Contrast Reflux Into IVC in Acute Pulmonary Embolism Predicts Clinical Deterioration Risk

ABSTRACT

Reflux of contrast medium into the inferior vena cava (IVC) on computed tomography pulmonary angiogram (CTPA) is an independent risk factor for mortality in patients with acute pulmonary embolism (PE). This study's aim was to determine if reflux into the IVC correlates with objective scores of clinical risk in acute PE. Patient‐level data were collected for adults diagnosed with acute PE by CTPA at University of California, San Diego Health between January 1 and June 30, 2023. Contrast reflux into the IVC was graded on a 4‐point scale. The primary endpoint was the correlation between reflux severity and risk for clinical deterioration as measured by the National Early Warning Score (NEWS). The secondary endpoint was interobserver reliability grading reflux amongst 16 reviewers. Fifty‐six subjects with acute PE were included. Extent of contrast reflux into the IVC correlated with NEWS (Spearman's ρ 0.2932; p = 0.028). Subjects with grade 3 reflux had increased NEWS (7.80) compared to those with reflux grades 0 (3.63; p = 0.014) or 1 (4.29; p = 0.05). NEWS was elevated in those with contrast below the diaphragm compared to those without when grading was pooled (p = 0.036). Interobserver reliability for reflux grading was “moderate” (κ = 0.58), yet increased to “almost perfect” (κ = 0.87) for pooled grading. Contrast reflux into the IVC during acute PE correlates with risk for clinical deterioration, as reflected by NEWS. Simplification of grading to contrast above or below the diaphragm improves reliability without compromising this relationship.

The clinical course for patients diagnosed with acute pulmonary embolism (PE) ranges from outpatient treatment of an incidental discovery to critical illness with right heart failure, hemodynamic collapse, and death [1, 2]. However, accurate prediction of which patients are at risk for clinical deterioration remains challenging and continues to be a key focus of research. Although computed tomography pulmonary angiography (CTPA) is used to diagnose the overwhelming majority of these patients, clot burden itself does not correlate well with outcomes [3, 4, 5]. Therefore, risk‐stratification calculators such as the pulmonary embolism severity index (PESI) and its simplified version (sPESI) continue to rely heavily on vital sign parameters and history to guide prognostication and management [6, 7]. Recently, Bavalia et al. validated a more universal system, the National Early Warning Score (NEWS) in acute PE [8, 9]. When compared to PESI and sPESI scores, NEWS was found to have similar, if not improved predictive utility with regard to 7‐day admission to the intensive care unit (ICU) and 30‐day mortality. Despite its broader application and potential superiority, similar to PESI and sPESI, NEWS remains a conglomerate calculation of nonspecific vital signs and observations.

In contrast to this, various studies have evaluated the use of specific CTPA findings in the prognosis of acute PE. Other than clot burden, measures including right ventricle‐to‐left ventricle (RV/LV) diameter ratio, RV/LV volume ratio, right atrium‐to‐left atrium (RA/LA) volume ratio, pulmonary artery (PA) diameter, and interventricular septal bowing have all been evaluated for correlation with 30‐day mortality [5, 10, 11]. Results have been mixed, and more immediate patient outcomes related to these findings have not been well studied. One exception to this has been contrast reflux into the inferior vena cava (IVC), which has been shown to correlate both with 30‐day mortality and the more immediate outcome of 24‐h mortality in acute PE [12].

We performed this study to determine if reflux of contrast into the IVC may function as a means of risk‐stratification for patients with acute PE. To investigate the relationship between the degree of contrast reflux into the IVC on CTPA and risk of clinical deterioration as measured by NEWS, we designed a retrospective study collecting clinical, laboratory, and radiographic data of patients diagnosed with acute PE at a large academic healthcare system. To further determine the feasibility of grading IVC reflux severity in practice, we provided brief instructional material and performed interobserver reliability testing with physicians of various levels of training and background.

