Customize your JAMA Network experience by selecting one or more topics from the list below.
von Kodolitsch Y, Schwartz AG, Nienaber CA. Clinical Prediction of Acute Aortic Dissection. Arch Intern Med. 2000;160(19):2977–2982. doi:10.1001/archinte.160.19.2977
Copyright 2000 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2000
Clinical criteria for aortic dissection are poorly defined. Thus, 35% of aortic dissections remain unsuspected in vivo, and 99% of suspected cases can be refuted.
To identify independent predictors of acute aortic dissection and create a prediction model for facilitated estimation of the individual risk of dissection.
Two hundred fifty patients with acute chest pain, back pain, or both; absence of an established differential diagnosis of the pain syndrome; and clinical suspicion of acute aortic dissection were evaluated for the presence of 26 clinical variables in a prospective, observational study. Multivariate analysis was performed to create a prediction model of aortic dissection.
Aortic pain with immediate onset, a tearing or ripping character, or both; mediastinal widening, aortic widening, or both on chest radiography; and pulse differentials, blood pressure differentials, or both (P<.001 for all) were identified as independent predictors of acute aortic dissection. Probability of dissection was low with absence of all 3 variables (7%), intermediate with isolated findings of aortic pain or mediastinal widening (31% and 39%, respectively), and high with isolated pulse or blood pressure differentials or any combination of the 3 variables (≥83%). Accordingly, 4% of all dissections were assigned to the low-probability group, 19% to the intermediate-probability group, and 77% to the high-probability group of aortic dissection.
Assessment of 3 clinical variables permitted identification of 96% of the acute aortic dissections and stratification into high-, intermediate-, and low-probability groupings of disease. With better selection for prompt diagnostic imaging, this prediction model can be used as an aid to improve patient care in aortic dissection.
ACUTE AORTIC dissection is the most frequently fatal condition in the spectrum of chest pain syndromes.1,2 With undelayed diagnosis and optimal medical and surgical therapy, however, 30-day survival can exceed 90%.1-4 Noninvasive imaging modalities permit rapid and safe diagnosis of dissection even in the early stage of intramural hematoma.3,4 With poorly defined criteria for imaging of the aorta, however, 35% of dissections are not subjected to imaging,5,6 and more than 99% of suspected dissections can finally be refuted.7-9 The attempt of the American College of Emergency Physicians to define a policy for the initial approach to patients with chest pain recently highlighted the continuing need for clinical criteria for aortic dissection.10-12 To define predictors of dissection we analyzed 26 clinical variables in a prospective, observational study of 250 study patients before emergency imaging of the aorta. The analysis was intended to design a simple prediction model for more rapid estimation of the individual risk of dissection on arrival in the emergency department. Such models have been developed to improve coronary care unit admission practices in ischemic heart disease and might also permit better selection for prompt diagnostic imaging with improved survival after dissection.13
Three clinical criteria were applied to selected patients admitted to the emergency department for inclusion in the study group: (1) Patients with onset of chest pain, back pain, or both within 2 weeks of admission were subjected to history taking, physical examination, routine laboratory evaluation, and a 12-lead electrocardiogram performed on arrival in the emergency department. (2) Evidence of an acute coronary syndrome or another established diagnosis explaining the chest pain, back pain, or both was an exclusion criterion. In uncertain cases, additional diagnostic procedures were performed using a strategy similar to that described elsewhere14 before definite exclusion. (3) All other patients were included when jointly considered clinically suspicious of acute aortic dissection by 2 experienced emergency department physicians.11 Patients with a history of recent trauma, cardiovascular intervention, or gastrointestinal surgery and comatose patients with no history available were excluded.
All study patients were evaluated for 26 clinical variables, which were assessed unblinded to each other. A plain chest radiograph was obtained as frontal and lateral projections in 190 patients and as frontal bedside examination in 60 patients.15,16 After clinical evaluation, each study patient was subjected to contrast-enhanced computed tomography, magnetic resonance imaging, transesophageal echocardiography, or digital angiography. Diagnostic findings on these imaging procedures and findings at surgery or autopsy were used to establish the final diagnosis in each patient4 (Table 1).
