European Radiology

, Volume 17, Issue 8, pp 1954–1959

Computed tomography pulmonary embolism index for the assessment of survival in patients with pulmonary embolism


    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Gero Wieners
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Przemyslaw Dul
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Frank Fischbach
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Oliver Dudeck
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Enrique Lopez Hänninen
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg
  • Jens Ricke
    • Klinik für Radiologie und NuklearmedizinUniversitätsklinikum Magdeburg

DOI: 10.1007/s00330-007-0577-2

Cite this article as:
Pech, M., Wieners, G., Dul, P. et al. Eur Radiol (2007) 17: 1954. doi:10.1007/s00330-007-0577-2


This study was an analysis of the correlation between pulmonary embolism (PE) and patient survival. Among 694 consecutive patients referred to our institution with clinical suspicion of acute PE who underwent CT pulmonary angiography, 188 patients comprised the study group: 87 women (46.3%, median age: 60.7; age range: 19–88 years) and 101 men (53.7%, median age: 66.9; age range: 21–97 years). PE was assessed by two radiologist who were blinded to the results from the follow-up. A PE index was derived for each set of images on the basis of the embolus size and location. Results were analyzed using logistic regression, and correlation with risk factors and patient outcome (survival or death) was calculated. We observed no significant correlation between the CTPE index and patient outcome (p = 0.703). The test of logistic regression with the sum of heart and liver disease or presence of cancer was significantly (p< 0.05) correlated with PE and overall patient outcome. Interobserver agreement showed a significant correlation rate for the assessment of the PE index (0.993; p< 0.001). In our study the CT PE index did not translate into patient outcome. Prospective larger scale studies are needed to confirm the predictive value of the index and refine the index criteria.


Pulmonary embolismPulmonary arteriesCTIndex



pulmonary embolism


computed tomography


multi-detector computed tomography


deep venous thrombosis


Non-invasive diagnostic imaging using computed tomography (CT) or magnetic resonance imaging has been suggested to be a useful tool for the diagnostic work-up in patients with suspected pulmonary embolism (PE) [1, 2]. Thus, multi-detector computed tomographic (MDCT) pulmonary angiography has gained wide acceptance as the first-line investigation for the detection of acute PE [3, 4]. The sensitivity and specificity reported in previous studies have ranged from 53 do 89%, and 78 to 100%, respectively [410]. With the development of MDCT, narrow collimation and modern workstations for image post-processing have opened a new chapter in the diagnostics of this frequent diagnosis. Therefore, today CT pulmonary angiography is being recommended as the modality of choice for the assessment of patients with suspicion of PE [4, 11, 12].

The evaluation of the degree of pulmonary arterial obstruction from PE expressed as the CT PE index had been proposed by Qanadli et al. [13], but the correlation between the massiveness of the pulmonary embolism assessed by imaging modalities and clinical outcome still remains underinvestigated. In a more recent study by Wu et al., standardized indices for the quantification of PE have been suggested as a valuable predictor of patient outcome, and according to their study, the CT PE index has been shown to be an important predictive factor of patient death, since patients with pulmonary vascular obstruction of more than 60% tend to have a poor clinical outcome [4].

The interesting issue is however that it is possible to visualize segmental and subsegmental arteries [14, 15] with high reproducibility for the diagnosis [16] and that the mortality and morbidity of patients with PE decrease at much lower rates, and rather disproportionately to the development of diagnostics. But, on the other hand, it may mean that the degree of obstruction evaluated by radiologists is not related to the severity of the patient’s outcome as was addressed in a study by Alpert et al. 30 years ago [17]. In this pioneering paper, the authors demonstrated that the mortality of a patient with PE is mainly driven by the clinical consequence of the disease and that pulmonary vascular obstruction has no impact on mortality.

In our study, we decided to assess if Wu et al.’s hypothesis was applicable in our institution for patients with clinical suspicion of acute PE without cardiogenic shock. Thus, the main objective of our analysis was to define the correlation between the CT PE index and patient survival.

