Loss of Pulmonary Vascular Volume as a Predictor of Right Ventricular Dysfunction and Mortality in Acute Pulmonary Embolism

Abstract

Background:

In acute pulmonary embolism(PE), chest computed tomography angiography(CTA) derived metrics, such as the RV:LV ratio, are routinely used for risk stratification. Paucity of intraparenchymal blood vessels has previously been described, but their association with clinical biomarkers and outcomes has not been studied. We sought to determine if small vascular volumes measured on CT scans were associated with an abnormal RV on TTE and mortality. We hypothesized that decreased small venous volume would be associated with greater RV dysfunction and increased mortality.

Methods:

A retrospective cohort of patients with intermediate risk PE admitted to Brigham and Women’s Hospital between 2009–2017 was assembled, and clinical and radiographic data were obtained. We performed 3D-reconstructions of vasculature to assess intraparenchymal vascular volumes. Statistical analyses were performed using multivariable regression and cox-proportional hazards models, adjusting for age, sex, lung volume and small arterial volume.

Results:

722 subjects were identified of whom 573 had documented TTE. A 50% reduction in small venous volume was associated with an increased risk of RV dilation(RR=1.38,95%CI:1.18–1.63 p<0.001), RV dysfunction(RR=1.62, 95%CI:1.36– 1.95,p<0.001), and RV strain(RR=1.67,95%CI:1.37–2.04 p<0.001); increased cardiac biomarkers, and higher 30-day and 90 day mortality(HR=2.50,95%CI:1.33 –4.67, p=0.004 and HR=1.84,95%CI:1.11–3.04, p=0.019 respectively)

Conclusions:

Loss of small venous volume quantified from CTA is associated with increased risk of abnormal RV on TTE, abnormal cardiac biomarkers and higher risk of 30- and 90-day mortality. Small venous volume may be a useful marker for assessing disease severity in acute PE.

Introduction

Chest computed tomography angiography (CTA) has become the standard of care for initial diagnosis of acute pulmonary embolism (PE). CTA not only establishes the diagnosis of PE but also provides prognostic information¹. Metrics such as the ratio of right to left ventricular (RV:LV) diameter are now routinely used by clinicians for real time prognostication. Paucity of visible lung blood vessels, referred to as the Westermark sign on plain radiographs², has long been observed and described in the PE literature. While often mentioned as a sign of the presence of PE, less is known about its prognostic significance. In chronic lung disease, the loss of vascular volume automatically derived from CT imaging has been associated with clinical and hemodynamic measures of disease severity in chronic obstructive pulmonary disease (COPD)³, asthma⁴, pulmonary arterial hypertension (PAH)⁵, and chronic thromboembolic pulmonary hypertension (CTEPH)⁶. Recently, we described the loss of small venous volume as a predictor of RV:LV volume ratio in a cohort of patients undergoing ultrasound facilitated catheter guided fibrinolysis for acute PE⁷.

In this investigation, we sought to determine if measurement of small vascular volume derived from the CT scan obtained at the time of acute PE diagnosis was associated with abnormal appearance of the RV on TTE and mortality. We focused on patients in whom a clinical diagnosis of intermediate risk PE was suspected at the time of initial presentation. We hypothesized that decreased small venous volume as assessed by CTA would be associated with greater RV dysfunction and increased mortality.

Methods

This study was approved by the IRB(#2016P002693) at Brigham and Women’s Hospital (Boston, MA). The data that support the findings of this study are available from the corresponding author upon reasonable request.

Cohort Selection

With a focus on establishing a cohort of patients with a clinical suspicion of intermediate risk PE, we identified patients admitted to Brigham and Women’s Hospital between 2009 – 2017 using the Research Patient Data Registry (RPDR). Patients were defined as having an intermediate risk PE if they met criteria outlined by the 2019 European Society of Cardiology guidelines.⁸ This was defined as the presence of acute PE with normal hemodynamics, increased markers of risk (PESI score) and possibly increased RV size on imaging (CTA or TTE) and/or positive biomarker. RPDR was queried for the following clinical data: a diagnosis of intermediate risk PE identified via ICD-9 or ICD-10 codes; medical history, comorbidities, transthoracic echocardiogram (TTE) reports, treatments administered for PE abstracted from discharge summaries and mortality data. An abnormal appearing RV was defined by the presence of RV dilation, dysfunction, strain, or a combination of these findings on the TTE report. Serum biomarkers were defined as abnormal if values were above the upper limit of the normal reference range for Troponin I, Troponin T, BNP or NT-pro BNP levels obtained at the time of initial diagnosis. Mortality data were extracted from RPDR using the Social Security Death Index.

