Incidence of acute pulmonary embolism among patients hospitalized with COVID-19: a systematic review and meta-analysis [version 2; peer review: 1 not approved]

Background: Coronavirus disease 2019 (COVID-19) is a global pandemic, which is associated with venous thromboembolism and pulmonary embolism (PE). This study aimed to estimate the pooled incidence of PE among patients hospitalized with COVID-19 within the published literature. Methods: This systematic review and meta-analysis was performed according to PRISMA guidelines. An electronic search using MEDLINE /PubMed, ScienceDirect, Cochrane, and OpenGray databases was conducted May 19th, 2020. Eligible studies included sufficient data to calculate the incidence of PE diagnosed during hospitalization in patients with COVID-19. Case reports were excluded. Quality was assessed using the Newcastle-Ottawa scale (observational cohort and Open Peer Review

case-control), AXIS tool (cross-sectional), and quality assessment tool (case series). Demographics and PE incidence data were extracted from the included studies and analyzed with R language. The pooled incidence of PE in patients hospitalized with COVID-19 was calculated. Results: The database search identified 128 records. Ten observational studies were eligible and were included in the metaanalysis with a total of 1722 patients (mean age= 63. 36). .The incidence of PE was noted to be higher in males. The D-dimer levels were specified between PE group and non-PE group in only three studies,

Introduction
In December 2019, pneumonia of unknown cause was detected in Wuhan, China 1 . The causative agent was identified and named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) 2 . On March 11th, 2020, the World Health Organization characterized the coronavirus disease of 2019 (COVID-19) as a pandemic 1,3 , resulting in 53,507,282 and 1,305,164 COVID-19-related cases and deaths, respectively, as of November 15 th , 2020 4 . While COVID-19 is primarily a pulmonary disease, there are multiple other pathologic manifestations and complications, including pulmonary embolism (PE) 5 .
The relationship between COVID-19 and thromboembolism is becoming established in the literature 5 . Thromboembolism has been previously associated with zoonotic coronaviruses 6 and may be attributed to several factors including; a hypercoagulable state associated with severe infection or inflammation 7 , COVID-19 associated hemostatic abnormalities 8,9 , recumbence 7,10-12 , and possible drug interactions between investigational COVID-19 therapies (Lopinavir/ritonavir) and antithrombotics 5 .
This systematic review and meta-analysis analyzed and estimated the pooled PE incidence from published literature of patients hospitalized with COVID-19 who developed PE.

Methods
This meta-analysis was performed following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline 13,14 . Institutional review board approval was not required for this study.
Duplicate citations and older versions of the same study population were removed. All included study bibliographies were screened for additional articles investigating PE in patients hospitalized with COVID-19. All eligible studies underwent full-text screening.

Eligibility criteria
Studies reporting data sufficient to estimate PE incidence of patients hospitalized with COVID-19 were included in the analysis. Only English and published studies were included. Studies with insufficient data, case reports, editorials, proposals, and abstracts without full-text were excluded. The title and abstract of retrieved articles were screened by two independent reviewers for potential inclusion. Any discrepancy between the reviewers was resolved by consensus with two additional reviewers.
Quality appraisal and data extraction Appraisal of individual study quality was performed by two independent reviewers using the Newcastle-Ottawa scale, AXIS tool, and quality assessment tool for observational cohort and case-control, cross-sectional, and case series, respectively 15-17 . Articles were deemed of good quality if they scored 50% or more using an arbitrary cut-off.
Information was sought from each included study are: characteristics of participants (including age, gender, BMI, D dimer of patient with COVID-19,, and total patients in the study), summary about included studies (first author, country, year of publication, study period, study design, method of diagnosing COVID-19) and the main data for meta-analysis (total patients in each study i.e., sample size and number of patients developing PE from the total). The data was subsequently extracted by five independent reviewers utilizing Microsoft Excel ® 2016 (Microsoft Corporation, Redmond, WA).

Analysis
Due to skewed proportions, the random effect model was used to pool the individual estimates through LOGIT transformation instead of double arcsine transformation to avoid a paradoxical effect upon back-transformation 18,19 .
Statistical analysis was performed using R language v.4 20 , using "meta" and "metafor" packages 21-23 . Random effects models were utilized to accommodate for the heterogeneity in the reported pooled incidences. Statistical heterogeneity was estimated using I 2 statistics and further assessed using subgroup analysis, meta-regression, influence analysis, and Gosh analysis. Publication bias was evaluated by both the Egger test and funnel plot visual analysis.

