Computed Tomography Pulmonary Perfusion for Prediction of Short-Term Clinical Outcome in Acute Pulmonary Embolism

Background  Computed tomography pulmonary angiography (CTPA) is the imaging modality of choice for the diagnosis of acute pulmonary embolism (PE). With computed tomography pulmonary perfusion (CTPP) additional information on lung perfusion can be assessed, but its value in PE risk stratification is unknown. We aimed to evaluate the correlation between CTPP-assessed perfusion defect score (PDS) and clinical presentation and its predictive value for adverse short-term outcome of acute PE. Patients and Methods  This was an exploratory, observational study in 100 hemodynamically stable patients with CTPA-confirmed acute PE in whom CTPP was performed as part of routine clinical practice. We calculated the difference between the mean PDS in patients with versus without chest pain, dyspnea, and hemoptysis and 7-day adverse outcome. Multivariable logistic regression analysis and likelihood-ratio test were used to assess the added predictive value of PDS to CTPA parameters of right ventricle dysfunction and total thrombus load, for intensive care unit admission, reperfusion therapy and PE-related death. Results  We found no correlation between PDS and clinical symptoms. PDS was correlated to reperfusion therapy ( n  = 4 with 16% higher PDS, 95% confidence interval [CI]: 3.5–28%) and PE-related mortality ( n  = 2 with 22% higher PDS, 95% CI: 4.9–38). Moreover, PDS had an added predictive value to CTPA assessment for PE-related mortality (from Chi-square 14 to 19, p  = 0.02). Conclusion  CTPP-assessed PDS was not correlated to clinical presentation of acute PE. However, PDS was correlated to reperfusion therapy and PE-related mortality and had an added predictive value to CTPA-reading for PE-related mortality; this added value needs to be demonstrated in larger studies.


Introduction
Computed tomography pulmonary angiography (CTPA) is the current imaging modality of choice for the diagnosis of pulmonary embolism (PE). 1 In recent years, technical advances have been made in the diagnostic management of PE including the introduction of computed tomography pulmonary perfusion (CTPP) imaging. With CTPP additional information of hemodynamic and functional impact of the PE as expressed by measures of pulmonary perfusion can be assessed. 2 Available studies using CTPP have mostly focused on its diagnostic performance for acute PE. The addition of CTPP to CTPA has been reported to improve the specificity for a PE diagnosis 3 and to improve the detection rate of small, subsegmental emboli. 4,5 Also, perfusion defects on CTPP were found to be correlated to PE thrombus load and signs of right ventricular dysfunction on CTPA. 6-10 Therefore, perfusion defects on CTPP may be relevant for prognostication of PE patients as well, although this is less well studied. For instance, the quantification of perfusion defects may predict PE-related death, hemodynamic collapse, or need for oxygen therapy. This information is relevant for initial risk stratification and treatment or to consider home treatment in patients with good prognosis. 11 In this study, we aimed to evaluate the correlation between perfusion defects on CTPP and clinical symptoms at presentation and its predictive value for adverse short-term outcome of acute PE.

Study Design and Population
This was a prospective observational study in a convenience sample of 100 consecutive hemodynamically stable adult patients (!18 years) with CTPA-confirmed acute symptomatic PE, diagnosed between July 2017 and October 2019 in the Leiden University Medical Center in whom CTPP was performed as part of routine clinical practice. Patients were excluded in case of nonassessable CTPP scan due to imaging artifacts. The diagnostic management of patients with suspected acute PE started with assessment of the clinical pretest probability in combination with D-dimer testing, following the YEARS algorithm. 12,13 In patients with CTPAconfirmed acute PE, anticoagulant treatment was started or modified in patients already on anticoagulant treatment according to international standards. The Hestia rule, consisting of 11 clinical criteria, was used to identify low-risk PE patients for outpatient treatment. 11,14,15 This study was approved by the institutional review board of the Leiden University Medical Center, and informed consent requirement was waived due to its observational nature.

Primary and Secondary Aim
The primary aim was to investigate the correlation between quantification of CTPP-measured perfusion defects with clinical symptoms at presentation and its predictive value for adverse short-term 7-day outcome. The secondary aim was to investigate the added value of CTPP reading to right ventricle to left ventricle diameter ratio (RV/LV ratio), pulmonary artery trunk diameter and total thrombus load on CTPA for prediction of intensive care unit (ICU) admission, reperfusion therapy, and PE-related mortality. Furthermore, the correlation between PDS on CTPP and total thrombus load on CTPA was evaluated.

