Introduction

Impairment of right ventricular (RV) function resulting from acute RV pressure overload plays a critical role in the pathophysiology and prognosis of pulmonary embolism (PE) [1, 2]. In particular, the combination of RV dilatation with RV ischaemia, injury and inflammation may lead to overt RV failure causing haemodynamic instability and death [3]. RV dysfunction, indicated by abnormal echocardiographic signs or elevated cardiac biomarkers, has been shown to predict short-term mortality in patients with PE even in the absence of clinically evident haemodynamic compromise at presentation [4], while signs of RV dysfunction at discharge have previously been associated with PE-related death [5].

Among survivors of the acute phase of PE, approximately 20% have been reported to present with persistent RV dysfunction at follow-up [5]. In fact, in patients with intermediate-risk PE participating in the Pulmonary Embolism Thrombolysis (PEITHO) trial, absence of complete RV recovery at 6 months, as assessed by echocardiography, predicted persisting RV dysfunction over the entire two-year follow-up period [6]. However, the definition of RV dysfunction has not been standardised, with various echocardiographic parameters having been used over the years, while laboratory values have also been considered in some cohorts [7,8,9].

The prospective multicentre single-arm Pulmonary Embolism International Trial (PEITHO)-2 (ClinicalTrials.gov Identifier NCT02596555, EudraCT Identifier 2015-001830-12) investigated if the early switch from parenteral heparin to oral anticoagulation using dabigatran in patients with intermediate-risk PE is effective and safe [10]. The present predefined analysis from the PEITHO-2 study sought (a) to determine the temporal pattern of recovery of RV function, as assessed by echocardiographic and biochemical parameters, in the PEITHO-2 study population; and (b) to identify baseline predictors of RV dysfunction during follow-up.

Methods

The rationale and design of the PEITHO-2 study have been previously described [11]. The main inclusion criteria in the study were an age of at least 18 years and the objective diagnosis of intermediate-risk PE, based on the classification proposed by the 2014 European Society of Cardiology (ESC) guidelines [12]. Key exclusion criteria were: pregnancy, reduced life expectancy, haemodynamic instability at presentation, presence of active bleeding or high risk for bleeding, contraindications to dabigatran, need for long-term anticoagulation/reperfusion treatment and impaired kidney/liver function. The primary efficacy endpoint was recurrent symptomatic venous thromboembolism (VTE) or PE-related death within 6 months after the index PE event.

According to the study protocol, echocardiographic and laboratory examinations were performed at baseline, i.e. upon enrolment, as well as at the 6-day and 6-month follow-up. To permit a standardised, coherent and complete assessment of the echocardiographic follow-up and comparison with the baseline status, all measured echocardiographic parameters (predefined; based on the protocol of the PEITHO-2 study [11]) were prospectively classified for the present analysis into four categories or groups, each one corresponding to a key manifestation of RV pressure overload cardiac imaging (Table 1): (i) RV size; (ii) RV free wall and tricuspid annulus motion; (iii) RV pressure overload; and (iv) right atrial (RA) and central venous pressure. This classification was not designed on the assumption that the above groups of findings are pathophysiologically ‘independent’ from each other; instead, it was implemented to ensure complete and reproducible echocardiographic reports in each patient and at each visit, based on the main pathophysiologic mechanisms implicated in RV dysfunction. In that sense, abnormal RV size (dilatation) was primarily confirmed by the documented right-to-left ventricular (RV/LV) end-diastolic diameter ratio; as a second option, if this parameter was missing, by the basal (D1) end-diastolic diameter of the RV measured in the 4-chamber view; and as a third option, if both of the above parameters were not available, by the RV end-diastolic diameter measured in the parasternal view. With a similar rationale, reduced RV free wall and tricuspid annulus motion was primarily evaluated by the tricuspid annular plane systolic excursion (TAPSE); or, if TAPSE was missing, by visual confirmation of RV free wall hypokinesia. RV pressure overload was indicated by visual confirmation of paradoxical septal wall motion; or by an increased LV eccentricity index indicating septal flattening and LV diastolic compression; and, as a third option, by an elevated (estimated) RV systolic pressure documented by measuring the tricuspid regurgitant jet velocity and calculating systolic RV pressure via the Bernoulli equation. Finally, increased RA and central venous pressure were primarily diagnosed by the documented absence of inspiratory collapse of the inferior vena cava and semi-quantitative estimation of right atrial pressure; or by the presence of pericardial effusion not explained by an alternative diagnosis (Fig. 1).

