Author + information
- Received March 29, 2011
- Accepted April 11, 2011
- Published online August 1, 2011.
- Doo Kyoung Kang, MD⁎,†,‡,
- Christian Thilo, MD⁎,†,§,
- U. Joseph Schoepf, MD⁎,†,⁎ (, )
- J. Michael Barraza Jr, BS⁎,†,
- John W. Nance Jr, MD⁎,†,
- Gorka Bastarrika, MD, PhD⁎,†∥,
- Joseph A. Abro, MA⁎,†,
- James G. Ravenel, MD†,
- Philip Costello, MD† and
- Samuel Z. Goldhaber, MD¶
- ↵⁎Reprint requests and correspondence:
Dr. U. Joseph Schoepf, Heart and Vascular Center, Medical University of South Carolina, Ashley River Tower, 25 Courtenay Drive, MSC 226, Charleston, South Carolina 29425
Objectives The purpose of this study was to compare the prognostic role of various computed tomography (CT) signs of right ventricular (RV) dysfunction, including 3-dimensional ventricular volume measurements, to predict adverse outcomes in patients with acute pulmonary embolism (PE).
Background Three-dimensional ventricular volume measurements based on chest CT have become feasible for routine clinical application; however, their prognostic role in patients with acute PE has not been assessed.
Methods We evaluated 260 patients with acute PE for the following CT signs of RV dysfunction obtained on routine chest CT: abnormal position of the interventricular septum, inferior vena cava contrast reflux, right ventricle diameter (RVD) to left ventricle diameter (LVD) ratio on axial sections and 4-chamber (4-CH) views, and 3-dimensional right ventricle volume (RVV) to left ventricle volume (LVV) ratio. Comorbidities and fatal and nonfatal adverse outcomes according to the MAPPET-3 (Management Strategies and Prognosis in Pulmonary Embolism Trial-3) criteria within 30 days were recorded.
Results Fifty-seven patients (21.9%) had adverse outcomes, including 20 patients (7.7%) who died within 30 days. An RVDaxial/LVDaxial ratio >1.0 was not predictive for adverse outcomes. On multivariate analysis (adjusting for comorbidities), abnormal position of the interventricular septum (hazard ratio [HR]: 2.07; p = 0.007), inferior vena cava contrast reflux (HR: 2.57; p = 0.001), RVD4-CH/LVD4-CH ratio >1.0 (HR: 2.51; p = 0.009), and RVV/LVV ratio >1.2 (HR: 4.04; p < 0.001) were predictive of adverse outcomes, whereas RVD4-CH/LVD4-CH ratio >1.0 (HR: 3.68; p = 0.039) and RVV/LVV ratio >1.2 (HR: 6.49; p = 0.005) were predictive of 30-day death.
Conclusions Three-dimensional ventricular volume measurement on chest CT is a predictor of early death in patients with acute PE, independent of clinical risk factors and comorbidities. Abnormal position of the interventricular septum, inferior vena cava contrast reflux, and RVD4-CH/LVD4-CH ratio are predictive of adverse outcomes, whereas RVDaxial/LVDaxial ratio >1.0 is not.
Right ventricular (RV) failure is the most common cause of early death among patients with acute pulmonary embolism (PE) (1–5). While bedside echocardiography remains the first-line test for diagnosing RV dysfunction (3,5–7), chest computed tomography (CT) has become the preferred imaging modality for PE diagnosis in stable patients, because it is accurate and available 24 h daily at most institutions (8,9). Several studies have evaluated chest CT signs for determining the severity of RV dysfunction and for predicting patient outcome (10–31), including flattening or displacement of the interventricular septum toward the left ventricle, reflux of contrast material into the inferior vena cava (IVC), and measurements of RV diameter (RVD) divided by left ventricular diameters (LVD) on transverse chest CT sections or reconstructed 4-chamber (4-CH) views.
However, because the ventricles have a complex 3-dimensional (3D) shape, 3D assessment of right ventricular volume (RVV) and left ventricular volume (LVV) may be more useful as a surrogate marker for RV dysfunction in patients with acute PE. Image post-processing methods enabling 3D volumetry of the ventricles have recently become available. A previous study (15) demonstrated that ventricular volume measurements at electrocardiogram (ECG)-gated chest CT were more sensitive for identifying patients with possible RV dysfunction than diameter measurements on non–ECG-gated chest CT. However, routine ECG-gating of chest CT studies for diagnosing PE is not currently recommended because of the additional radiation exposure involved with ECG-gated techniques (8,9) and because of the only limited incremental diagnostic improvement over routine non–ECG-gated chest CT (23).