Radiological data obtained included the initial diagnostic CTA that patients received during their hospital stay. All scans were performed using a standard bolus tracking CTA acquisition timed for maximum opacification of the pulmonary arteries. Patients were excluded if the CTA slice thickness was >1.25 mm or if scans had poor quality of lung imaging (missing fields or > 50% collapse/infiltrate of either lung).

Lung Segmentation and Vascular Quantification

Automated lung segmentation,^{9, 10} assessment of the total lung volume by summation of all voxels within the lung parenchyma based on the generated segmentation, and subsequent vascular reconstruction within that parenchymal region were performed on imaging at the time of diagnosis using the using the Chest Imaging Platform (www.chestimagingplatform.org)^{11, 12}. Detection and estimation of the size of the vasculature were performed using scale-space particles^13–15, a fully automated methodology which takes advantage of the geometry of the blood vessels for detection and quantification of vessel segments and their respective diameters. Vascular segmentations were visually inspected for quality prior to quantification.

Separation of the intra-parenchymal arterial and venous vascular trees was accomplished using a convolutional neural network.¹⁵ An approximation of the distribution of the vascular size as a function of cross-sectional area was used to estimate the volume of different vascular size ranges using the Chest Imaging Platform for each subject. Small blood vessel volume fractions defined as volume in vessels less than 10mm² in cross section were calculated for the arterial and venous trees. Small vessel volumes were further differentiated into small venous volume and small arterial volume, each less than 10mm² in cross section and measured in ml.^{4, 7} Examples of intraparenchymal vascular reconstruction for three subjects, one with regional loss of vascularity, one with a saddle PE and RV dysfunction and significant loss of distal vessel volume and one with normal RV and relatively preserved intraparenchymal vasculature are shown in Figure 1.

Figure 1: Pulmonary vascular reconstructions.

A) shows an example of a sagittal CT image through the right lung in a patient with acute pulmonary embolism and regional loss of visible vasculature in multiple regions including the right middle lobe (Red Circle). B) shows the vascular reconstruction for the same subject as (A) but shows the loss of vessels in the right middle lobe. Coloring represents vessel size with red representing small vessels and blue representing large vessels. C) shows a vascular reconstruction from the right lung of a subject with acute pulmonary embolus but no evidence of RV dysfunction with preserved small vascular volume. D) shows a patient showing loss of distal vasculature and with RV dysfunction.

Measurement of RV:LV End-Diastolic Diameter Ratio

The RV:LV ratio was measured from the CTA in the axial plane for each patient as described previously^16–18. Diameter for each chamber was measured between the endocardial border and the interventricular septum using a line drawn perpendicular to the long axis of heart. Specific axial planes were selected such that the maximum diameter for each ventricle was used to calculate a ratio as illustrated in Figure 2. We defined a CT derived RV:LV ratio of ≥ 1 as abnormal.

Figure 2: RV:LV Ratio measurements in two patients.

Measurements were performed on axial images, perpendicular to the longitudinal axis and extend to the endocardial border. The axial sections were chosen to measure the maximal extent for each ventricle. The first patient (Images A and B) has a RV:LV ratio of 1.38. The second patient (Images C and D) has a RV:LV ratio of 0.97. Both subjects had evidence of RV dysfunction by echocardiogram and had elevated biomarkers.

Statistical Methods

Data are presented as medians and interquartile range. Clinical variables and distal vascular volumes were compared between groups using Wilcoxon rank sum tests. Small venous was the primary variable of interest in our analyses. Univariable logistic regression models were run, to assess the association of small venous volume with outcomes of interest – abnormal RV on TTE and abnormal serum biomarkers. Separate models were run for RV dilation, RV dysfunction, and RV strain.