Table 1 Echocardiographic criteria for detecting and classifying right ventricular dysfunction at baseline and during follow-up, with corresponding cut-off values
Fig. 1
figure 1

Algorithm for echocardiographic assessment of right ventricular function during follow-up based on four key categories of ultrasound parameters. D1 basal end-diastolic diameter, IVC inferior vena cava, LV left ventricular, RA right atrial, RV right ventricular, SRVP systolic right ventricular pressure, TAPSE tricuspid annular plane systolic excursion, TRV tricuspid regurgitant jet velocity

The analysis of echocardiographic data required assessment of each one of the four categories as defined above, both at baseline and during follow-up. The ‘hierarchical’ order applied to define an abnormal status within each category was based on existing evidence and expert consensus on the relative prognostic strength and reproducibility of individual ultrasound parameters [1, 13]. Comparing the echocardiographic parameters at different time points and using the cut-off values shown in Table 1, the course of RV function during follow-up was described as: (a) persistently abnormal; (b) deteriorating; (c) improving; or (d) remaining normal. The laboratory parameter used for the assessment of RV dysfunction was the N-terminal pro-brain natriuretic peptide (NT-proBNP) level. The levels of NT-proBNP were considered abnormal if being above the cut-off value of 125 pg/ml [14, 15].

Statistical analysis was performed in the intention-to-treat population [10]. Categorical variables are reported with absolute and relative frequencies; continuous variables, with the corresponding median and interquartile range. Alluvial plots were designed to depict the course of RV function, as assessed by echocardiography, during follow-up. Univariable and multivariable logistic regression models were applied for examining predictors of abnormal RV function at 6 days and 6 months. The variables included in the models were selected on the basis of existing literature and current medical knowledge [1]. The results are presented as odds ratios with the corresponding 95% confidence intervals. The R software (R: A language and environment for statistical computing. R Foundation for Statistical Computing) was used for the statistical analysis.

Results

A total of 402 patients with intermediate-risk PE (48% women, median age of 69.5 years; 70% in the intermediate-high risk category) were enrolled in the PEITHO-2 study between January 2016 and July 2019. Patients were followed for 180 days after enrolment. During follow-up, 7 (2%) patients developed recurrent symptomatic VTE or PE-related death, 8 (2%) died from any cause and 11 (3%) had major bleeding [10].

Echocardiographic signs of RV dysfunction based on any abnormal parameter were present in 84% (n = 338) of the patients at baseline. RV dilatation was the most frequently abnormal echocardiographic finding (163 patients, 40.6% of the total study population), followed by increased RA and central venous pressure (139 patients, 34.6%), RV pressure overload (129 patients, 32.1%), and reduced RV free wall and tricuspid annulus motion (84 patients, 20.9%; Supplementary Table S1). The baseline characteristics of the patients with complete as opposed to those without completely available echocardiographic data at 180-day follow-up are shown in Table S2. After the acute phase of PE, the frequency of all markers of RV dysfunction decreased substantially; at 6 months, RV dysfunction had recovered in the vast majority of the patients, with RV enlargement and dilated inferior vena cava being the only findings documented somewhat more frequently, in 8.2% and 16.4% of the patients, respectively (Table S1).