Materials and methods

This study was approved by the institutional review board; informed consent was not required. Between September 2001 and September 2004, 694 consecutive patients referred to our institution with clinical suspicion of acute PE underwent CT pulmonary angiography procedures. All 211 examinations with diagnosed PE at the time of CT interpretation were included. Among the 211 examinations, 9 were excluded because of poor image quality, another 8 cases were excluded because the examinations were performed on a single-row detector CT scanner (no secure statement for subsegmental arteries), and 6 examinations were follow-up procedures in the same patients, and thus were also excluded. A total of 23 examinations were excluded from further analysis, and finally 188 patients were eligible to enter the study group.

All analyzed images were obtained on multi-detector row scanners; 58 examinations (30.9%) were performed on 4-channel detector CT, 76 examinations (40.4%) on 8-channel detector CT, and 54 examinations (28.7%) on 16-channel detector CT scanners. Breath-hold or shallow-breath acquisition was employed depending on the patient's respiratory capacity. The scanning area encompassed the chest and upper abdomen to the level of adrenal glands. Images were acquired in the cranio-caudal direction.

CT scans were obtained by using the following scanners: CT Light Speed 16 (GE, Milwaukee, WI) and CT Light Speed 8 (GE, Milwaukee, WI); collimation 16 × 1.25 mm or 8 × 1.25 mm, table speed 15 mm/s, 120 KV and 150 mAs. Images were reconstructed with an effective slice thickness of 1.25 mm or 2.5 mm. Delay in contrast administration was defined by SmartPrep software or was fixed for 15 s. For VolumeZoom 4 (Siemens, Erlangen, Germany) CT scanner collimation was 4 × 2.5 mm, slice thickness 3 mm, table speed 7.5 mm/s, 120 KV and 150 mAs, and reconstruction slice thickness 3 mm. Scanning delay was determined by C.A.R.E software or fixed 15-s delay was applied.

All studies were performed after the injection of 100 ml of non-ionic contrast medium iopromide (Ultravist 370, Schering AG, Germany), which was administered intravenously to an antecubital vein at a rate of 3.5 ml/s (GE scanners) and 3 ml/s (Siemens scanner) with an automated double-piston injector device. After termination of contrast agent administration, 50–60 ml of saline was injected.

Images were analyzed in standard light on 19-inch screens calibrated with Adobe Photoshop (Adobe Systems, Mountain View, CA) and were reviewed with Acculite software (Version 3.122, AccuImage Diagnostic Corporation, CA). Evaluation was performed in the mediastinal window (350–400 W/50 L) and pulmonary window (1,200 W/−700 L). Occasionally individual window settings with variable levels were chosen at the discretion of the investigator.

The CT criterion used to diagnose pulmonary emboli consisted of visualization of the thrombus present within the contrasted pulmonary artery, causing a filling defect on at least two subsequent scans. Locations of thrombi were described in accordance with the anatomy of the pulmonary vasculature. Subsequently, an extension of the embolism was estimated with the use of the PE index. The PE index was defined as:
$$ PE\;index = {\sum {{\left( {N_{{seg}} \times D} \right)}} }. $$
where Nseg is the value of the proximal clot site, equal to the number of segmental branches arising distally, and D is the degree of the obstruction. The maximal N value was 10 per each lung and was equal to the number of segments. The value D could be expressed in two possible scores: either as the value 2 when the vessel was occluded completely or as the value 1 when the segmental artery was occluded partially. Emboli within subsegmental arteries were treated as partial occlusion of segmental vessels and were scored as value 1. The thrombus present within the proximal vessel (lobar arteries or pulmonary trunk) influenced the index values of the distal segmental vessels depending on the degree of occlusion (D). According to these rules, when the lumen of all segmental vessels was closed completely, the maximal number of points possible to assign was 40, which means 20 points for each lung. Afterwards, all index scores were expressed as proportionate values.

CT scans were reviewed independently by two experienced radiologists blinded to each other's findings and to the patient's survival status.