Accordingly, the purpose of this current investigation was to compare, in patients with acute PE, the prognostic value for adverse outcomes and 30-day mortality of 3D volumetric ventricular analysis based on routine, non–ECG-gated chest CT with all other described chest CT signs of RV dysfunction.
From April 2007 through August 2009 we performed 1,606 contrast-enhanced chest CT studies for suspected PE, and 302 of these studies were interpreted as positive for new onset, acute PE. Of these, a total of 260 patients were included in the present analysis after 36 patients were excluded because of incomplete follow-up during the study period. Six patients were excluded because of insufficient contrast enhancement of the ventricular chambers for reliable delineation of the endocardial borders.
In all, 173 patients (66.5%) were inpatients already hospitalized at the time of acute PE diagnosis. Primary reasons for hospitalization and complications other than acute PE during the hospital stay included major surgery in 93 patients, trauma in 32, pneumonia in 29, stroke in 14, sepsis in 11, major hemorrhage in 9, myocardial infarction in 2, and multisystem organ failure in 1 patient.
For indicating imaging work-up of suspected PE, our institution generally follows the recommendations as outlined by the Christopher Study Investigators (32) and by Piazza et al. (33). All patients included in this study were hemodynamically stable and had undergone a clinical assessment of the probability of PE, using the original Wells scoring system (34). Our study population included 172 patients (66.2%) with a Wells score >4 and 88 patients (33.8%) with abnormal D-dimer test.
Adverse clinical outcomes
Adverse clinical outcomes were defined as death within 30 days or escalation of therapy, according to the MAPPET-3 (Management Strategies and Prognosis in Pulmonary Embolism Trial-3) study criteria (4), including cardiopulmonary resuscitation, endotracheal intubation, vasopressors for systemic hypotension, thrombolysis, or surgical embolectomy. Demographic information (age and sex) and medical history (history of cancer, coronary artery disease, congestive heart failure, diabetes mellitus, chronic lung disease, and renal insufficiency) were recorded on the basis of patients' medical records.
Chest CT procedures
All patients had undergone non–ECG-gated chest CT on a 64-slice multidetector-row CT system (Somatom Sensation 64, Siemens, Forchheim, Germany). Image acquisition parameters consisted of 0.6 mm collimation, 120 kV, 160 mAseff, pitch of 1.2, and a reconstructed section thickness of 1 mm. Contrast medium enhancement was achieved with 100 ml of a nonionic iodinated contrast medium (Ultravist 370, Bayer-Schering, Wayne, New Jersey) injected at 4 ml/s using a power injector (Stellant D, Medrad, Indianola, Pennsylvania). Automated bolus triggering was used with a region of interest in the main pulmonary artery and a threshold of 100 Hounsfield units for initiating data acquisition.
Chest CT interpretation
In case of multiple CT examinations for serial follow-up, we included only the first CT study performed at the time of the initial presentation. In a pilot investigation (21), chest CT studies of 50 consecutive patients had been independently reviewed by 2 experienced observers (D.K.K., C.T.) for analysis of interobserver variability. The remaining 210 patients were evaluated by 1 of the observers (D.K.K.). Observers were blinded to the patients' clinical characteristics. All CT studies were reviewed and analyzed on a clinical workstation (MultiModality Workplace, Siemens).
Deviation of the interventricular septum was evaluated as follows: normal (convex toward the RV), flattened, and septal bowing (convex toward the left ventricle) (25). Flattened septum (Fig. 1A) and septal bowing (Fig. 2A) were considered abnormal positions of the septum indicating RV strain.
The severity of reflux of contrast medium into the IVC or hepatic veins was graded according to a previously published scale (12): 1 = no reflux; 2 = trace of reflux into IVC only; 3 = reflux into IVC but not hepatic veins; 4 = reflux into IVC and proximal hepatic veins; 5 = reflux into IVC and hepatic veins down to the mid-portion of the liver; 6 = reflux into IVC with opacification of distal hepatic veins (Fig. 1B). The degree of reflux was grouped into nonsubstantial (grades 1 to 3) and substantial (grades 4 to 6).
The 2 axial sections that showed the maximal distance between the ventricular endocardium and the interventricular septum, perpendicular to the long axis of the heart, for the RV and LV, respectively, were identified. RVDaxial and LVDaxial were subsequently measured, and the RVDaxial/LVDaxial ratio was calculated (Figs. 1 and 2). Next, 4-CH views were reconstructed on the workstation using our previously described approach (21,26), and RVD4-CH/LVD4-CH ratios were calculated.