Figure 2 displays the temporal changes of RV function across visits as assessed by echocardiography. As early as day 6, remaining normal or improved (from baseline) RV size was observed in 260 patients (64.7%), RV free wall and tricuspid annulus motion in 301 (74.9%), RV pressure overload in 297 (73.9%), and RA and central venous pressure in 254 (63.2%) patients. At day 180, the frequency of findings remaining normal or having improved rose slightly to reach 68.7% for RV size, 76.1% for RV free wall and tricuspid annulus motion, 79.8% for RV pressure overload, and 65.9% for RA and central venous pressure. A detailed description of these changes is shown in Table 2. Based on the proposed algorithm, at least one abnormal echocardiographic category was present in 264 (65.7%) patients at baseline, 146 (36.3%) at 6 days and 104 (25.9%) at 6 months.

Fig. 2
figure 2

Changes in echocardiographic parameters of right ventricular function across the 6-month follow-up: green color represents ‘normal’ and red ‘abnormal’ result at the end of follow-up. RA right atrial, RV right ventricular

Table 2 Changes of right ventricular function in the study population during follow-up, as assessed by echocardiography

The median value of NT-proBNP was 1448 (406.5–3417) pg/ml at baseline, being almost twice as high (1873 [533–4391] versus 956 [304.3–2468.5] pg/ml) in patients compared to those without RV dilatation; three times as high (2964 [860–5007] versus 1084 [310.3–2662.8] pg/ml) in patients with reduced compared to normal RV free wall and tricuspid annulus motion; more than two-fold increased (2036 [901–4360] versus 955 [303.5–2616] pg/ml) in patients with (versus those without) signs of RV pressure overload; and similarly, also twice as high (1862.5 [700–4537.3] versus 979 [333.8–2664.5] pg/ml) in patients with increased versus normal RA and central venous pressure. Overall, median NT-proBNP levels decreased sharply to 256.5 (94.6–799) pg/ml on day 6, and to 127 (61–280.5) pg/ml on day 180. Nevertheless, the median value of NT-proBNP continued to be relatively high, i.e. 519 (144.8–1463.5) pg/ml in patients with at least one abnormal echocardiographic finding at 6 days; it fell to 148.2 (57.8–405.7) pg/ml in patients with at least one abnormal ultrasound parameter at 6 months.

As shown in Table 3, body-mass index at baseline was associated with the presence of at least one abnormal echocardiographic category 6 days after the acute event, after adjusting for age, sex, prior VTE, history of cancer, history of chronic cardiopulmonary disease, hypotension, hypoxia and tachycardia. Table S3 shows the association of baseline parameters with the presence of at least one abnormal echocardiographic category at 6 months.

Table 3 Predictors of abnormal findings of right ventricular function in at least one echocardiographic category at 6 days

Discussion

The aim of the present analysis was to examine the temporal pattern of changes in RV function, i.e. complete or partial recovery versus persistence or deterioration of abnormal findings, in patients having suffered acute, intermediate-risk PE. We examined the patient population participating in a prospective multicentre multinational single-arm study. Our main results are the following: (i) RV dilatation was the most frequently (41%) reported abnormal echocardiographic finding at the time of the index PE; (ii) in almost two-thirds of the patients, RV function parameters as assessed by echocardiography had already recovered by day 6, and this improvement was maintained over the 6-month follow-up; however, almost one out of four patients had at least one abnormal echocardiographic finding at 6 months; and (iii) the median levels of NT-proBNP remained elevated in patients with at least one abnormal echocardiographic finding at 6 days but decreased and approached normal values at 6 months even among patients with persisting echocardiographic abnormalities.