To assess the outcome of the patients with PE medical data concerning each patient were retrospectively collected from medical records:
  • Place of onset of clinical symptoms (home, hospital)

  • Survival or death (date of death)

  • Occurrence of deep venous thrombosis (DVT) diagnosed on the basis of Doppler ultrasound

  • Possible surgery preceding PE

  • D-dimers level

  • Coagulation factors deficit or C and S proteins deficit, evaluated mostly in young patients

  • Malignant disease (the type of malignancy was defined on the base of examination)

  • Coexisting liver disease such as cirrhosis, hepatitis B or C

  • Coexisting heart disease, based on history and ECG (ischemic disease, history of myocardial infarction, serious valvular insufficiency, cardiomyopathy, etc.)

  • Therapy (therapy decision clinical-condition-dependent: thrombolysis, heparin bolus, low-molecular-weight heparins, warfarine derivatives)

Malignant disease was defined on the basis of histopathological or cytological examinations.

Data concerning survival or death were obtained from the registry. Death related by PE was defined as death within 30 days following the occurrence of pulmonary embolism.


All analyses were performed by using a software program (SPSS for Windows, version 10.0.0, Chicago, IL). For calculation of interobserver agreement of the PE index calculation the Kendall-W test was used. Statistical differences and correlations of the PE index and survival were calculated using Spearman’s coefficient of rank correlation.

The relationship between clinical findings and PE was in the test of logistic regression.


Our results are summarized in Figs. 1 and 2 and Tables 1 and 2. Distribution of the clot burden as derived from the PE index for all patients is shown in Fig. 1, for the 30 day-period after PE in Fig. 2, and the mean and median value of the PE index is shown in Table 1. Patient characteristics are summarized in Table 2.
Fig. 1

Histogram shows the distribution of clot burden as measured with the CT PE index for all patients
Fig. 2

Histogram shows the distribution of clot burden as measured with the CT PE index for the 30-day-mortality period

Table 1

PE index


All patients

Death (all)

Death within 30 days post PE

1st reviewer













2nd reviewer













Table 2

Characteristics of 188 patients with PE

No. of patients (%)

All patients, n = 188

Death (all), n = 62 (33%)

Death within 30 days post PE, n = 22 (11.7%)

Patient sex


87 (46.3%)

28 (45.2%)

11 (50%)


101 (53.7%)

34 (54.8%)

11 (50%)

Clinical symptoms


86 (45.7%)

31 (50%)

13 (59.1%)


102 (54.3%)

31 (50%)

9 (40.9%)

Preexisting comorbidity (also multiple findings)


83 (44.1%)

25 (40.3%)

8 (36.4%)


48 (25.5%)

13 (21.0%)

6 (27.3%)

 Hypercoagluable state

26 (13.9%)

9 (14.6%)

4 (18.2%)

 Presence of cancer

57 (30.3%)

33 (53.2%)

13 (59.1%)

 Liver disease

30 (16.0%)

23 (37.1%)

9 (40.9%)

 Heart disease

50 (26.6%)

24 (38.7%)

11 (50.0%)

 History of PE

16 (8.5%)

11 (17.7%)

4 (18.2%)

Type of therapies (also multiple findings)


7 (3.7%)

4 (6.5%)

2 (9.1%)

 Heparin i.v. (bolus + perfusor)

89 (47.3%)

22 (35.5%)

11 (50.0%)

 Heparin s.c.

121 (64.4%)

35 (56.5%)

8 (36.4%)


68 (36.2%)

9 (14.5%)

1 (4.5%)

Kendall-W-test used for calculation of interobserver agreement showed a correlation rate for the assessment of PE index of 0.993 (p< 0.001). The median value was equal for both radiologists with a value of 13.

Median survival of seven patients receiving thrombolysis with recombinant tissue plasminogen activator (rt-PA) was 131.5 days (minimum: 0 days; maximum: 992 days), three patients survived (42.9%) within the frame of the follow-up period. The criteria for the use of thrombolysis were independent of the CT PE index and depended on the clinician's decision about the patient's condition and therapy risks. Eighty-nine patients received heparin as a bolus (5,000 IE) and continuous perfusion i.v., and 92 patients got heparin as as a s.c injection.

