Background
Hemodynamic status and comorbidities are key factors in the prognosis of pulmonary embolism (PE) [
1,
2]. In addition to hemodynamic variables, cardiac biomarkers such as troponins and natriuretic peptides are risk factors for patients with acute PE [
3]. The PE prognostic prediction model that is based on these variables is widely accepted [
2,
4].
Initial risk stratification of patients with PE is based on the presence of shock or hypotension [
5]. If the patient is hemodynamically stable, right ventricular function is then assessed by echocardiography, and cardiac biomarkers are measured [
5]. However, these variables are not accurate predictors of PE mortality, especially in hemodynamically stable patients [
6,
7].
The triad of vessel wall injury, venous stasis, and blood hypercoagulability has historically been considered a major risk factor for venous thrombosis [
8]. Infection is another established risk factor for PE [
9], and in certain cases, PE has been associated with influenza [
10] or cytomegalovirus infection [
11].
In deep vein thrombosis (DVT), an inflammatory reaction triggers endothelial cell dysfunction [
12,
13] and results in high serum concentrations of the inflammatory marker C-reactive protein [
14]. Such inflammatory reactions frequently induce well-studied DVT risk factors; however, few studies have investigated a prognostic prediction model for PE. Respiratory and pulse rates are included in the representative PE severity index (PESI) [
15]. Nevertheless, there is a dearth of studies on the potential association between the systemic inflammatory response and PE patient prognosis.
Studies have suggested that leukocytes contribute to venous thrombosis by damaging the endothelium [
16,
17]. Animal models have shown that genetic knock-out of the adhesion molecules E- and P-selectin results in a reduction in thrombus size, which is associated with altered leukocyte accumulation in the surrounding vein wall [
18]. However, the role of leukocytes in the prognosis of PE has not been well studied. We hypothesize that both a systemic inflammatory response and leukocytosis may be negative prognostic factors in PE patients.
Methods
Study design
This retrospective single center observational cohort study was conducted between January 2005 and December 2011. This study was approved by the institutional review board at Dongsan Hospital, Keimyung University School of Medicine.
Study subjects
A total of 667 PE patients were enrolled at Dongsan Hospital from January 2005 through December 2011. All PE patients treated during the study period were included; no additional selection criteria were used. Subjects were either admitted to the hospital or were Emergency Department or outpatient clinic patients.
Image studies
PE was defined either as a filling defect in the pulmonary artery detected through chest computed tomography (CT) or CT pulmonary angiography, or diagnosed based on a ventilation perfusion scan. DVT diagnosis was confirmed via ultrasound examination of the lower extremity veins in patients with clinically suspected PE.
Study methods
Electronic medical records of all PE patients were examined. Risk factors included renal dysfunction, defined as serum creatinine level >1.3 mg/dL; active cancer, defined as treatment with an anti-cancer agent within 3 months of PE diagnosis; hospital admission for supportive therapy within 3 months of PE diagnosis; or outpatient use of analgesics for end-stage malignancy. Shock was defined as systolic blood pressure of <90 mmHg.
CT was performed using a 16 or 64-slice detector. The reconstruction interval of the scan was 3 mm. The ratio of right ventricular diameter to left ventricular diameter (RV/LV ratio) was calculated using CT scan images showing the interventricular septum and myocardium from the longitudinal axis of the heart [
19,
20]. A radiologist reviewed all the CT scans in a blind fashion.
Patients were diagnosed with systemic inflammatory response syndrome (SIRS) if they met 2 or more of the following criteria: peripheral white blood cell (WBC) count of <4,000/μL or >12,000/μL, respiratory rate >20 breaths per min, pulse rate >90 beats per min, and body temperature of >38.3°C or <36.0°C [
21]. We retrospectively collected the clinical characteristics and laboratory data on factors such as comorbidity, symptoms, chest CT findings, vital signs, RV/LV ratio, NT Pro-BNP, and complete blood count at the time of diagnosis of PE.
We defined 30-day all-cause mortality as a primary outcome.
Statistical analysis
All values are expressed as mean ± standard deviation. Data were analyzed using SAS version 9 (SAS Institute, Cary, NC, USA) and MedCalc version 11.0 (MedCalc Software, USA). A χ
2 test was used to compare frequencies. Student’s
t test was used for statistical significance analysis of continuous variables, and cross-analysis was used for categorical variables. Statistical significance was set at <0.05. Univariate analysis of statistically significant variables was conducted using logistic regression. Multivariate analysis was performed for variables with a
p-value of <0.1. We used the area under the receiver operating characteristic (ROC) curve or the area under the curve (AUC) to quantify the ability of the model to distinguish between high- and low-risk subjects. To compare the ROC curves, we used the Delong et al. [
22] method to calculate the standard error of the AUC and the difference between 2 AUCs with the exact binomial confidence interval for the AUCs. The improved discriminative and predictive values of the WBC count + PESI score and the prognostic model were examined by calculating the net reclassification improvement (NRI), as described by Pencina et al. [
23].
