Introduction

The ongoing coronavirus disease 2019 (COVID-19) pandemic has caused dramatic effects on society. On March 21, 2021, more than 122 million people have been infected and more than 2.7 million people have died of the disease worldwide [1]. Although approximately 80% of patients with COVID-19 have a favorable clinical course without hospitalization [2], approximately 20% of patients experiences severe respiratory disease [2]. A high incidence of thromboembolic events, including pulmonary embolism (PE), has been observed in COVID-19, which suggests that COVID-19 may induce intravascular coagulopathy [3,4,5,6]. PE may be a direct cause of death in patients with COVID-19, despite the use of antithrombotic prophylaxis [4, 6, 7]. Patients with COVID-19 who experience PE should be managed in a timely manner with therapeutic doses of anticoagulant therapy [8]. Computed tomography pulmonary angiography (CTPA) is the preferred imaging modality to detect PE [9]. To date, the frequency of PE in patients with COVID-19 is not completely clear. As such, it is still unclear which patients should undergo CTPA to detect PE. Unfortunately, clinical pretest probability scores, such as the Wells criteria [10], are unreliable to predict the occurrence of PE in patients with COVID-19 [11,12,13,14]. It has been suggested that assessment of D-dimer levels may help to improve risk stratification for PE [5, 15], but the exact value is also not completely clear. In order to overcome these gaps in knowledge, it was our purpose to investigate, in a meta-analysis, the frequency of PE in patients with COVID-19, and whether D-dimer assessment may be useful to select patients with COVID-19 for CTPA.

Materials and methods

This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guideline [16]. Institutional review board approval was not applicable.

Study retrieval and selection

MEDLINE and Embase were searched using the following search string: (Corona OR Coronavirus OR Covid-19 OR SARS-Cov-2 OR 2019nCoV OR Wuhan-virus) AND (Computed tomography OR Computerized tomography OR Computed tomographic OR CT OR CAT OR CTPA) AND (pulmonary embolism OR PE OR pulmonary thromboembolism OR PTE OR pulmonary thrombosis). Furthermore, the journal Radiology: Cardiothoracic Imaging was manually searched for relevant studies (articles published by this journal were not listed yet in MEDLINE and Embase). The search was updated until March 14, 2021. Bibliographies of studies which were included in our meta-analysis were screened for potentially suitable references.

Original studies which reported the frequency of PE on CTPA scans performed in at least 10 patients with COVID-19 (regardless of PE frequency) were eligible for inclusion. Review articles, abstracts, and studies involving fewer than 10 patients were excluded. Using these selection criteria, titles and abstracts of studies were reviewed. The full text versions of potentially relevant studies were then reviewed to determine whether studies could be included in our meta-analysis. Bibliographies of included studies were screened for other potential relevant studies. Two reviewers (R.M.K. and H.J.A.A.) independently performed the study selection. Any discrepancies were solved by consensus with a third reviewer (T.C.K.).

Study data extraction and study quality assessment

Main study characteristics (country of origin, patient inclusion period, number of patients, age and gender of patients, indication for CTPA, use of antithrombotic prophylaxis before CTPA, and CT interpreter(s)) were extracted for each included study. The proportions of patients with and without PE were extracted. If reported, data were extracted for patients with COVID-19 who presented at the emergency department (ED), for patients with COVID-19 who had been admitted to general wards, and for patients with COVID-19 who had been admitted to the intensive care unit (ICU). Furthermore, we extracted the association between severity of COVID-19 at chest CT and PE, if reported by the included studies. We also extracted the locations of PEs (i.e., main, lobar, segmental, and subsegmental pulmonary arteries) on a per-patient basis if reported by the included studies. In addition, for studies which reported D-dimer levels for patients with and without PE, we extracted the mean values and standard deviations (SDs). Corresponding authors of studies which did not report mean values and SDs were contacted and requested to provide these values. We also extracted sensitivity and specificity values for different D-dimer cutoff levels, if reported by the included studies.

Study quality aspects were assessed by two independent reviewers (R.M.K. and H.J.A.A.) using items from the Newcastle-Ottawa quality assessment scale [17] which were adapted to our meta-analysis (Table 1). Any discrepancies were solved by consensus with a third reviewer (T.C.K.).