Between January 1, 1988, and December 31, 1996, 41,495 consecutive patients presented to the emergency department of the University Hospital Eppendorf, Hamburg, Germany, with acute chest pain, back pain, or both; of these, 250 constituted the study group (77 women and 173 men; mean ± SD age, 54 ± 15 years). Dissection was confirmed in 128 study patients, including 16 diagnosed in the stage of intramural hematoma4; 105 patients had acute dissection with onset of symptoms within 48 hours and 23 had subacute dissection with onset of pain within 2 weeks of emergency department admission.2,4 Seventy-eight patients with dissection in the ascending aorta were identified as type A (61%) and 50 with no affliction of the ascending aorta were identified as type B (39%) dissections17 (Table 1). Chronic arterial hypertension was prevalent in 151 patients (60%), and Marfan syndrome was present in 11 (4%).18
Dissection was refuted in the other 122 study patients.4 The remainder of 41,245 patients with chest pain, back pain, or both (20,062 women and 21,183 men; mean ± SD age, 55 ± 16 years) did not meet the inclusion criteria. The final diagnoses in the 2 cohorts of the excluded 41,245 patients and in the 122 study patients with refuted dissection were acute coronary syndrome (24.4% vs 14.7%; P<.001); neuroradicular syndrome (17.1% vs 0.8%; P<.001); pulmonary disease (15.0% vs 0.8%; P<.001); cardiac arrhythmia (6.5% vs 0%; P<.001); hypotension with vagal reaction (6.2% vs 0%; P<.001); heart failure (5.0% vs 0%; P<.001); hyperventilation (3.3% vs 0%; P<.001); hypertensive crisis (2.2% vs 9.0%; P<.03); chest wall syndrome (1.8% vs 14.7%; P<.001); gastrointestinal diseases such as esophagitis, peptic ulcer, gastritis, and pancreatitis (1.2% vs 9.8%; P<.001); pneumothorax (0.6% vs 1.6%; P = .67); pulmonary embolism (0.4% vs 4.9%; P<.001); pleuritis (0.2% vs 4.0%; P<.001); and pericarditis (0.1% vs 5.7%; P<.001). In 5.9% and 3.3%, respectively, a definite diagnosis could not be established (P<.002), and in the remainder miscellaneous diagnoses were made.
The 26 most frequently reported clinical features in aortic dissection were chosen as study variables.5,7,15 Among these, onset of pain within less than 2 minutes was considered immediate; a trigger of pain was antecedent to strain of an emotional or physical nature15; and prodromal symptoms comprised giddiness, swaying, or apprehension before pain7,15 (Table 2). Intense severity of pain was present with the notion of excruciating, unbearable intensity,7,15 and tearing or ripping pain was considered with any tearing, ripping, or lacerating feature of pain, also described as "if something inside tore loose."2,7,15 Shifting of pain location was considered migratory, whereas inspiratory aggravation of pain was considered pleuritic.2,5,7,15 Episodic interruption of consciousness without focal neurologic defect was considered syncope; prolonged loss of consciousness or coma and focal neurologic signs were defined by standard criteria.2,5,7,15,19 Systolic blood pressure greater than 150 mm Hg on hospital admission was considered hypertension and less than 100 mm Hg was considered hypotension; a difference in systolic blood pressure of more than 20 mm Hg between both arms was a blood pressure differential. Absence of a proximal extremity or carotid pulse was considered a pulse differential.20 Acute renal failure was diagnosed in the presence of anuria, oliguria, and/or increased creatinine levels with a systolic blood pressure greater than 100 mm Hg and no history of chronic renal failure. A murmur of aortic regurgitation was detected by auscultation.15 Pulsating abdominal mass and abdominal boardlike tenseness were considered abdominal signs.15 Mediastinal and/or aortic widening (including the aortic knob) and pleural effusion on chest radiographs were diagnosed at the discretion of a board-certified radiologist. Electrocardiographic left ventricular hypertrophy, previous Q-wave infarction, and acute myocardial infarction syndrome were assessed by standard criteria.21-23
Logistic regression analysis was performed with only 0.6% missing data.24 Risk of aortic dissection associated with the presence of each sign or symptom was initially evaluated using univariate, unconditional logistic regression models and statistical software.25,26 Risk factors significant for predicting risk at the 5% level were evaluated for inclusion in a multivariate model using stepwise procedures. To simplify the final model, a new variable was created in which anyone with pain of immediate onset or of a tearing or ripping nature (both variables were the only pain variables, which were independent predictors of aortic dissection) was labeled "aortic pain." The final model obtained from the stepwise procedure was then refitted using this simplified pain variable. Estimates of risk (odd ratios) were calculated based on coefficients from the logistic models. Analyses were repeated for patients with type A and type B dissection. Probability of dissection was calculated as the number of patients with dissection and the clinical feature(s) divided by the number of patients presenting with the clinical feature(s). Comparison between groups and characteristics was performed using the Fisher exact test; significance was set at 5%.27 Data are given as mean ± SD.