We then ran multivariable logistic regression models adjusted for a priori selected variables, including age, sex, lung volumes, and small arterial volumes. Adjustment by lung volume was performed to account for variability in blood volume based on lung size while adjustment by arterial volume was chosen to account for variability in regional perfusion. Dependent variables in models were binary. There was no multicollinearity noted among variables. Observations were noted to be independent. We performed a sensitivity analysis whereby missing values for missing echo variables were imputed using multiple imputations by chained equations. Results were reported as relative risks to allow more intuitive understanding by clinicians.

Cox proportional hazard models, adjusted for covariates described above in addition to the presence of malignancy and RV:LV ratio on CT, were used to assess all cause 30- and 90-day mortality. We considered evaluation of only cardiovascular mortality, but due limited number of events, a subset was analyzed and reported in the supplement.

As a secondary analysis, multivariable logistic regression models adjusted for age and sex were used to assess the relationship between an abnormal RV:LV ratio on CT imaging with abnormal appearing RV on TTE and cardiac biomarkers. P-values < 0.05 were considered statistically significant. All statistical analyses were performed in R 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria) and Stata 16.

Results

RPDR was used to identify 774 subjects meeting the inclusion criteria with acceptable CT images as illustrated in Figure 3. Of these, 52 were excluded due to reduced image quality including severe motion artifacts, so 722 patients were included in the final analyses. Demographics and characteristics of the subjects are shown in Table 1. The cohort was mostly white (75%) with a slight female predominance (56%). A total of 573 patients had echocardiograms available. Of these, 571 commented on RV dilation, 561 on RV dysfunction and 276 on RV strain. These measures were abnormal in 62%, 54% and 67%, respectively. Troponin levels were measured in 602 patients and were elevated in 295 patients. BNP levels were measured in 374 and were elevated in 172. The most frequently used treatment was anticoagulation (98%), with some patients receiving additional catheter directed therapy (9%), systemic tissue plasminogen activator (tPA) (1.8%), or inferior vena cava (IVC) filter placement (8.3%). 321 patients (44.5%) within the cohort had an underlying malignancy. 65 patients died within 30 days, and 113 died within 90 days of the diagnosis of acute PE.

Figure 3: Patient selection for cohort generation.

RPDR= Research patient data registry, CTA = CT angiogram, RV = Right Ventricle, BNP = B-type natriuretic peptide.

Table 1:

Patient characteristics in the entire cohort

	N = 722
Age (Median, IQR)	63 (51 – 73)
Female Gender (N, %)	402 (55.7%)
Race
Non- Hispanic White	544 (75.3)
African American	102 (14.1)
Hispanic	29 (4.0)
Asian	9 (1.2)
Other	13 (1.8)
Indian/Alaskan	2 (0.3)
Not Recorded	23 (3.2)
Cancer (N, %)	321(44.5)
Echo (Abnormal/total measured (% abnormal))
RV dilation	353/571 (61.8)
RV dysfunction	301/561 (53.7)
RV strain	185/276 (67.0)
Serum Markers (Abnormal/total measured (% abnormal))
Troponin	295/602 (49.0)
BNP	172/374 (46.0)
Treatment (N, %)
Anticoagulation	714 (98.9)
TPA	21 (2.9)
TPA via CDT	8 (1.1)
Systemic TPA	13 (1.8)
Mechanical CDT	59 (8.2)
IVC filter	60 (8.3)
CT Measures (Median, IQR) in ml
Total blood volume	195.27 (158.04 – 234.62)
Total lung volume	3.077 (1.06 – 4.13)
Total Arterial volume	129.67 (108.53 – 156.79)
Total Venous volume	64.18 (48.94 – 80.88)
Small vessel volume	40.61 (36.01 – 45.11))
Small venous volume	14.41 (11.91 – 16.47)
Small arterial volume	26.14 (23.38 – 29.22)
RV:LV ratio	1.2 (1.06 – 1.37)
Mortality (N, %)
30 days	65 (9.0)
90 days	113 (15.65)

An abnormal RV:LV ratio on CT was associated with increased risk of an abnormal RV on TTE (RR: 3.5, 95%CI: 2.75 – 4.48, p<0.0001) and elevated cardiac biomarkers (troponin: RR:3.30, 95%CI:2.39 – 4.57, p<0.0001) but was not a significant predictor of 30- or 90-day mortality. (Supplement Table IA and IB).