Discussion
The independent variables identified in this study as predictors of mortality within 30 days of hospital admission for PE were: SIRS satisfying the peripheral blood WBC count criterion, altered mental state, shock, and the RV/LV ratio. In patients with PE, WBC count and SIRS satisfying the peripheral WBC count criteria were significantly associated with mortality within 30 days of hospital admission (OR = 1.05, 95% CI, 1.01–1.09; OR = 2.8, 95% CI, 1.5–5.4, respectively) (Tables
2,
3).
Patients with PE initially present with various clinical profiles, ranging from clinically stable to shock status. Moreover, 30-day mortality rates in PE patients with hemodynamic abnormalities have been reported to range from 5% to 58% [
1,
24]. The prevalence of shock in our sample was 5.3% (36 cases). The 30-day mortality rate for PE patients presenting with shock was 25%. The RV/LV ratio measured by chest CT indicated potential right ventricular failure, which also reflects the severity of PE [
20]. The mean RV/LV ratio observed in the present study was similar to that reported in a previous study [
19]. Moreover, the RV/LV ratio was significantly associated with 30-day mortality (OR = 1.7, 95% CI, 1.0–2.8), which is consistent with previous reports that death within 30 days of hospital admission is predictable [
25]. However, the levels of N-terminal prohormone of brain natriuretic peptide were not independently associated with mortality our study. This may be, in part, due to the limitations imposed on data collection, which are inherent to retrospective studies.
The percentage of PE patients with 3 or more risk factors, including infection, has been reported to be as high as 50% [
26]. This is expected, as WBCs are involved in the coagulation process [
27]. In the present study, we demonstrated that in addition its role as a risk factor for PE, SIRS also plays an important role in patient prognosis. Analysis of SIRS criteria showed that heart rate and the WBC criterion were significantly associated with mortality (Table
2). Multivariate analysis including all significant variables, identified SIRS satisfying the WBC count criterion as a significant prognostic factor (Table
3). These results indicate that WBC count in the systemic inflammatory response is an important prognostic factor in PE patients. Although there was a significantly higher percentage of infection in non-survivors compared to survivors (Table
1), multivariate analysis showed SIRS satisfying leukocytes criteria was an independent predictor of outcome, while infection was not. These results suggest that SIRS satisfying the leukocytes criteria is a predictor of outcome, independent of coincident events such as an infection.
In contrast to studies that reported a significant association between the presence of a malignant tumor and PE mortality [
1,
28], no such relationship was observed in this study. We speculate that this lack of association may be due to the recent progress in cancer management. Indeed, this may also result in a difference in the ROC curve from the original PESI [
4]. Cancer survival rates are currently improving [
29‐
31], which may influence the reclassification of the prognostic index in the future.
The area under the ROC curve of shock + altered mental state + active cancer was 0.64, which is lower than the previously reported value [
28]. However, when RV/LV ratio, renal dysfunction, and SIRS satisfying the WBC criterion were included in the risk prediction model, a marginal increase in the AUC from 0.64 to 0.76 (
P = 0.05) was observed. We classified the predicted risks obtained from both models (old and new) into 3 categories (0%–10%, 10%–30%, >30% of 30-day PE mortality) and then cross-tabulated these 2 classifications. Consequently, classification was less accurate for approximately 36% of non-survivors according to the new model, compared to the old model. In contrast, classification was more accurate for approximately 11% of survivors (Table
4).
The prognostic significance of WBC count was evaluated using the previously established PE prognostic index. Huang CM et al. also showed WBC count (≥11,000 mm
3) was an independent predictor of 30-day mortality in PE patients [
32]. Accordingly, we found that WBC count was an independent prognostic factor, apart from PESI, for 30-day PE mortality (Table
5). The area under the ROC curve of PESI was 0.68, which is lower than the previously reported value [
4]. However, when WBC count was added to the PESI, an increase in AUC from 0.72 to 0.76 (
P = 0.008) was observed. We classified the predicted risks obtained from both models (PESI and PESI + WBC) into 3 categories (0%–10%, 10%–30%, >30% of 30-day PE mortality), and then cross-tabulated these classifications. Consequently, the classification of approximately 5% of non-survivors was less accurate when comparing PESI + WBC with PESI alone. Conversely, approximately 3% of survivors were reclassified down when WBC count was added to PESI (Table
6).
Thrombolytics, inferior vena cava filter, and surgical embolectomy were performed in 16, 7, and 2 of the cases included in this study, respectively. Thus, our analysis of prognostic factors did not account for treatment method.
Competing interests
All authors declare that they have no competing interests.
Authors’ contributions
JYJ was responsible for data analysis, and for drafted this manuscript. WIC was responsible for the content of the manuscript, for study design, for data analysis, and for drafted this manuscript; JWL was responsible for the data collection; BHR was responsible for the data collection and analysis; MYL was responsible for the data analysis and interpretation. All authors contributed to the drafting and revisions of the manuscript. All authors read and approved the final manuscript.