sought to test if the ECG scoring system could distinguish patients with severe pulmonary hypertension induced by PE from other patients in group A. To accomplish this objective, a ROC curve was constructed to show how the ECG score performed at each of 21 cutoff points, considering the 12 Figure 2. Plots of ECG scoring as a function of the spap. Each dot represents the mean ECG score of 11 readers for each patient. Top: plot for 34 patients without PE. Bottom: plot for 26 patients with PE. Correlations tested with Spearman rank order test, r values. p 0.001; data not shown). The ECG score and spap were also evaluated for the 10 patients with low-probability V /Q scan readings: for the six patients with normal pulmonary angiogram findings, the ECG score was and the spap was mm Hg; for the four patients with positive pulmonary angiogram findings, the ECG score was and the spap was mm Hg. To test the second hypothesis, group A was subdivided into three categories, based on the spap: low, 0 to 30 mm Hg; moderate, 31 to 49 mm Hg; severe, 50 mm Hg. Patients without PE were distributed as follows: low (n 13), moderate (n 16), and severe (n 5). Patients with PE were distributed as follows: low (n 7), moderate (n 7), and severe (n 12). The mean ECG score of the 12 patients with PE and severe pulmonary hypertension was , which was significantly higher than the mean ECG score ( ; Figure 3. Mean ECG scoring grouped according to severity of pulmonary hypertension (low spap, 0 to 30 mm Hg; moderate spap, 31 to 49 mm Hg; severe spap, 50 mm Hg). Top: data for patients without PE (PE ). Bottom: data for patients with PE (PE ). Data compared with analysis of variance on ranks, with pairwise Dunn s test as post hoc. *p 0.05 considered significant. 478 Clinical Investigations

6 patients with PE and severe pulmonary hypertension as the true-positives and all others in group A as true-negatives (Fig 4). The area under the ROC curve (62 12%, calculated from the trapezoidal rule) suggests that the ECG score had a marginal overall diagnostic performance. An optimal cutoff was chosen at a score of 10 points to preserve a very high specificity with the best possible sensitivity. At this cutoff, a 2 2 diagnostic table was constructed for the ECG score compared to the results of pulmonary angiography (Table 2). The ECG scoring system demonstrated a sensitivity of 23.5% (95% CI, 16 to 31%) and a specificity of 97.7% (95% CI, 96 to 99%) [likelihood ratio positive, 11.1] for the detection of severe pulmonary hypertension from PE in group A. These indexes were derived from the 660 observations made by 11 ECG readers who evaluated 60 ECGs (43% with PE). In group B (25 patients with fatal PE), the mean ECG score for the last ECG prior to death was (median, 10). For the seven patients from whom only one ECG was obtained prior to death, those ECG scores were considered the last ECG score. In the 18 patients with two ECGs, when the first ECG score ( ; median, 4.5) was compared to the last ECG score ( ; median, 10), the increase was significant (p 0.001, paired t test). Thirteen of the last ECGs (52%) obtained prior to death had an ECG score of 10. All 13 patients Table 2 Diagnostic Performance of the ECG Score at a Cutoff of 10 Points Compared to the Results of Pulmonary Angiography* ECG Scores Results of Pulmonary Angiography Severe spap (PE ), No. Low/Moderate spap (PE ) orpe, No *Results of all 660 scores obtained by 11 readers for the study group of 60 patients (24 patients with PE, and 34 patients without PE. Sensitivity 23.5% (95% CI, 16.3 to 30.7); specificity 97.7% (95% CI, 96.1 to 98.8); likelihood ratio positive 11.1, where likelihood ratio positive (sensitivity/[1 specificity]). died within 24 h of obtaining the ECG tracing. The most frequently observed changes between the first and second ECG were the development of an incomplete or complete right bundle branch block (n 8), the progression of a partial S 1 Q 3 T 3 pattern to its complete form (n 6), and the progression of an incomplete right bundle branch block to complete right bundle branch block (n 3). Eight of 18 patients developed new anterior T-wave inversion in more than one precordial lead. Three patients developed preterminal bradycardia on ECG and therefore lost the points that had been contributed by the previous presence of tachycardia. Discussion Figure 4. ROC curve for the ability of the ECG score to distinguish patients with PE who had a spap of 50 mm Hg from patients with PE who had a spap of 50 mm Hg, or patients without evidence of PE on pulmonary angiography. This curve shows how the true-positive rate increases as the falsepositive rate increases. The curve was made by plotting the true-positive rate and false-positive rate at each of 21 cutoff points, based on the 21-point ECG scale. Because the truepositive rate and false-positive rate were identical at several cutoff points, the figure appears to show only 15 points. The area under the curve is 62 12%. This study demonstrates a positive relationship between an ECG scoring system and the severity of pulmonary arterial pressure in subjects with PE demonstrated on pulmonary angiography. The 21- point ECG scoring system was derived from previous work that demonstrated the association of each element of the scoring system with pulmonary hypertension from PE, including increased heart rate 2,7 ; impairment in right-sided cardiac conduction, manifested as various degrees of right bundle branch block 3,6,7 ; anterior T-wave inversions 1,4,13,14 ; and displacement of the QRS axis, manifested as various degrees of the classic S 1 Q 3 T 3 pattern. 