A total of 41,495 consecutive admissions to the emergency department fulfilled criterion 1 of acute chest pain, back pain, or both as described in the "Patients and Methods" section; in 38,819 patients (93.5%), aortic dissection was excluded by establishing alternative diagnoses, as defined by criterion 2. Using criterion 3, 2 experienced emergency department physicians agreed on the exclusion of an additional 2426 patients (5.8%) and shared a clinical suspicion of acute aortic dissection in the remaining 250 patients eligible for our study (0.6%) (Table 1).
Study patients presented to the emergency department 49 ± 44 hours after initial onset of symptoms. Clinical evaluation and emergency diagnostic imaging were completed within 1.8 ± 1.5 hours, confirming aortic dissection in 128 patients (51%) and excluding dissection in 122 (49%). At the time of initial presentation, no patient had electrocardiographic signs of acute myocardial infarction; 5 patients, however, developed new ST-segment elevation (4%). Five patients with dissection were falsely diagnosed as having acute coronary syndrome: 1 survived fibrinolytic therapy and the other 4 died and were diagnosed at autopsy.
Univariate analysis identified 13 variables associated with acute dissection (Table 2). The final model included aortic pain with immediate onset of pain, tearing or ripping pain, or both (odds ratio, 6.41; 95% confidence interval, 3.06-13.46); mediastinal widening, aortic widening, or both (odds ratio, 57.86; 95% confidence interval, 7.04-475.47); and pulse differentials, blood pressure differentials, or both (odds ratio, 10.19; 95% confidence interval, 4.92-21.11; P<.001 for all) as independent predictive variables of acute aortic dissection. Each of these 3 variables was confirmed as an independent predictor separately for both subtypes of dissection. In addition, multivariate analysis revealed that a murmur of aortic regurgitation predicted type A (P<.003) and excluded type B (P<.005) dissection.
Furthermore, the probability of dissection was stratified according to the presence of its 3 clinical predictors. In the absence of all 3 predictors, the probability of dissection was 7% and thus low. The isolated findings of aortic pain and mediastinal widening, aortic widening, or both were associated with intermediate probabilities of dissection of 31% and 39%, respectively. The combinations of aortic pain with mediastinal widening, aortic widening, or both or with pulse differentials, blood pressure differentials, or both were associated with high probabilities of dissection (83% and 92%, respectively). All other combinations and isolated pulse differentials, blood pressure differentials, or both revealed 100% probabilities (Table 3). Accordingly, based on 3 clinical variables, 4% of dissections would have been considered at low risk, 19% at intermediate risk, and 77% at high risk for dissection.
This prospective, observational study of a selected group of patients with acute chest pain, back pain, or both identifies aortic pain; pulse differentials, blood pressure differentials, or both; and mediastinal widening, aortic widening, or both on chest radiography as independent predictors of acute aortic dissection. These variables permit identification of 96% of all acute dissections and rapid stratification according to high, intermediate, or low probability of disease. With better selection for prompt diagnostic imaging, this prediction model can be used as an aid to improve survival after dissection.
Selection of candidates for imaging of the aorta depends on the experience of the attending physician and his or her clinical impression,7 which was also used as a criterion for enrollment in the present study. Our data confirm that in specialized cardiovascular centers, such subjective clinical criteria might yield diagnostic confirmation of dissection in 50% of suspected cases.28,29 Identification of dissection in unselected outpatients, however, is much less efficient.9 Thus, use of defined clinical criteria for aortic dissection seems particularly helpful when dealing with unselected patients in settings without specialized clinical experience or adequate diagnostic or therapeutic facilities available for the management of potential acute aortic disease.7-9
Patients without pain were not included in our study group. However, all documented cases of so-called painless dissection revealed atypical chest pain along with other signs of dissection.30,31 Because such patients were included, we believe that our study group was not biased by the absence of completely painless dissections.