Table 1 Criteria to evaluate the quality of included studies
Full size table

Statistical analyses

The frequencies of PE in patients with COVID-19 were determined for each included study and pooled with a random effects model. For studies which reported the frequency of PE in both patients with COVID-19 and those without COVID-19, differences were assessed by a chi-square test [18]. The frequencies of central (main and lobar) and peripheral (segmental and subsegmental) PEs were also pooled with a random effects model and the odds ratio (OR) of peripheral vs. central PEs was calculated. Differences in D-dimer levels between patients with COVID-19 vs. those without PE were assessed by calculating the standardized mean difference (SMD). The pooled SMD was estimated by a random effects model. Heterogeneity was tested by the I2 statistic [19]. If significant heterogeneity was present (defined as I2 > 40% [20]), subgroup analyses were performed to explore potential sources of heterogeneity. Predefined covariates were “indication for CTPA” (study reported that was CTPA only performed if PE was clinically suspected vs. study reported that CTPA was performed for triaging or on a routine basis, but not necessarily because of clinically suspected PE), “use of antithrombotic prophylaxis before CTPA” (100% vs. < 100% of included patients who used antithrombotic prophylaxis before CTPA), and “way of CTPA interpretation” (blinding vs. no blinding of CTPA interpreters to clinical information). Statistical analyses were performed using the Open Meta-Analyst software package [21], and MedCalc Statistical Software (MedCalc Software) [22]. p values < 0.05 were considered statistically significant for all analyses.

Results

Study retrieval and selection

The study selection process is summarized in Fig. 1. After screening titles and abstracts, 130 potentially relevant studies remained and were retrieved in full text. After reviewing the full text, 52 studies were excluded because there was no reporting of PE frequency data with respect to the number of CTPA scans performed in patients with COVID-19, 6 studies were excluded because these studies did not allow separate data extraction of patients with and without COVID-19, 2 studies were excluded because they comprised fewer than 10 patients with COVID-19, and 1 study was excluded because in this study PE was also determined based on clinical grounds rather than by CTPA only. Two additional references were found by screening bibliographies of remaining studies. Finally, 71 studies [11, 23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92], which comprised a total of 8086 patients with COVID-19 who underwent CTPA to evaluate for PE (median of 55 patients per study, range 10–1240), were included in our meta-analysis. Main study characteristics are displayed in Table 2.

Fig. 1
figure 1

Flow diagram of the study selection process

Full size image
Table 2 Main characteristics of the included studies
Full size table

Study quality

Details with regard to individual study quality are displayed in Supplemental Table 1. Eight studies (11.3%) had a prospective design, whereas 58 included studies (81.7%) had a retrospective design, whereas in 5 studies (7.0%) it was not reported whether the study design was prospective or retrospective. All but one of the included studies consecutively or randomly selected patients, or obviously comprised a representative series of patients. In 55 studies (77.5%), patient selection criteria for CTPA were reported, in 15 studies (21.1%), patient selection criteria for CTPA were not reported, whereas in 1 study (1.4%), patient selection criteria for CTPA were only reported for a subset of patients. CTPA interpreters were blinded to clinical information in 15 studies (21.1%), and unblinded in 2 studies (2.8%) whereas this was not clear (not reported) in the remaining 54 studies (76.1%).

Frequency of PE in patients with COVID-19

Pooled frequency of PE in all included patients with COVID-19 was 32.1% (95% confidence interval [CI]: 28.5–35.9%). Pooled frequency of PE was lowest in patients who presented at the ED (17.9% [95% CI: 12.0–23.8%]) (Fig. 2), followed by patients who had been admitted to general wards (23.9% [95% CI: 15.2–32.7%] (Fig. 3). In patients with COVID-19 who had been admitted to the ICU, pooled frequency of PE was highest (48.6% [95% CI: 41.0–56.1%]) (Fig. 4).