1,15 This ECG scoring system was rapidly assessed by 11 primary-care emergency physicians, and the interobserver agreement was adequate to suggest that the ECG scoring system will be reproducible among different observers when it is tested in prospective fashion. At a cutoff point of 10 points, the ECG scoring system was highly specific for the detection of PE with severe pulmonary hypertension, and the ECG score was 10 points in 52% of patients with fatal PE. These data support the hypothesis that pulmonary CHEST / 120 / 2/ AUGUST,

7 hypertension induced by acute PE causes a set of pathophysiologic processes that alter the surface ECG in a specific pattern, which can be reasonably well modeled by an explicit ECG score. First, the regression data indicate that the ECG scoring system increases significantly in relation to the pulmonary arterial pressure in patients with PE, whereas no positive relationship was observed in patients without PE. Second, when patients were grouped according to severity of pulmonary hypertension, a clearly positive relationship was observed between the ECG score and the severity of pulmonary hypertension for patients with PE, but this finding was not evident in patients without PE. These observations suggest that when RV pressure is elevated by acute severe PE, that right-sided cardiac conduction and repolarization are affected to a greater degree than when RV pressures are chronically elevated. The present data do not allow a mechanistic explanation for this observation, but previous animal models have shown that experimental acute pulmonary hypertension can produce RV subendocardial ischemia. 16 The present study included five subjects without PE who had severe elevation of pulmonary arterial pressures all secondary to chronic diseases: three patients with severe COPD, one patient with primary pulmonary hypertension, and one patient with obesity-hypoventilation syndrome. Patients with suspected chronic PE were excluded in the present study. As a point of speculation, it is possible that patients who were categorized as having severe PE were the only patients in group A who experienced rapid and severe enough increase in right-sided pressures to produce RV subendocardial ischemia. The data also support the hypothesis that the ECG score can assist the clinician in the rapid recognition of severe pulmonary hypertension from PE. Toward this goal, this study primarily assessed the performance and reproducibility of a structured ECG score, rather than the frequency or importance of certain findings of PE on ECG. At the central hospital for this study (Carolinas Medical Center), most fatal PE cases were initially suspected or diagnosed by emergency physicians, and most of these patients died rapidly. As such, emergency physicians can play a pivotal role in the management of massive PE, especially if they can rapidly and accurately recognize the presence of cardiac stress from severe PE. We reasoned that if the ECG has any value in this process, then it was more important to show that a group of emergency physicians would all interpret ECGs in a similar way, rather than have a specialist interpret the tracing and enumerate the specific abnormalities. This study design therefore helps to show that a competent primary-care provider can accurately use the derived ECG score. In particular, an ECG score 10 was 97.7% specific and 23.5% sensitive for the presence of severe pulmonary hypertension caused by a PE. These diagnostic indexes were derived from group A, which had a pretest probability of PE (43%) that could be considered typical of outpatients with V /Q scan findings that are read as nondiagnostic for PE (either low or intermediate probability), but who elicit a high enough clinical suspicion to warrant pulmonary angiography. 