The overall median follow-up was 12.7 months (minimum-maximum, 0–37.5 months). Within a 36-month period covering our study, a total of 62 patients died (33%), resulting in a median overall survival of 54 days (minimum-maximum, 0–995 days). Within the 30 day-period after PE, 22 patients died. In this analysis, no appreciable cut-off on any index level consistently separated survival and death. Logistic regression models demonstrated that the PE index was not a significant predictor of patient death. Correlation between the PE index and mortality calculated with the use of Spearman’s coefficient of rank correlation was −0.028 (95% CI = −0.170 to 0.116). We observed no significant correlation between the PE index and patient survival (p = 0.703).

The test of logistic regression with the sum of variables for heart (p = 0.049) and liver disease (p = 0.007) and presence of cancer (p = 0.004) showed a significant (p = 0.039) relationship with patient outcome on day 30.


The establishment of a quantitative method for the assessment of pulmonary artery obstruction [13] has definitely enabled a more accurate evaluation and thus a more precise diagnosis of PE. Therefore, the natural consequence of this step was the attempt to correlate the CT PE index with the patient’s clinical outcome, as evaluated by Wu et al.; according to their results, there was a close negative correlation between the index and patient survival [4]. Unfortunately, the results of our study given here did not confirm the hypothesis of Wu et al. In contrast, the analysis of 188 cases of PE performed in our institution instead showed no correlation between the CT PE index and patient survival. This finding may be attributed to the fact that different medical centers may have their own and slightly different algorithms for the work-up and treatment of particular diseases; e.g., in our hospital, each outpatient with suspicion of acute PE and no contraindication receives a heparin bolus at the latest at the moment of tomographically confirmed diagnosis. Perhaps such a management, although risky, may help to prevent further development of the ongoing disease and thus may significantly impact on the results obtained in our study. On the other hand, however, the median value of CT PE index described by Wu et al. was lower than the median value obtained from our study group, which, simply speaking, means that the emboli we detected were more massive than the emboli found in the aforementioned study with 38 massive PEs in our patient group vs. six in the group of Wu et al. [4].

Another issue that was different when compared with the results from Wu et al. was the fact that the major proportion of patients presented in our study had multiple risk factors, including pre-existing malignant disease, heart disease or liver disease as comorbidity factors.

Several limitations of this study should be addressed. The definition of death related to the presence of PE in patients without cardiogenic shock in our retrospective study (30-day mortality) with a small group of patients who died (n = 22) can change the statistical analysis. Van der Meer et al. evaluated patient deaths (n = 18) by an independent adjudication committee that had full access to all available clinical and diagnostic data for patient deaths that were definitely or most probably attributed to PE or were attributed to a cause unrelated to PE [18]. The results of this study showed that the presence and severity of right ventricular dysfunction and the extent of obstruction of the pulmonary tree help predict mortality within 3 months of clinical presentation in initially hemodynamically stable patients with PE with a similar cut off for vascular obstruction of 60% ± 28% for patients who most probably died from PE as reported by Wu et al. [4, 18]. Our 30-day mortality definition checked the advantage of the CT PE index for all in- and outpatients examined in our department independent of later reasons of death.

The second limitation is the number of available positive PE studies and especially the number of patients with massive PE was limited to 188 and 38, respectively. More patients would provide more power for the study. The test of logistic regression with the sum of variables for heart and liver disease and presence of cancer showed a significant relationship with patient outcome, but only in our patient group, and means at last that mortality depends on associated diseases whatever the vascular obstruction may be and may explain the inhomogeneous treatment in our institution.

In light of our findings, we may state that the clinical outcome of patients must be considered a result of a variety of factors that need to be considered as a whole, and among them, the massiveness of pulmonary artery obstruction was one factor that may not be applicable as the gold standard for the prognosis of patients with PE in clinical practice for radiologists and clinicians.

Our results suggest that the CT PE index may not be simply translated into a patient’s clinical outcome; nevertheless, we consider it a very useful and precise diagnostic tool that may help to attain a more objective assessment of image findings and provide a reproducible standard for measuring the response to thrombolytic therapy. Maybe a combination of right ventricular dysfunction measurements on CT scans and perfusion imaging for visualization of PE [19], including the PE index, will be more objective for all patients with PE in the question of outcome. Further large-scale studies on are needed to confirm the predictive value of the index, and prospective studies will be necessary to refine the definitions and the index criteria.

Copyright information

© Springer-Verlag 2007