Chest pain is consistently reported in more than 90% of patients with aortic dissection.7,15 Its immediate, catastrophic onset of a ripping, tearing, and migrating nature is reported in more than 80% of dissections.15,28 Our study confirms the high incidence of pain and identifies 2 variables—immediate onset and a ripping nature—characteristic of aortic pain. Widening of the mediastinum and the aortic knob has previously been found in more than 90% of all dissections.5,7,16 Mediastinal widening, aortic widening, or both were associated with 39% probability of dissection in this study; combined with aortic pain or pulse differentials, blood pressure differentials, or both, probability increased to 83% and 100%, respectively. Widening of the mediastinal, aortic shadow, or both in patients with normal aortas were not more prevalent in frontal bedside examinations than in combined frontal and lateral projections (Table 2).
Twenty-eight percent of patients revealed signs of limb ischemia and 38% had pulse differentials with a probability of dissection greater than 90%.7,20,32 Our data confirm a high prevalence of aortic regurgitation in type A dissection, emphasizing the usefulness of diastolic aortic murmurs to distinguish proximal from distal dissection.7,15 The scarce occurrence of malperfusion signs disqualifies them as independent predictors of dissection. However, focal neurologic signs and renal signs were found exclusively in patients with dissection and thus might be used as additional diagnostic clues of dissection.20,33
In 21 study patients with exclusion of dissection (17%), other aortic disorders, such as penetrating aortic ulcer and unstable nondissecting aneurysms with rapid expansion, rupture, or severe aortic regurgitation, were diagnosed, more than half of which required immediate surgical intervention. Chest pain, mediastinal widening, and murmurs of aortic regurgitation make such patients difficult to separate from those with acute dissection1,8,17,28 (Table 4).
Moreover, in a setting of acute chest pain, back pain, or both, acute coronary syndromes outweigh the frequency of acute aortic dissection by 80:1, with their diagnostic separation failing in 0.01%.34 In some rare cases, dissection with peripheral malperfusion might be confused with arterial embolism caused by acute myocardial infarction.32 More important, 4% of acute aortic dissections present with electrocardiographic signs of acute myocardial infarction34; fibrinolysis was administered to 0.4% of patients with aortic dissection and 2% of reported dissections received fibrinolytic therapy, with lethal outcome in 64%.34 In our study, 5 patients with aortic dissection were misdiagnosed as having acute coronary syndromes, with atypical ST-segment elevations due to left ventricular hypertrophy in 2 and typical ST-segment elevations caused by aortic dissection with coronary artery involvement in 3. Thus, clinical evaluation for signs and symptoms of dissection in patients even with electrocardiographic evidence of acute myocardial infarction might avoid confusion of both diseases.
Assessment of risk is likely to help (1) administer the appropriate initial medical therapy, (2) choose the optimal imaging modality, (3) assess the urgency for definite diagnosis, and (4) define the need for potential emergency intervention. In patients at high risk for dissection, initiation of intravenous therapy with β-adrenergic blocking agents, precautions for surgical intervention, and diagnostic confirmation have priority, and transportation to a tertiary care center is warranted for adequate imaging.35 Patients with isolated aortic pain or mediastinal widening are at intermediate risk for dissection and should be treated as high-risk patients.
In patients with a low probability of dissection, the priority of diagnostic evaluation might be shifted to other diseases and imaging of the aorta conducted later in the differential workup. Knowledge of rare and variable alternative signs of dissection might help individualize the risk in each patient, and it should be emphasized that our prediction model is designed for more rapid identification of patients with dissection rather than as an instrument for clinical rule out.
Accepted for publication May 18, 2000.
This study was supported by grants KO 1828/1 and KO 1828/1-2 from the Deutsche Forschungsgemeinschaft, Bonn, Germany (Dr von Kodolitsch).
We thank Kim Eagle, MD, University of Michigan Hospital, Ann Arbor; Markus Fliegner, MD, Marien Hospital, Hamburg, Germany; Frauke Janssen, MD, and Michael Raghunath, MD, Westfälische Wilhelms University, Münster, Germany; and Nathaniel Reichek, MD, Walter J. Rogers, MD, and Reed E. Pyeritz, MD, the MCP Hahnemann School of Medicine, Pittsburgh, Pa, for their thoughtful and stimulating comments; and Marina Vogler, the German Marfan Foundation, Cologne, for discussion of the manuscript.
Reprints: Christoph A. Nienaber, MD, Department of Cardiology, University Hospital Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany (e-mail: email@example.com).