17 Thus, a high ECG score could be particularly useful in the ED patient in whom PE is suspected, but when imaging procedures are either unavailable, or noninvasive imaging study results are inconclusive. A high ECG score used together with clinical probability assessment for PE based on other criteria 18 could help drive the decision to pursue more invasive testing for PE. The data from group B also show that the ECG scores were high in patients with fatal PE, and that the ECG score increased as the severity of cardiac stress intensified before cardiac arrest. In these 18 patients, the dominant ECG manifestation of impending cardiac failure was the development of a new incomplete or complete right bundle branch block, or the development of new anterior T-wave inversion. These findings point out the potential importance of using serial ECGs, together with other clinical criteria, to alter the clinical probability of massive PE. In group B, four ED patients all with ECG scores 10 suffered cardiac arrest while waiting for pulmonary vascular imaging. In all cases, physicians had already documented a high suspicion for PE; in all four patients, heparin infusion was the only treatment. This study has several limitations. First, readers were not tested prospectively to determine the accuracy of the ECG score when it is assessed in the clinical setting. We are currently performing a prospective study to determine the ability of a high ECG score to predict two findings: (1) the presence of RV dysfunction on echocardiography in patients with PE diagnosed by either V /Q scan or contrast-enhanced spiral CT scan, and (2) the frequency of PE proven by pulmonary angiography in patients with a nondiagnostic V /Q scan or a finding on contrast-enhanced spiral CT that is negative for PE. This prospective phase will also better define the best cutoff point for the ECG score, and clarify if specific criteria should be given more or less weight in the scoring system. Post hoc analysis of the present data suggests that the incomplete right bundle branch block should have been given more weight, and that isolated T-wave inversion in V 1 should have been given less weight, given that the latter finding is often a normal variant. Nevertheless, when the ECG scoring system was adjusted in various ways post hoc, the area under the ROC curve did not increase. Another limitation to 480 Clinical Investigations

8 the study is that the subgroup without PE who had severe pulmonary hypertension was small, and no patient had a process which would have been likely to produce a rapid increase in pulmonary arterial pressure (eg, ARDS, air embolism, or fat embolism syndrome). It remains possible that when the ECG scoring system is evaluated in more subjects with severe pulmonary hypertension, but without PE, the specificity of the ECG scoring system may decrease. Conclusion Massive PE causes specific abnormalities on ECG that can be quantitated by an explicit scoring system. Massive PE causes a significant increase in the derived ECG score compared to that in patients without massive PE. References 1 Stein PD, Dalen JE, McIntyre KM, et al. The electrocardiogram in acute pulmonary embolism. Prog Cardiovasc Dis 1975; 17: Stein PD, Terrin ML, Hales CA. Clinical, laboratory, roentgenographic, and electrocardiographic findings in patients with acute pulmonary embolism and no pre-existing cardiac or pulmonary disease. Chest 1991; 100: McIntyre KM, Sasahara AA, Littman D. Relation of the electrocardiogram to hemodynamic alterations in pulmonary embolism. Am J Cardiol 1972; 30: Ferrari E, Inbert A, Chevalier T, et al. The ECG in pulmonary embolism: predictive value of negative T waves in precordial leads; 80 case reports. Chest 1997; 111: Ribeiro A, Lindmarker P, Juhlin-Dannfelt A, et al. Echocardiography Doppler in pulmonary embolism: right ventricular dysfunction as a predictor of mortality rate. Am Heart J 1997; 134: Smith M, Ray CT. Electrocardiographic signs of early right ventricular enlargement in acute pulmonary embolism. Chest 1970; 58: Sreeram N, Cheriex EC, Smeets JLRM, et al. Value of the 12-lead electrocardiogram at hospital admission in the diagnosis of pulmonary embolism. Am J Cardiol 1994; 73: Chia BL, Tan HC, Lim YT. Right sided chest lead electrocardiographic abnormalities in acute pulmonary embolism. Int J Cardiol 1997; 61: Sasahara AA, Hyers TM, Cole CM. The urokinase pulmonary embolism trial: a national cooperative study. Circulation 1973; 47(suppl 2): Hanley JA, McNeil BJ. The meaning and use of the area under the receiver operating characteristic (ROC) curve. Radiology 1983; 143: Gardner MJ, Altman DG. Calculating confidence intervals for proportions and their differences. In: Altman DG, MacHin D, Bryant TN, eds. Statistics with confidence. 1st ed. London, UK: BMJ, 1989; Karras DJ. Statistical methodology: II. Reliability and validity assessment in study design, part A. Acad Emerg Med 1996; 4: Yoshinaga T, Ikeda S, Nishimura E, et al. Serial changes in negative T wave on electrocardiogram in acute pulmonary thromboembolism. Int J Cardiol 1999; 72: Lui CY. Acute pulmonary embolism as the cause of global T-wave inversion and QT prolongation. J Electrocardiol 1993; 26: McGinn S, White PD. Acute cor pulmonale resulting from pulmonary embolism. JAMA 1935; 104: Gold FL, Bache RJ. Transmural right ventricular blood flow during acute pulmonary artery hypertension in the sedated dog. Circ Res 1982; 51: Perrier A, Desmarais S, Miron M-J, et al. Non-invasive diagnosis of venous thromboembolism in outpatients. Lancet 1999; 353: Wells PS, Anderson DR, Rodger M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the model s utility with the SimpliRED D-dimer. Thromb Haemost 2000; 83: CHEST / 120 / 2/ AUGUST,