Study characteristics
The search yielded 81 and 47 records in both Medline/PubMed and ScienceDirect, respectively. No records were identified in OpenGrey and Cochrane databases. After eliminating duplicate data, 125 studies were included for title and abstract screening, of which 76 were excluded due to insufficient data. Full-text screening of the remaining 49 studies excluded 38 records with an agreement kappa of 0.813. Of the remaining studies, one scored <50% on the quality assessment tool, leaving ten studies for pooled analysis. Details of the selection process are summarized in (Figure 1). Nine studies were from Europe and

Amendments from Version 1
We shortened the method section of the abstract and widened the result section of it, also we add the missing data of the flowchart. we remove the fixed effect model sentences from Figure 1 and Figure 2. one study was from North America (Table 1). Studies were divided into descriptive and analytic categories for subgroup analysis ( Table 1). The reported PE incidence among patients hospitalized with COVID-19 in all included studies ranged from 3-35% (Table 2).
Patients had a mean age of 63 years. The incidence of PE was noted to be higher in males ( Table 3). The D-dimer levels were specified between PE group and non-PE group in only three studies, while the remaining either reported it improperly or had missing data (Table 3).
Though anatomical distribution of PE was mentioned in most studies (except Lorant et al. and Grillet et al.) 29,30 , the clinical classification of PE into massive and sub-massive was not mentioned.

Meta-regression
When categorizing the included studies into analytic or descriptive, the meta-regression model showed that the study design was significantly associated with the difference in PE incidence (p<0.05) ( Figure 5).

Discussion
PE is the most common thromboembolic complication occurring in patients with COVID-19 24,30,32 . This systematic review     and meta-analysis estimated a pooled PE incidence among patients hospitalized with COVID-19 at 17% (95% CI: 10-26%) (Figure 2) 24-33 . This pooled PE incidence is higher than the PE incidence of the general population, and most importantly, higher than that of hospitalized patients with other medical conditions 34-40 .
This reported PE incidence could represent an over or under estimation of the true incidence. The increasing knowledge of higher incidences of venous thromboembolism in patients with COVID-19 may have led to a selection bias due to lower threshold of CT pulmonary angiography (CTPA) utilization 41 . An underestimation is known to occur in the general population for the diagnosis of PE related to the majority of massive PE being diagnosed on post-mortem examination. Similarly, patients with COVID-19 could be assumed to have a lower reported incidence of PE diagnosis due to; limited CT scan availability, patients instability, concern over hospital exposure to others related to transportation, early hospital mortality and death outside of healthcare 5,[42][43][44] .
Most of the patients included in this meta-analysis were males and aged sixty years and above. Not only was PE higher in elderly and males patients hospitalized with COVID-19, but also other COVID-19 related complications and death were seen more in this population 45,46 . This observation is consistent with the higher PE incidence across the elderly in the general population as well 47 . Therefore, hospitalized male and elderly  patients with COVID-19 could be at a higher risk of developing PE compared to their COVID-19-afflicted female and young counterparts, respectively.
To account for the variability of CTPA timing effect on PE incidence, a study assigned each patient with COVID-19 two time points in the CTPA and demonstrated a higher incidence at the later time point 27 . Such an observation may help in interpreting the variation in PE incidence across different studies with similar population characteristics and may also highlight underestimation of PE incidence in population with single time CTPA.
PE mortality was reported in only 3 out of the 10 studies using different formats 24,25,27,28,33 . Therefore, the estimated pooled mortality could not be calculated.

Strength, limitations, and recommendations
This study was designed and executed in accordance with PRISMA guidelines. The heterogonous severity of the included COVID-19 cohort included the complete spectrum of hospitalized patients with COVID-19. Therefore, this estimated incidence could apply to any hospitalized patients with COVID-19, regardless of disease acuity. However, these findings need to be interpreted in the context of some limitations. First, neither articles in non-indexed journals nor non-published papers were searched, which may have introduced some publication bias. Second, the inclusion of studies published only in English literature may have led to language bias. Third, the presence of a subpopulation in whom PE was suspected and CTPA could not be performed for various reasons, such as allergy to contrast material, and could result in skewed incidence. Fourth, both clinical indications for CTPA and PE classification or risk stratification were not specified in the different studies. Sixth, due to variability of the way of reporting specific data on some variables (d-dimer, mortality data), it was difficult to make meaningful predictions. Lastly, some relevant clinical and para-clinical variables were not mentioned.
Further research is warranted to accurately characterize the patients hospitalized with COVID-19 who develop PE in terms of; risk factors profile, PE diagnostic indications, CTPA timing, PE prophylaxis and management, and PE pathogenesis.

Conclusion
This systematic review and meta-analysis reported a pooled PE incidence among patients hospitalized with COVID-19 at 17%, suggesting that almost in every five hospitalized patients with COVID-19, one may develop PE. This represents around a 243-fold increase in incidence when compared to the general population. Healthcare professionals should be aware of this observed increase in risk of PE incidence among patients hospitalized with COVID-19. Development of accurate and precise risk stratification scores and point of care biomarkers to guide prophylactic and therapeutic strategies in this vulnerable population are warranted.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required. This project contains the following extended data: -Code for publishing plots (.R file).

Extended data
-Code for meta-analysis (.R file). It is more appropriate to use MOOSE guidelines for a systematic review and meta-analysis of articles discussing the incidence.

Reporting guidelines
Current research is not registered on PROSPERO.
The authors say, "A systematic literature search was performed on May 19 th , 2020". This paper is a bit older. It should be updated and collect more new data.
The data analysis is a bit simple and the significance of the p value was not defined.
The research flow chart is not concise enough. What is the meaning of "36 records with no sufficient"?