Automated vascular reconstructions were performed as outlined in methods above, examples of which are shown in Figure 1. Small vessel volume quantified from CT scans was lower in patients with abnormal RV when compared to those with normal appearing RV assessed by TTE (Table 2). Patients with impaired RV function had a lower median small venous volume compared to those with intact RV function (13.4ml vs 14.9ml, p <0.001). This difference was also seen in the arterial small vessels (26.5ml vs 25.7ml, p=0.02), but was more pronounced on the venous side. Similarly, those with elevated cardiac biomarkers had lower small venous volumes compared to those with normal cardiac biomarkers levels (15.07ml vs 13.42ml, p <0.001).

Table 2:

CT derived metrics and their values in the complete cohort, patients with normal/abnormal RV on echocardiograms and those with normal/elevated biomarkers. All values are presented as median (IQR)

	Full Cohort	RV Parameters on Echocardiogram			Cardiac biomarkers
CT Measures, ml		Normal	Abnormal	p value	Normal	Elevated	p value
Distal blood volume	40.61 (36.01 – 45.11)	42.07 (37.22 – 46.33)	39.41 (34.77 – 43.55)	<0.001	41.71 (37.20 – 45.92)	39.36 (34.38 – 43.74)	<0.001
Distal venous blood volume	14.41 (11.91 – 16.47)	14.93 (12.73 – 17.18)	13.4 (11.09 – 15.52)	<0.001	15.07 (12.78 – 17.15)	13.42 (10.95 – 14.45)	<0.001
Distal arterial blood volume	26.14 (23.38 – 29.22)	26.47 (23.67 – 29.65)	25.67 (22.85 – 28.73)	0.02	26.52 (23.63 – 29.50)	25.63 (22.79 – 28.74)	0.013
RV:LV ratio on CT	1.2 (1.06 – 1.37)	1.11 (0.99 – 1.25)	1.32 (1.16–1.46)	<0.001	1.14 (1.02 – 1.27)	1.3 (1.13 – 1.46)	<0.001

Univariable analyses showed that loss of small venous volume was associated with an increased risk of abnormal RV on TTE and elevated cardiac biomarkers These models were then adjusted for age, sex, lung volumes, and small arterial volumes. A 1 standard deviation (SD) decrease in small venous blood volume was associated with an increased risk of RV dilation (RR:1.38, 95%CI:1.20 – 1.58, p<0.001), RV dysfunction (RR:1.63, 95%CI:0.71 – 1.86, p<0.0001), and RV strain (RR:1.68, 95%CI:1.41 – 2.08, p<0.0001). A 1 SD decrease in small venous volume was also associated with an increased risk of elevated serum cardiac biomarker levels. Decrease in small venous volumes was associated with a rise in levels of both troponin (RR, 1.66 [95% CI, 1.42–1.98], P<0.0001) and BNP (RR, 1.64 [95% CI, 1.30–2.10], P<0.0001) (Table 3). 1 SD of venous blood volume was 17.84 mL in this cohort. These results were reproducible in a sensitivity analyses that accounted for missing echocardiographic and biomarker data (Supplement Table II).

Table 3:

Multivariable logistic regression models assessing the association decrease in small venous volume with right ventricular function on TTE and cardiac biomarkers after adjusting for age, sex, lung volume and distal arterial volumes. Relative risk is reported per 1 standard deviation (17.84ml) decrease in small venous blood volume.

Parameter	RR	CI	P value
Abnormal RV on Echo	1.35	1.19 – 1.55	<0.0001
Right ventricular dilation	1.38	1.20 – 1.58	<0.0001
Right ventricular dysfunction	1.63	0.71 – 1.86	<0.0001
Right ventricular strain	1.68	1.41 – 2.08	<0.0001
Elevated Troponin	1.66	1.42 – 1.98	<0.0001
Elevated BNP	1.64	1.30 – 2.10	<0.0001

A 1 SD reduction in small venous volume was also independently associated with 30-day increased mortality (HR:2.52, 95%CI:1.51 – 4.45, p<0.001) and 90-day increased mortality (HR:1.66, 95%CI: 1.10 – 2.50, p=0.016) after multivariate adjustment for a diagnosis of malignancy and abnormal RV:LV ratio in addition to covariates above. (Table 4, Figure 4) The associations remained significant for 30-day mortality, with a trend towards significance for 90-day mortality when adjusting the above models for elevated troponin levels (Supplement Table III). In the analysis of a subset of these patients with only cardiovascular mortality, the associations between small venous volume and mortality were maintained. (Supplement Table IV)