The 3D volumetric analysis of both ventricles was performed by using the volume analysis application of the workstation (Volume Analysis, Siemens). The endocardial contours were semiautomatically segmented from the valvular plane down to the apex of both ventricles. That involved manually outlining the endocardial contours on the transverse sections comprising the minimal and maximal expanse of the ventricle, which were then automatically propagated to the neighboring sections. The RVV/LVV ratio was subsequently calculated (Figs. 1 and 2). The time needed to perform the volumetric measurements ranged between 3.7 min and 10.9 min (21).
We used MedCalc (version 10.4.8, MedCalc Software, Mariakerke, Belgium) for all statistical analyses. We used the chi-square test or Fisher exact test for comparisons of categorical variables and the Mann-Whitney test for comparisons in the distributions of continuous variables. Using receiver-operator characteristic curves (in accordance with the methods of DeLong et al. ), the area under the curve (AUC) of RVDaxial/LVDaxial, RVD4-CH/LVD4-CH, and RVV/LVV for predicting adverse events and 30-day death were compared, and optimal cutoff values, weighted for higher sensitivity, for the above ratios were assigned, similar to previous investigations (26,28); these cutoffs were used for subsequent analyses. The Cox proportional hazard model was used to calculate the hazard ratio of clinical variables and CT measurements for predicting adverse outcomes. Multivariate analysis was then performed to identify predictors of adverse outcomes, using the proportional hazards model with calculation of 95% confidence intervals (CIs). Each CT sign was evaluated in a separate model accounting for clinical characteristics and comorbidities that were significant (p < 0.05) predictors of adverse outcomes and 30-day death.
Clinical characteristics of study population
The mean age of the patient population was 55 ± 18 years, and there were 139 (53.5%) men. In all patients, the initial clinical diagnosis of PE was confirmed by the presence of at least 1 filling defect in the pulmonary artery tree (8,9). One hundred forty-five patients (71.4%) had relevant comorbidities, including cancer in 59, coronary artery disease in 35, congestive heart failure in 13, diabetes mellitus in 40, chronic interstitial lung disease in 26, and renal insufficiency in 36. There were no statistical differences between the demographics or CT findings of the 36 excluded patients and final study population (p > 0.05 for all measures).
Fifty-seven patients (21.9%) had adverse clinical outcomes, including 20 (7.7%) who died within 30 days. Of the 37 (14.2%) surviving patients with adverse outcomes, 35 required endotracheal intubations, 15 were treated with vasopressors, 11 received thrombolysis, 7 required cardiopulmonary resuscitation, and 4 underwent surgical embolectomy. Adverse outcomes occurred in the range from 1 day to 30 days (mean 10.4 ± 9.6 days) from the time of diagnosis. Patients with adverse outcomes were older and more commonly had diabetes mellitus, renal insufficiency, and sepsis (Table 1), whereas cancer, diabetes mellitus, renal insufficiency, and sepsis were more common among patients who died within 30 days (Table 2); these parameters were subsequently included in the relevant multivariate analyses.
The pilot investigation (21) in 50 patients had shown fair interobserver reproducibility for describing the position of the ventricular septum (k = 0.32), with moderate reproducibility (k = 0.44) for differentiating normal versus abnormal (i.e., flattening or bowing) position. There was good agreement (k = 0.68) for classifying the degree of contrast medium reflux into the IVC and hepatic veins as nonsubstantial versus substantial. Correlation coefficients for measurements of RVDaxial/LVDaxial ratio, RVD4-CH/LVD4-CH ratio, and RVV/LVV ratio measurements between the 2 observers were 0.88, 0.85, and 0.93 respectively (p < 0.001 for all correlations).
Prognostic value of CT signs
There were no differences in the AUC of RVDaxial/LVDaxial, RVD4-CH/LVD4-CH, and RVV/LVV with AUCs of 0.658 (95% CI: 0.597 to 0.715), 0.659 (95% CI: 0.598 to 0.717), and 0.677 (95% CI: 0.617 to 0.734), respectively, for predicting adverse events (p = 0.565 to 0.940); and AUCs of 0.698 (95% CI: 0.638 to 0.753), 0.694 (95% CI: 0.634 to 0.750), and 0.664 (95% CI: 0.603 to 0.721), respectively, for predicting 30-day death (p = 0.603 to 0.937). Receiver-operator characteristic analysis identified RVDaxial/LVDaxial >1.0, RVD4-CH/LVD4-CH >1.0, and RVV/LVV >1.2 as optimal cutoffs with high sensitivity (>70%) for predicting 30-day death; these values were used for subsequent analyses. Table 3 illustrates the prevalence of the measured CT signs in patients with and without adverse events and 30-day death. The sensitivities and specificities of CT signs for predicting adverse outcomes and 30-day death are presented in Table 4.