Outcomes
For the primary outcome, clinical symptoms at presentation and adverse short-term outcome were evaluated. Clinical symptoms included (non)pleural chest pain, dyspnea, and hemoptysis. Adverse short-term outcome included hospital or ICU admission, need for supplemental oxygen therapy or intravenous pain medication >24 hours, reperfusion therapy, vasopressor or inotropic therapy, and PE-related death within 7-day follow-up. All symptoms and outcomes were assessed from digital patient files.
For the secondary outcome, we assessed prognostic imaging signs on CTPA including RV/LV ratio, pulmonary artery trunk diameter, and total thrombus load. The predictive capacity of these CTPA clinical imaging signs and of PDS for the outcome of ICU admission, reperfusion therapy, and PE-related mortality was evaluated.

Image Acquisition and Analysis
Since June 2017, CTPP is part of the standard CT angiography protocol in adult patients with suspected PE at our hospital. CT examinations were performed on a 320-multislice detector row CT scan (Canon). CTPP images were acquired by using subtraction technique, in which the precontrast image is subtracted from the contrast-enhanced image. The subtraction image is then color coded and fused with the CTPA images; normal perfusion: yellow to orange, moderately decreased perfusion: red to pink, and severely decreased or absent perfusion: purple to dark blue/black (►Fig. 1).
For this analysis, CTPP and CTPA image reading was performed independently by two different readers, who were unaware of presenting symptoms and occurrence of adverse events. CTPP assessment was performed by a researcher (L.F.V.D) trained by an expert thoracic radiologist (L.J.M.K.). Perfusion defects were quantified per segment by using the score proposed by Chae et al and expressed as mean PDS in percentage. 10 To assess the interobserver agreement for PDS reading, CTPP images of 25 consecutive patients were independently evaluated by a second reviewer (L.J.M.K.). RV/LV ratio, pulmonary artery trunk diameter, and total thrombus load on CTPA were evaluated by one expert thoracic radiologist (L.J.M. K.) with over 20 years of experience in pulmonary CTPA reading. The maximum diameters of both the right and left ventricle were measured in the standard axial view with measurement of the maximal distance between the ventricular endocardium and the interventricular septum. The pulmonary artery trunk was measured at its largest transverse diameter. The total thrombus load was assessed by using the CT obstruction index according Qanadli et al. 16 Definitions Acute PE was defined as at least one filling defect in the pulmonary artery tree on CTPA. 17 defined as sharp chest pain that worsens during breathing. Nonpleural chest pain was defined as pressure on or squeezing sensation in the chest. PE-related death was defined as objectively confirmed clinically severe PE before death in the absence of an alternative diagnosis. 20

Statistical Analysis
Baseline characteristics are described as mean with standard deviation (SD) or median with interquartile range. To evaluate the correlation between PDS to clinical symptoms and adverse outcomes, the difference between the mean PDS with corresponding 95% confidence interval (CI) in patients with versus without chest pain, dyspnea, and hemoptysis and adverse short-term outcome was calculated. To evaluate the agreement in PDS scoring between the two reviewers, the mean difference between PDS assessed by reviewer 1 and 2 was calculated.
The added predictive value of PDS to CTPA assessment for ICU admission, reperfusion therapy, and PE-related death was assessed by comparing two prediction models. In the first prediction model, CTPA parameters including RV/LV ratio, pulmonary artery trunk diameter, and total thrombus load were included. In the second prediction model, PDS assessment was added to these CTPA parameters. Multivariable logistic regression analysis and the likelihood-ratio test were performed to assess the predictive value of the two models for ICU admission, reperfusion therapy, and PE-related death and whether PDS assessment significantly improved the predictive value of the model. Additionally, to quantify the performance of the prediction models, we determined the discrimination and calibration. Discrimination refers to the ability to discriminate between those with and those without the outcome and calibration to the agreement between observed outcomes and predictions. Discrimination was expressed with the concordance (c) statistic, by calculating the area under the receiver operating characteristic curve (AUC) with a 95% CI, with discrimination considered perfect if AUC of 1, good if AUC >0.8, moderate if AUC 0.6 to 0.8, poor if AUC <0.6, and no better than chance if AUC ¼ 0.5. Calibration was assessed by using the Brier score, which ranges from 0 to 0.25, with a score of zero signifies a perfect prediction model and a score of 0.25 a noninformative model. 21 The correlation between PDS and total thrombus load was evaluated by using the Pearson's correlation test. A two-sided p-value of p <0.05 was considered as statistical significant. All statistical analyses were performed in SPSS version 25 (IBM, Armonk, New York, United States).

Study Population
A total of 100 patients with CTPA proven acute PE were eligible for analysis. The baseline characteristics of the 100 included patients are shown in ►Table 1. Four patients were transferred to another hospital within 48 hours because of  logistical reasons. Results for adverse short-term outcome were thus available for 96 patients. The mean PDS of all included patients was 27% (SD ¼ 13%) and mean Qanadli score was 30% (SD ¼ 23%). Forty-nine patients (49%) had a RV/LV ratio >1 (►Table 1). The agreement in PDS scoring between the two reviewers was good with a mean difference in PDS of 4.2% (SD ¼ 6.9%).