Table 4:

Cox proportional hazard model assessing the Hazard Ratio of a reduction in small venous blood volume with 30- and 90-day mortality after adjusting for age, sex, lung volume, small arterial blood volume, abnormal RV:LV ratio and presence of cancer. Hazard's ratios reported here are per 1 standard deviation (17.84ml) decrease in small venous volume

	Univariable Analysis			Multivariable Analysis
Mortality	HR	CI	P value	HR	CI	P value
30-day	1.47	0.95 – 2.32	0.08	2.52	1.51 – 4.45	<0.001
90-day	1.06	0.68 – 1.33	0.77	1.66	1.10 – 2.50	0.016

Figure 4. Kaplan Meier survival curves for patients with acute submassive pulmonary embolism.

Solid green line represents patients above the 50th percentile of small venous volume. Dashed red line represents patients below the 50th percentile of small venous volume.

Discussion

In this single center retrospective study of patients with acute submassive PE, loss of small venous volume was associated with increased risk of abnormal RV size and function on TTE and abnormal cardiac biomarkers. Additionally, loss of small venous volume was also associated with increased risk of 30- and 90-day mortality, even after adjusting for an abnormal RV:LV ratio and the presence of underlying malignancy.

The loss of small pulmonary vascular volume has been previously described and often referred to as pruning. This finding has been observed in several chronic lung conditions, including pulmonary vascular diseases, and when present has been associated with disease severity, morbidity, and mortality.³ In acute PE, loss of small vessels (focal oligemia) has been described as Westermark’s sign on plain radiographs^{19, 20}. Prior studies have described differential ventilation and worsening gas exchange in relationship to this paucity²¹. However, the overall prognostic importance of its presence has not been studied extensively or quantitatively. In a prior, smaller study we demonstrated that loss of small venous volume was associated with increased CT derived RV to LV volume (obtained from 3D reconstructions of the heart). In this study we demonstrate the quantification of the small venous volume is associated with functional metrics of abnormal appearance of RV on TTE, abnormal cardiac biomarkers, as well as increased mortality.

The loss of small venous volume in acute PE and its association with an abnormal appearance of the RV on TTE has multiple potential explanations. Obstructive clot burden leads to regional hypoperfusion. Pulmonary veins contain less smooth muscle and elastic intima than pulmonary arteries.²² Their inherent collapsibility may explain the relative loss of venous as compared to the arterial volume observed in this study. Decreased lumen cross-sectional area leads to increased pulmonary vascular resistance which is a non-pulsatile component of the increased afterload experienced by the right ventricle.²³ However, a significant portion of the work performed by the right ventricle is related to pulsatile flow and the capacitance of the pulmonary circulation which is an important prognostic marker in pulmonary hypertension.²⁴ The capacitance has been previously described to derive from the proximal and intermediate arteries^{25, 26} but appears to be particularly sensitive to proximal pressure rise in occlusion experiments²⁷ where there is presumably no remodeling of the vessel walls. The decrease in distal blood volume may then reflect the integration of increased resistance in the occluded regions and decreased capacitance due to increasing pressures in the network. It is also plausible that reduced cardiac output results in a delay in contrast reaching the venous system. Based on timing of image acquisition in relation to contrast administration, this delayed flow of contrast may contribute to the observed reduced small venous vascular volume.

The association between loss of small venous volume and mortality may be caused by an impaired RV, which would cause decreased distal perfusion and resultant vascular loss. However, we found that the loss of small venous volume was an independent predictor even in models adjusted for RV:LV ratio. In our cohort, an abnormal RV:LV ratio was a strong predictor of an abnormal RV on TTE but did not predict an increased mortality risk (Supplement table IA and IB). It is possible that while RV/LV ratio is useful in prognostication in the general PE population, that its prognostic capacity is limited once the analysis is limited to submassive PE. Indeed, most subjects in this cohort already had elevated RV/LV ratios.