Multivariate analysis demonstrated that RVDaxial/LVDaxial >1.0 was not independently predictive of adverse outcomes or 30-day death. Abnormal position of the interventricular septum, substantial IVC contrast reflux, RVD4-CH/LVD4-CH >1.0, and RVV/LVV >1.2 were independent predictors of adverse events, while RVD4-CH/LVD4-CH >1.0 and RVV/LVV were predictive of 30-day death, with hazard ratios of 3.68 (95% CI: 1.08 to 12.60; p = 0.039) and 6.49 (95% CI: 1.77 to 23.84; p = 0.005), respectively (Table 5). Among 149 patients with RVV/LVV ratio >1.2, 45 had adverse events within 30 days (representing a positive predictive value of 30.2%), and 17 died (positive predictive value of 11.4%). Among 111 patients with RVV/LVV ≤1.2, 99 experienced no adverse events (negative predictive value 89.2%) and 108 survived (negative predictive value 97.3%).
We show that 3D measurement of ventricular volumes is superior to other chest CT signs of RV dysfunction for predicting adverse outcomes and 30-day death in patients with acute PE. In our cohort, among 149 patients with RVV/LVV ratio >1.2, 45 (30.2%) had adverse events within 30 days, and the risk of death within 30 days increased approximately 6-fold. Conversely, in the absence of an RVV/LVV ratio >1.2, most patients (97.3%) survived. The routine availability of contemporary advanced image post-processing workstations, which automatically propagate the segmentation of cardiac chambers, has greatly facilitated the performance of 3D ventricular volume measurement. Performing 3D volumetric ventricular measurements can be accomplished on most current image post-processing platforms. This evaluation method may more appropriately account for the complex shape of the cardiac ventricles than a single image or plane.
In our cohort, the mortality rate was 7.7% within 30 days, and thus was very similar to the mortality rates in other studies (3,5,36,37). The proportion of patients who had nonfatal in-hospital adverse events or required escalation of therapy was similar to that in the MAPPET-3 study (4). The patients enrolled in our study suffered from a variety of severe diseases and may therefore have died of other reasons than PE. However, although, overall, most deaths in PE patients occur because of pre-existing underlying disease (37), the main cause of early death, namely, within 30 days as in our study, is acute RV failure (1,3).
The interventricular septum, which normally bows toward the RV, may shift toward the LV related to increased right-side heart pressure with severe pulmonary arterial obstruction (16). Several studies reported that ventricular septal bowing indicated severe PE (13,25) as manifested by RV dysfunction (14,20,22) and predicted short-term death (10). Conversely, Araoz et al. (11), Van der Meer et al. (31), Ghaye et al. (17), and Aviram et al. (12) did not find this sign to be predictive of death from acute PE. These latter findings correspond to our current results, where abnormal position of the interventricular septum was overall predictive of adverse events but not of 30-day death.
Reflux of contrast medium into the IVC is an indirect sign of increased RV pressure and can be seen in various underlying conditions (21). While substantial IVC reflux was predictive of adverse outcomes in our study, it failed to identify patients at risk of early death. Increased RVD/LVD ratio on chest CT has been proposed as a sign of RV dysfunction. However, multiple quantitative methods and cutoff points have been described to assess dilatation of the complex-shaped RV in patients with acute PE. Araoz et al. (10) and Contractor et al. (14) measured both ventricles at the level of the atrioventricular valves, whereas van der Meer et al. (31) used maximum minor axis measurements. Contractor et al. (14) and Lim et al. (22) found a RVDaxial/LVDaxial ratio of >1 as measured on transverse sections indicative of severe PE, while others proposed a threshold of >1.5 (13,27). Several studies report that an increased RVDaxial/LVDaxial ratio on transverse chest CT sections indicates RV dysfunction (14,20,22) and predicts short-term death (10), although others found this sign not to be associated with increased mortality in stable patients without shock (29).