Fig. 2
figure 2

Frequency of PE in patients with COVID-19 who presented at the ED

Full size image
Fig. 3
figure 3

Frequency of PE in patients with COVID-19 who had been admitted to general wards

Full size image
Fig. 4
figure 4

Frequency of PE in patients with COVID-19 who had been admitted to the ICU

Full size image

Significant heterogeneity was present across the included studies (I2 ≥ 80%). No potential sources of heterogeneity were identified (I2 > 85%) for subgroups according to “indication for CTPA,” “use of antithrombotic prophylaxis before CTPA,” and “way of CTPA interpretation.” In two studies which routinely performed CTPA at the ED (regardless of clinical suspicion of PE), PE frequencies in COVID-19 patients were 2.1% (1/48) and 5.7% (4/70), respectively [29, 50]. In two other studies which routinely performed CTPA at the ICU (regardless of clinical suspicion of PE), PE frequencies in COVID-19 patients were 47.2 % (34/72) and 60.0% (12/20), respectively studies [68, 92]. Six studies reported a significant association between severity of COVID-19 at chest CT and PE, whereas 13 studies did not find a significant association (Supplemental Table 2).

PE location

PE was more commonly located in peripheral than in main pulmonary arteries (pooled frequency of 65.3% [95% CI: 60.0–70.1%] vs. 32.9% [95% CI: 26.7–39.0%]; OR = 3.540 [95% CI: 2.308–5.431%]).

PE in patients with COVID-19 and association with D-dimer levels

Patients with COVID-19 and PE had significantly higher D-dimer levels than patients with COVID-19 and no PE (pooled SMD of 1.096 [95% CI, 0.844–1.349]; I2 = 89%) (Fig. 5). Sensitivity and specificity values for different D-dimer cutoff levels are displayed in Table 3. D-dimer cutoff levels which have been used to identify patients with PE varied between 1000 and 4800 μg/L. All studies listed in Table 3 used the conventional D-dimer score. Only one study also used age-adjusted D-dimer cutoffs [93], yielding a sensitivity of 94% and a specificity of 35% [33].

Fig. 5
figure 5

Association between D-dimer levels and PE in patients with COVID-19

Full size image
Table 3 Sensitivity and specificity values for different D-dimer cutoff levels to diagnose PE
Full size table

Discussion

This meta-analysis showed that the frequency of PE in COVID-19 was highest in patients who were in the ICU (pooled frequency of 48.6%), followed by patients who were in general wards (pooled frequency of 23.9%), and by patients who presented at the ED (pooled frequency of 17.9%). PE was more commonly located in peripheral than in main pulmonary arteries (pooled frequency of 65.3% vs. 32.9%). Patients with PE had significantly higher D-dimer levels than patients without PE.

Fifty-eight of the 71 included studies (81.7%) had a retrospective design. However, there was no evidence of selection bias, as all but one of the studies included a consecutive, randomly selected, or obviously representative series of patients. Selection criteria for CTPA were reported in the majority of included studies (77.5%), which benefits the generalizability of study results. In only 21.1% of included studies, it was reported that CTPA interpreters were blinded to clinical information. Non-blinding could have biased the results to either overcalling or undercalling PE frequency on CTPA.

The findings of our meta-analysis suggest that the frequency of PE in patients with COVID-19 increases with increasing disease severity (ICU > general wards and ED). This is supported by six studies which reported a significant association between severity of lung parenchymal abnormalities at CT and PE [32, 40, 44, 45, 71, 74]. However, such an association was not demonstrated in 13 other studies [24, 30, 31, 37, 42, 43, 49, 50, 52, 60, 62, 78]. Therefore, there are probably other COVID-19- and host-related factors that are associated with the occurrence of PE. Further studies are required to improve our understanding of the pathophysiology of PE in COVID-19. Furthermore, the observed frequency of PE depends on the selection criteria for CTPA. In the far majority of included studies, it was reported that CTPA was performed because of clinically suspected PE. In only two studies, CTPA was routinely performed at the ED (regardless of clinical signs of possible PE), with relatively low yields of only 2.1% and 5.7% [29, 50]. In two other studies which routinely performed CTPA in COVID-19 patients at the ICU (regardless of clinical signs of possible PE), PE frequencies were high: 47.2% and 60.0%, respectively [68, 92]. These findings in unselected samples of patients confirm that frequency of PE is higher in ICU patients compared to patients who present at the ED.