There are several limitations to this study. This is a retrospective study which relies on the ICD-9 and ICD-10 codes for cohort generation. The validity of these codes to specifically identify patients with submassive PE has not been studied extensively in the inpatient setting. Several studies have looked at validity of these codes for the inpatient diagnosis of venous thromboembolism and have reported variable sensitivities and specificities. However, these codes have performed best in patients at a higher risk of VTE.²⁸ The social security administration death master file (SSDMF) was utilized to assess mortality for this cohort. SSDMF has appeared to underestimate overall mortality in prior studies.^29–30 However, the estimates of 30 day post hospital discharge mortality have been more accurate³¹. We anticipate that this underestimation of mortality would bias our results to the null. While we included all-cause mortality as an endpoint, we were unable to ascertain cardiovascular mortality for patients who expired outside of the BWH health care system (Supplement Table IV). The methods used for measurement of vascular volume have been largely developed and tested in chronic lung conditions and have not been optimized for detection of disease in acute pulmonary embolism. Specifically, contrast bolus timing may have led to variable detection of vasculature between the arterial and venous beds. The arterial venous labeling has been validated in multiple disease states and scan types but may also be affected by the specific scans or disease context. Additionally, since these studies were obtained during routine clinical care, there may be additional scanner specific variability that our study could not account for. Concurrent existence of lung disease, such as emphysema or those associated with air trapping, could also lead to decreased vascular volume. Finally, 149 patients within this cohort had missing TTE data. We performed a sensitivity analysis to account for the missing data and noted no significant changes in our results.

In summary, the loss of small vessel volume may be a useful CT based marker for assessing disease severity and can be obtained in an automated fashion from clinically available CT imaging at the time of diagnosis producing measurable differences in the vascular morphology of patients with submassive PE. These differences appear to be useful for both stratification and prognostication. Further work is needed to determine if CT based imaging biomarkers of vasculature could be used to enable precision therapy in the setting of acute PE or in detecting already established conditions such as chronic thromboembolic disease. Additionally, quantifying the relationship between changes in vascular morphology, clot burden, and RV dysfunction as well as clinical data may help us understand the pathophysiologic evolution of pulmonary embolus and impaired RV function. Combining vascular reconstruction with perfusion maps derived from the same CT scan using emerging technologies such as Dual Energy CT may provide a complete measurement of micro as well as macro vascular changes in acute pulmonary embolism. Prospective multicenter studies of the baseline imaging may allow for validation of this measurement as well as discovery of other derived imaging markers that will allow for better subtyping and assessment of potential response to intervention.

Clinical Perspective.

Pulmonary embolism is a life-threatening condition with significant morbidity and mortality. There is increasing appreciation for the importance of systematic assessment and nuanced treatment upon presentation. This has led to the development of novel methods of establishing risk stratification and treatment of pulmonary embolism. With this interest comes the opportunity to examine quantitative methods of assessing the potential state of the cardiopulmonary system in light of the presence of the clot, methods that in time may allow a more refined approach to selecting patients for different treatment modalities and monitoring for long-term sequelae. Oligemia, the loss of distal blood vessels in imaging, has been described in pulmonary embolism until recently has not been quantified in imaging. In this single-center retrospective study, of patients with acute, intermediate-risk pulmonary embolism. This class of patients has significant room for the investigation of biomarkers to guide different types of treatment. We sought to measure the loss of vascular volume in the CT scans obtained at the time of diagnosis using fully automated methods. We showed that loss of small vascular volume was associated with echocardiographic right ventricular dysfunction, elevated cardiac serum biomarkers, and mortality. These findings present a crucial step towards developing tools that leverage multiple aspects of imaging available at the time of diagnosis to improve prognostication. This study paves the way for future analyses targeting assessment of the impact of interventions, such as embolectomy or catheter-directed therapy, on lung perfusion.

Abbreviations list

BNP

type B natriuretic peptide

COPD

Chronic pulmonary obstructive disease

CTA

Chest computed tomography angiography

CTEPH

chronic thromboembolic pulmonary hypertension

IVC

inferior vena cava

LV

left ventricle

PAH

Pulmonary arterial hypertension

PE

Pulmonary embolism

PESI

pulmonary embolism severity index

RPDR

Research Patient Data Registry

RV

Right ventricle

SSDMF

social security administration death master file

tPA

tissue plasminogen activator

TTE

Echocardiography