Although a recent report (30) described no statistically significant difference between measurements of RV enlargement on axial and 4-CH views, Dogan et al. (15) reported significant differences for RVD4-CH/LVD4-CH ratios between patients with and without PE but not for RVDaxial/LVDaxial ratios on transverse sections. These findings correspond to our own prior investigations (26,28), where we found that the proportion of patients with increased RVDaxial/LVDaxial ratios on transverse sections was similar for patients with and without adverse events. However, ventricular measurements on 4-CH views were superior, with an AUC of 0.753 of RVD4-CH/LVD4-CH >0.9 versus an AUC of 0.667 of RVDaxial/LVDaxial >0.9, for predicting adverse outcomes. In our current investigation, although RVDaxial/LVDaxial >1.0 was more common in patients who died within 30 days, this parameter was neither predictive of adverse outcomes overall nor of early death. Compared with our previous studies (26,28), receiver-operator characteristic analysis in our current investigation revealed a slightly higher RVD4-CH/LVD4-CH ratio cutoff (1.0 vs. 0.9), which, however, is identical to the RVD/LVD cutoff of 1.0 used in the MAPPET-2 trial (3) and the PIOPED (Prospective Investigation of Pulmonary Embolism Diagnosis) II study (38).
The most important limitation of our study is its retrospective nature. Also, we did not include patients without PE as a control group to determine whether these findings are unique to PE positive patients. However, our current investigation is 1 of the largest analyses to date evaluating the prognostic value of chest CT signs of RV dysfunction in patients with acute PE. It is the only study systematically comparing all signs that have been described for this purpose, including the rather recent method of 3D ventricular volumetry. Our study would have been strengthened by correlation with echocardiography findings; unfortunately, only a small percentage of our patients (100 of 260, 38.5%) had echocardiography findings available. Considering that prior studies have established a good correlation between echocardiography and CT findings (18,20,22,24,26), we chose not to formally analyze this small subset of our patients.
Another potential limitation of our present study is that chest CT image acquisition was not ECG gated. Non–ECG-gated CT is inevitably inaccurate for measuring ventricular chamber size, because the images are acquired in different phases of the cardiac cycle. However, the use of ECG-gated CT protocols over routine chest CT has been shown to result in only limited incremental diagnostic improvements (23). More importantly, because of the additional radiation exposure involved with ECG-gated techniques, this approach is not currently used for routine PE imaging (8), whereas our results obtained in non–ECG-gated chest CT studies are directly transferable to clinical practice. Finally, the clinical applicability of our results is limited by interobserver variability and 3D workstation requirements; however, we believe continuing evolution and implementation of CT technology should mitigate this limitation in the future.
In conclusion, we show that abnormal position of the interventricular septum, IVC contrast reflux, and RVD4-CH/LVD4-CH ratio are predictive of adverse outcomes. However, 3D measurement of ventricular volumes is superior to all other chest CT signs of RV dysfunction for predicting adverse outcomes and 30-day death in patients with acute PE. Future studies are needed to investigate the benefit of cardiac volumetric measurements for prospectively guiding patient management in comparison to and in combination with other risk assessment tools, such as echocardiography or cardiac biomarkers (39–43).
Dr. Schoepf receives research support from and is a consultant for Bayer-Schering, Bracco, General Electric, Medrad, and Siemens. Dr. Bastarrika is a consultant for General Electric, Medrad, and Siemens. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Kang and Thilo contributed equally to this work.
- Abbreviations and Acronyms
- area under the curve
- confidence interval
- computed tomography
- inferior vena cava
- left ventricle
- left ventricular diameter
- left ventricular volume
- pulmonary embolism
- right ventricle
- right ventricular diameter
- right ventricular volume
- Received March 29, 2011.
- Accepted April 11, 2011.
- American College of Cardiology Foundation
- Kasper W.,
- Konstantinides S.,
- Geibel A.,
- et al.
- Kasper W.,
- Konstantinides S.,
- Geibel A.,
- Tiede N.,
- Krause T.,
- Just H.
- Grifoni S.,
- Olivotto I.,
- Cecchini P.,
- et al.
- Schoepf U.J.,
- Goldhaber S.Z.,
- Costello P.
- Ghuysen A.,
- Ghaye B.,
- Willems V.,
- et al.
- Oliver T.B.,
- Reid J.H.,
- Murchison J.T.
- Quiroz R.,
- Kucher N.,
- Schoepf U.J.,
- et al.
- Schoepf U.J.,
- Kucher N.,
- Kipfmueller F.,
- Quiroz R.,
- Costello P.,
- Goldhaber S.Z.
- Piazza G.,
- Goldhaber S.Z.
- Giannitsis E.,
- Muller-Bardorff M.,
- Kurowski V.,
- et al.
- Konstantinides S.,
- Geibel A.,
- Olschewski M.,
- et al.
- Kucher N.,
- Goldhaber S.Z.
- Meyer T.,
- Binder L.,
- Hruska N.,
- Luthe H.,
- Buchwald A.B.