Our results extend those of previous reports on the temporal changes of RV dimensions and function after acute PE. They may help to improve the level of evidence regarding long-term outcomes since our patient population was included in a prospective management study with standardised initial and chronic (over 6 months) treatment of PE. In an early small observational study dating back to the vitamin K antagonist era, it was reported that RV function normalised in the vast majority of PE patients within the first 5–13 days following treatment initiation [16]. However, another observational cohort study, focusing on 109 patients with submassive PE, demonstrated that an abnormal RV function (indicated by RV dilatation or RV hypokinesia) was still present in 25% of the patients at 6-month follow-up [17]. Furthermore, a substantial proportion (40%) of the patients who were included in the first PEITHO trial and underwent long-term follow-up had one or more indicators of pulmonary hypertension and/or RV dysfunction documented by echocardiography, with no differences between patients randomised to early systemic thrombolysis and those having received placebo [18].

The present study adds to our knowledge on the patients’ long-term course after acute intermediate-risk PE in view of contemporary treatment with a direct oral anticoagulant (in this case, a thrombin inhibitor) over the entire 6-month period, and it also proposes a standardised approach to categorising, analysing and reporting echocardiographic follow-up data. In this regard, it must be emphasised that our ‘algorithm’ was not developed as a new prognostic score in acute PE but rather as a guidance for following RV dysfunction with echocardiographic parameters used in everyday clinical practice. We believe that the parameter groups and the steps developed for our analysis may be useful for future studies with serial assessments of RV (dys)function, especially when multiple centres in different countries are involved. Since the assessment of RV function may be crucial for the resumption of daily activities during follow-up after acute PE [19], an echocardiographic algorithm might also facilitate everyday clinical practice as well as help to harmonise follow-up programs.

Although RV function is expected to recover early after acute PE in the majority of the patients, there are cases where RV recovers at a later stage or even fails to recover completely [19, 20]. This may be due to the presence of pre-existing chronic PE/CTEPH at the time of acute PE and a persistent RV dysfunction already at baseline [21]. The impact of incomplete RV recovery on long-term prognosis, i.e. its correlation with functional limitation, persistent symptoms, poor quality of life and chronic thromboembolic pulmonary disease, with or without pulmonary hypertension, remains to be established.

The observation that obesity was an independent determinant of persistently compromised RV function in our study is of interest but not totally unexpected. Previous studies have reported that obesity negatively affects the cardiovascular system [22] and is a risk factor for RV dysfunction and abnormal RV morphology [23], while also affecting LV size and contractility [24]. An increased body-mass index at baseline has been associated with worse imaging parameters reflecting RV function and higher levels of NT-proBNP, thus underlining the prognostic value of the body-mass index for the outcome of patients with acute PE [25], and justifying the recent call to focus on achieving/maintaining a healthy lifestyle in PE survivors [19, 26].

Strengths of the present study include: (i) the participation of patients from 42 centres across 9 European countries, covering a broad spectrum of clinical settings in which the follow-up took place; (ii) the prospective follow-up of RV function at 6 days and 6 months after the diagnosis of acute PE; (iii) the prospective categorisation of echocardiographic parameters to permit standardised assessment of RV function and its changes over time, and (iv) the assessment of both echocardiographic and biochemical parameters for the evaluation of RV (dys)function and (presumably right) heart failure at baseline and during follow-up. However, our analysis also has some limitations. Firstly, not all parameters related to RV function were available for all patients at all visits. In addition, an association of the kinetics of echocardiographic RV parameters with the primary clinical endpoint of the study, recurrence of symptomatic or fatal VTE at 6 months, could not be established due to the small absolute number of recurrent events [10]. Finally, the algorithm of echocardiographic parameters proposed in the present study needs to be validated in external cohorts and associated with the patients’ clinical long-term prognosis before it can be proposed for broader investigational and clinical use.

In conclusion, our results indicate that in the majority of patients with acute intermediate-risk PE who were switched early (on the third day) to a direct oral anticoagulant, echocardiographic parameters reflecting RV function normalised within 6 days and remain normal throughout the first 6 months. The levels of NT-proBNP also improved during 6-month follow-up. Almost one in four patients, however, still had at least one abnormal echocardiographic finding suggesting some degree of persisting RV dysfunction and possibly the need for continued follow-up over the long term.