Our findings contrast those in patients from the general population without COVID-19, where PE has been reported to occur in main pulmonary arteries as frequent as or more frequently than in peripheral arteries [94,95,96]. Therefore, the underlying pathomechanisms may be different. The relatively high frequency of peripheral PE suggest that local thrombosis may play a more important role in the development of PE (or pulmonary artery thrombosis) in COVID-19 [37, 55, 97, 98] rather than the classic thromboembolism originating from the leg or pelvic veins in patients without COVID-19 [99]. This hypothesis is supported by in vivo chest CT studies, where vascular thickening, a potential sign of vascular inflammation, endothelial damage, and microthrombosis [55], is observed in most symptomatic patients with COVID-19 [100]. Pathological studies in patients with COVID-19 also confirm the local immunothrombosis hypothesis [97, 98]. Accordingly, the term MicroCLOTS (microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome) has been coined as a new name for severe pulmonary COVID-19, in which alveolar viral damage is followed by an inflammatory reaction and by microthrombosis [97, 101,102,103]. It may become obsolete to call this pathophysiological disorder PE.

Subgroup analysis did not indicate that the use of antithrombotic prophylaxis was associated with a lower frequency of PE in patients with COVID-19. This implies that physicians should remain alert for the occurrence of PE even in patients who receive antithrombotic prophylaxis. D-dimer levels were found to be significantly higher in patients with PE (pooled SMD of 1.096), which indicates that a rise in D-dimer levels is not only a marker of pneumonia severity but is also associated with a higher risk of PE. Therefore, D-dimer assessment may help to decide which patients with COVID-19 should undergo CTPA to detect PE. However, there is no uniformly accepted D-dimer threshold to discriminate COVID-19 patients with and without PE. Twelve studies used different D-dimer cutoff levels (varying between 1000 and 4800 μg/L), yielding sensitivity and specificity values which varied between 63–100% and 23–84%, respectively [11, 24, 33, 52, 58, 60, 71, 74, 78, 82, 86, 89]. These D-dimer cutoff levels were at least twice as high compared to the conventional D-dimer cutoff level of 500 μg/L, which is usually employed in the general population as a screening test for venous thromboembolism [104, 105]. In non-COVID-19 patients aged 50 or more, the application of age-adjusted D-dimer cutoffs has shown to increase specificity without modifying sensitivity [106]. Only one of the studies included in our meta-analysis also used age-adjusted D-dimer cutoffs, yielding high sensitivity (94%) but poor specificity (35%) [33]. More research is needed to investigate whether the use of age-adjusted D-dimer cutoffs can improve the clinical utility of D-dimer testing in patients with COVID-19.

Our study has some limitations. First, in the far majority of included studies, CTPA was only performed in case of clinically suspected PE. Therefore, the true prevalence of PE in patients with COVID-19 remains to be elucidated. Second, due to incomplete and unstandardized reporting, we could not adjust the frequency of PE for well-known risk factors for PE (such as cancer, history of previous venous thromboembolism, duration of hospitalization, obesity, and cardiovascular disease [107]) and type and dosage of antithrombotic prophylaxis. Third, there was a great deal of heterogeneity in the patient population and the indication for CTPA in each included study. Although we attempted to group the studies into ED, general wards, and ICU patients, this delineation may be problematic due to the unpredictable course of COVID-19 and the fact that a patient discharged from the ED could become an ICU ARDS patient within a matter of a week. Furthermore, statistical heterogeneity still remained in each of these groups. Fourth, PE was determined by CTPA, which has a good but not perfect sensitivity in PE detection [9]. Although they may be clinically less relevant [108], smaller subsegmental PEs may have been missed by CTPA. This could have resulted in an underestimation of PE frequency.

In conclusion, the frequency of PE in patients with COVID-19 is highest in the ICU, followed by general wards and the ED. PE in COVID-19 is more commonly located in peripheral than in central pulmonary arteries, which suggests local thrombosis to play a major role. D-dimer assessment may help to select patients with COVID-19 for CTPA, using D-dimer cutoff levels of at least 1000 μg/L.