1. Methods

1.1. Study Population and Design

We performed a retrospective observational cohort study of adult patients diagnosed with acute PE at University of California, San Diego Health between January 1, 2023 and June 30, 2023, as identified from hospital records through diagnostic codes. Patient characteristics (including age, sex, body mass index, self‐identified heritage, chronic and underlying disease diagnoses, and medications), bedside monitoring details at presentation (including vitals signs, symptom reports, and nursing documentation), laboratory results (cardiac enzymes and blood counts), radiographic images (CTPA, vascular ultrasonography, and echocardiography), administered therapeutics or interventions (anticoagulation, antiplatelet therapy, suction thrombectomy, and any associated adverse events), and other clinical course documentation (including hospitalization details, consultation to PE team, and mortality) at the time of acute PE diagnosis were collected. Of the 91 subjects identified in the study timeframe, 56 were included in the final cohort after exclusions as detailed in Figure 1. Exclusion criteria included those patients without CTPA imaging available for review, those with prior diagnosed pulmonary hypertension, and those with chronic rather than acute PE. The study was approved by the Institutional Review Board (IRB # 210377).

Figure 1.

Flow diagram of the study population. Fifty‐six patients with a diagnosis of acute pulmonary embolism were included in the final review, January 1, 2023, to June 30, 2023.

1.2. Variables and Measures

The NEWS is a clinical deterioration risk calculator validated in acute PE [8, 9]. NEWS was chosen for this study for its utility as a tool to predict clinical worsening early in the course of acute PE management, whereas scores such as PESI and its simplified version focus instead on mortality. The total score ranges from 0 to 20 points, the sum of six variable inputs valued on a 0–3 point scale, along with one additional input on a 0–2 point scale (the use of supplemental oxygen). Higher scores correlate with an increased risk of 24‐h unanticipated ICU admission or death in all‐comers. Scores of 7 or higher are considered in need of prompt critical care evaluation due to a high risk of decompensation and 24‐h mortality. Calculation of the NEWS for each patient was performed retrospectively from data recorded at the time of acute PE diagnosis. Similar to the Bavalia et al. validation study [8], clinical documentation of “altered mental status” was used in lieu of the NEWS variable “Alert Voice Pain Unresponsiveness” (AVPU), with 0 points marked in case of no altered mental status and 3 points added in its presence.

For each study patient, CTPA was performed at the time of the diagnosis of acute PE using a multi‐slice CT scanner and iodinated contrast medium injected into an upper extremity vein or other tributary to the superior vena cava using the test bolus technique [12]. These studies underwent independent image review to measure the degree of contrast reflux into the IVC using coronal series. The degree of reflux was assessed according to a 4‐point scale: grade 0 = no reflux of contrast into the IVC; grade 1 = reflux of contrast limited to the subcardiac IVC, superior to the liver and diaphragm; grade 2 = intrahepatic reflux of contrast into the IVC; grade 3 = subhepatic reflux of contrast into the IVC (Figure 2) [13]. Severity grading for each CTPA was then repeated independently by a second physician reviewer, and confirmed via independent review by a senior board‐certified chest radiologist.

Figure 2.

Schema for grading reflux of contrast into the inferior vena cava with coronal computed tomography imaging. (A) 0 (no reflux), (B) 1 (suprahepatic reflux), (C) 2 (intrahepatic reflux), and (D) 3 (infrahepatic reflux).

Sixteen physician participants independently reviewed and scored the IVC reflux severity on CTPA for each of the 56 study subjects. Reviewers were blinded to the patients' NEWS scores. The reviewers were organized into four groups of four members: internal medicine residents; pulmonary and critical care medicine (PCCM) fellows; PCCM faculty specializing in venous thromboembolism, and chest radiology faculty. The observers were trained with a standardized 9‐min instructional video tutorial detailing the scoring system (created for the purposes of this study) prior to grading the scans. No clinical history or additional clinical context was provided. Reviewers were provided with only coronal series imaging for each patient to perform their grading assessment. However, each reviewer was able to adjust image display parameters, including brightness and contrast, via digital interface. There were no time constraints. Observer scores were collected via predesigned survey.