Secondary Outcome
The results from logistic regression and likelihood-ratio test are shown in ►Table 3. With the use of the Pearson correlation test, a positive correlation between the total PDS and CTPA-assessed total thrombus load was found (r ¼ 0.523, p < 0.001).

Discussion
We showed that perfusion defects on CTPP are correlated to reperfusion therapy and PE-related mortality, and that the addition of PDS assessment to CTPA assessment of RV/LV ratio, pulmonary artery trunk diameter, and total thrombus load improved the predictive value of the model to predict PE-related mortality, but not ICU admission nor reperfusion therapy. Moreover, perfusion defects on CTPP did not correlate to clinical symptoms at presentation. Risk stratification of patients with acute PE is crucial for deciding on the optimal treatment, including hospitalization, close hemodynamic monitoring, and reperfusion therapy. 1,22,23 Previous studies found that right ventricle enlargement (RV/LV ratio > 1.0) is associated with an increased risk for PE-related mortality. [24][25][26] Current European guidelines therefore recommend the assessment of right ventricular dimensions or function as part of initial risk stratification. 22 As previous publications have shown that CTPP-assessed PDS is correlated to RV/LV ratio and total thrombus load, 6-10 perfusion imaging may play a role in this risk stratification. Although our results showed an improvement in the predictive capacity for PE-related mortality when PDS was added to CTPA-reading, the improvement in AUC was only marginal. Furthermore, we could not confirm an added value of PDS over CTPA assessment to predict ICU admission nor reperfusion therapy. A possible explanation may be the low incidence of these adverse outcomes (range between two and seven patients).
We also evaluated whether perfusion defects on CTPP were correlated to clinical symptoms at presentation. This is relevant, as pain and dyspnea for which treatment with intravenous pain medication, and oxygen therapy may be needed are also relevant for the decision for hospitalization or home treatment. 27,28 However, an association between PDS and presenting symptoms could not be established. Of note, as the generation of dyspnea and chest pain involve multiple underlying (complex and not fully understood) mechanisms, a discrepancy between chest pain and dyspnea and extent of perfusion defects in acute PE is possible. 29 PDS was also not correlated to hospital admission. However, the decision to admit a patient to the hospital is often based on multiple variables, some not included in this analysis, including pregnancy, active bleeding and the presence of a social reason for treatment in hospital.
In current literature, the addition of CTPP to CTPA was found to improve the specificity in the PE detection from 94% (95% CI: 89-97%) to 100% (95% CI: 100-100%) and the detection of occlusive (sub)segmental pulmonary emboli. 30 CTPP was also evaluated for PE prognostication and was shown to be correlated to adverse clinical outcome including ICU admission, all-cause and PE-related mortality, 7,31 but had no added value to RV/LV ratio to predict mortality. 32,33 Hence, based on our results and Table 3 Predictive value, area under the receiver operating characteristic curve and Brier score of model 1 (CTPA parameters including RV/LV ratio, pulmonary artery trunk diameter, and total thrombus obstruction score) and model 2 (CTPA parameters and perfusion defect score) and odds ratios for each parameter within model 2 for ICU admission, reperfusion therapy, and PE-related mortality available literature, the application of CTPP seems to be mostly relevant for the diagnostic management of acute PE rather than for prognostication. Limitations of the study are its observational design and the use of a convenience cohort without a specific sample size calculation. This latter may have resulted that the study was underpowered to detect a correlation between PDS and clinical symptoms and some adverse outcomes. On the other hand, the predictive value of perfusion defects for reperfusion therapy and PE-related mortality may be overestimated due to the low incidence of these adverse events and should therefore be interpreted with caution. Furthermore, the presence of clinical symptoms was self-reported and not assessed in a standardized manner, what may have introduced relevant bias. Bias may also be present in the perfusion defect quantification as perfusion defects may not only be the result of a PE but also of other pathology such as pneumonia. The strengths of this study are its prospective design and the inclusion of all comers, which supports the external validity of our findings. Also, CTPA and CTPP assessment was performed by independent readers who were unaware of the clinical presentation and course.
In conclusion, PDS was not associated with clinical presentation of acute PE. However, our data showed that CTPP-assessed PDS was correlated to reperfusion therapy and PE-related mortality and improved the predictive value of CTPA reading for PE-related mortality, but not for ICU admission or reperfusion therapy. Due to the limited number of adverse events and the design or our study, our observations should be considered hypothesis generating. Future larger studies including an upfront determined sample size calculation are needed to determine the clinical relevance of PDS quantification on top of CTPA assessment of right ventricle dysfunction in risk stratification of acute PE.