1.3. Statistical Analysis

Demographic data and clinical characteristics were described with descriptive statistics. Correlation between NEWS and graded severity of IVC contrast reflux on CTPA was evaluated using Spearman's rank correlation coefficient [14]. Further stratification by reflux severity was analyzed using analysis of variance. The Fleiss' kappa statistic was used to evaluate the reliability between study observers for each assigned reflux severity grade [15]. Observers were grouped for analysis based on their training level, as well as assessed altogether as a single cohort. Levels of agreement (κ value) matched those proposed by Landis and Koch [16], with kappa values of 0.00–0.20 considered slight agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 substantial agreement, and 0.81–1.00 almost perfect agreement. Statistical analysis was carried out using Qualtrics software (Qualtrics, Provo, UT) [17].

2. Results

Baseline characteristics of the 56 study subjects, including coexisting medical conditions, symptoms reported at time of diagnosis, and level of most proximal clot detected on CTPA are summarized in Table 1. The most common symptoms at presentation were dyspnea (39.3%) and chest pain (37.5%). The most proximal emboli were in the main pulmonary arteries in 19 (33.9%) subjects, in the lobar or interlobar arteries in 6 (10.7%), segmental arteries in 18 (32.1%), and subsegmental in 13 (23.2%). All‐cause in‐hospital mortality for the cohort was 11/56 (19.6%) though only two of these deaths were attributed to acute PE.

Table 1.

Baseline clinical characteristics of study patients.

Characteristic	Patients (N = 56)
Mean age (range) – years	60 ± 16 (20‐90)
Sex – no. (%)
Female	41 (73.21)
Male	25 (44.64)
Body mass index	28.01 ± 6.5
Self‐reported Ethnicity – no. (%)
Non‐Hispanic White	25 (44.64)
Other Hispanic, Latino(a), or Spanish Origin	10 (17.86)
Black	8 (14.29)
Asian	7 (12.50)
Other/Mixed Race	4 (7.14)
American Indian or Alaska Native	1 (1.79)
Unknown	1 (1.79)
Coexisting disorder – no. (%)
Cancer history	32 (57.14)
Cerebrovascular accident	3 (5.36)
Congestive heart failure	6 (10.71)
Connective tissue disease	4 (7.14)
Chronic obstructive pulmonary disease	5 (8.93)
Chronic kidney disease	2 (3.57)
Dementia	2 (3.57)
Diabetes mellitus	16 (28.57)
Hemiplegia	0 (0)
Human Immunodeficiency Virus infection	2 (3.57)
Liver disease	4 (7.14)
Myocardial infarction	2 (3.57)
Peptic ulcer disease	1 (1.79)
Peripheral vascular disease	3 (5.36)
Symptoms reported at diagnosis – no. (%)
Dyspnea	22 (39.29)
Chest pain	21 (37.50)
Cough	6 (10.71)
Extremity pain or swelling	5 (8.93)
Other pain	3 (5.36)
Syncope/Presyncope	3 (5.36)
Confusion	2 (3.57)
Weakness	2 (3.57)
Fever	1 (1.79)
Hemoptysis	1 (1.79)
Level of most proximal clot – no. (%)
Main pulmonary arteries, including pulmonary trunk	19 (33.93)
Lobar and Interlobar	6 (10.71)
Segmental	18 (32.14)
Subsegmental	13 (23.21)

Consensus interpretations by three physicians, including one chest radiologist disclosed no contrast reflux (grade 0) in 24 subjects (42.9%); reflux superior to the liver and diaphragm (grade 1) in 14 (25.0%); intrahepatic reflux (grade 2) in 13 (23.2%); and subhepatic reflux (grade 3) in 5 (8.9%). The degree of contrast reflux into the IVC correlated with NEWS (Spearman's ρ 0.2932, p = 0.028). Patients with grade 3 reflux had significantly higher NEWS than those with grade 0 (p = 0.014) or 1 (p = 0.050) reflux (Figure 3a). Inpatient all‐cause mortality was 19.6% (11/56), however only two of these deaths were attributed to PE by their care teams. Of the two deaths attributed to PE, one subject had grade 2 reflux and the other had grade 3. Table 2 highlights additional specific outcome data stratified by severity of contrast reflux into the IVC, including need for ICU admission, hospital length of stay, if suction thrombectomy was performed, and is vasopressors were needed. While much of this data trended toward significance, the only statistical significance was found in the increase in ICU admissions as the severity of contrast reflux increased (p = 0.003) overall. However, when comparing specific groups this was not preserved.

Figure 3.

National Early Warning Score stratified by severity grade of contrast reflux into the IVC in patients with acute pulmonary embolism. (a) Grade 3 reflux correlated to significantly higher NEWS when compared to those with grade 0 (p = 0.014) or 1 (p = 0.050) reflux. (b) Pooled reflux severity grading, groups organized by degree of reflux in relation to the diaphragm (0–1 above, 2–3 below). IVC reflux of contrast also correlated to NEWS with statistical significance (p = 0.036) between these pooled cohorts. Median NEWS with interquartile range is represented for each group with box plot, means noted with x.

Table 2.

Specific outcome data in acute PE stratified by degree of contrast reflux into the IVC.

Degree of reflux	0	1	2	3
All‐cause in‐hospital mortality	5/24 (20.8%)	1/14 (7.1%)	4/13 (30.8%)	1/5 (20.0%)
Death attributable to acute PE	0/24 (0.0%)	0/14 (0.0%)	1/13 (7.7%)	1/5 (20.0%)
Average hospital length of stay (days)	10.3	7.8	11.8	15.6
Required ICU admission or transfer	5/24 (20.8%)	4/14 (28.6%)	10/13 (76.9%)	3/5 (60.0%)
Suction thrombectomy performed	0/24 (0.0%)	2/14 (14.3%)	8/13 (61.5%)	1/5 (20.0%)
Need for vasopressors	1/24 (4.2%)	1/14 (7.1%)	3/13 (23.1%)	2/5 (40.0%)

Figure 4a illustrates the interobserver reliability of contrast reflux grading among independent readers, as evaluated by Fleiss' kappa. Overall reliability among all 16 reviewers was moderate (κ = 0.58). The κ score for internal medicine residents was 0.48, and for PCCM fellows was 0.53 (both “moderate” agreement). For PCCM faculty members specializing in venous thromboembolism κ was 0.70, and for chest radiology faculty was 0.67 (both “substantial” agreement) [16].

Figure 4.

Interobserver reliability by Fleiss' kappa (κ) for the measurement of reflux of CTPA contrast into the IVC in acute pulmonary embolism amongst different physician groups. (a) Reliability grading using the 0–3 scale. (b) Reliability using pooled severity grading into above (0–1) or below (2–3) the diaphragm.

Feedback from participants in the interobserver reliability portion of the study consistently noted that IVC contrast reflux grades 0 (no reflux) and 1 (subcardiac reflux only) were the most difficult to distinguish.

For these reasons, we performed post hoc analyses of both sections of the study, in which the 4‐point scale was simplified into two groups: contrast entirely above the diaphragm (grades 0 and 1) and reflux into the IVC below the diaphragm (grades 2 and 3). Subjects in whom consensus reading disclosed contrast reflux into the IVC below the diaphragm had significantly higher NEWS than those with contrast limited to above the diaphragm (p = 0.036), as illustrated in Figure 3b.

Consolidation of reflux grading into two groups dramatically improved interobserver reliability. The κ score improved to 0.79 (“substantial” agreement) for internal medicine residents, while it increased to 0.86 for PCCM fellows, 0.89 for PCCM faculty specializing in venous thromboembolism, and 0.96 for chest radiology faculty (all of which rated as “almost perfect” agreement) (Figure 4b). The overall score for the entire group improved to 0.87 (“almost perfect” agreement).

3. Discussion

Risk assessment in acute PE remains a challenge, even amongst highly trained subspecialty experts. Our study investigated the correlation between a single radiological finding in acute PE: the extent of contrast reflux into the IVC, and a patient's risk of clinical deterioration as estimated by the NEWS.

We demonstrated a correlation between severity of contrast reflux and NEWS. Moreover, this radiologic evaluation was demonstrated to be reliable among various physician groups at different stages of training (internal medicine residents, PCCM fellows and faculty, and chest radiologists) following a brief tutorial. We anticipate that similar reliability would be experienced among health professionals of various other backgrounds and training levels.

Interobserver reliability greatly increased when the degree of contrast reflux was dichotomized to a simple assessment of whether contrast reflux remained above or extended below the diaphragm, rather than the more complicated grades used previously [10, 11, 13, 18]. Those with contrast of reflux into the IVC below the diaphragm demonstrated a significantly higher risk of clinical deterioration compared with those who did not, as estimated by NEWS. Those with subdiaphragmatic contrast had average NEWS scores of 3.9, whereas those without had scores of 5.9, which have been associated with 81% and 64% specificities for the development of critical illness or death, respectively [8].

Our findings suggest that reflux of contrast into the IVC below the diaphragm should be evaluated by CTPA at the time of acute PE diagnosis. We speculate that the extent of contrast reflux can assist in determining appropriate patient level of care, involvement of expert consultants, and can further aid in the decision to pursue advanced therapies such as inotropic support or suction thrombectomy [19, 20, 21].

Our study has several limitations. First, the sample size is modest and only includes patients diagnosed at a single academic healthcare system. Additionally, the findings associated with pooled reflux severity grading were the result of post hoc analysis. This analysis we felt to be beneficial as it directly addressed the study question and simplified the grading schema, in the case of already statistically significant correlation between reflux severity and NEWS. Finally, while we included several groups of physicians, the benefit of our instructional video and degree of reliability assessing IVC reflux on CTPA could be further assessed by including other healthcare professionals, such as emergency department physicians, general radiologists, or advanced practice providers.

We anticipate that with concise instruction, a simple means of risk stratification for patients with acute PE can be implemented among clinicians with various levels of training and experience. Moving forward, interventions that simplify risk stratification, such as this, may better equip providers to more appropriately seek expert consultation or escalate care.

Author Contributions

All authors contributed significantly to the conception, design, data collection, data analysis, drafting, and/or editing of this work. Nicholas A. Leverone, Alisse G. Singer, Jenny Z. Yang, Elizabeth Weihe, and Timothy M. Fernandes contributed substantially to the conception and initial design of the study. Amanda A. Lopez, Charlotte Ellberg, Patrick J. Henry, Alisha A. Kabadi, Alyssa A. Self, Danielle M. Leverone, Andrew Yen, Lewis Hahn, Jennifer Karunamuni, W. Cameron McGuire, Demosthenes G. Papamatheakis, and Timothy A. Morris contributed substantially to refinement of the design, as well as to acquisition, analysis, or interpretation of data for the study. All authors provided critical revisions to manuscript drafts with regard to content and language, and all authors confirmed final approval of the version to be published.

Funding

The authors received no specific funding for this work.

Ethics Statement

The study was approved by the Institutional Review Board (IRB # 210377).

Conflicts of Interest

Timothy A. Morris and Timothy M. Fernandes report research support, unrelated to this project from Inari. Timothy M. Fernandes has been a paid consultant for Penumbra. This research has no sponsors.