Meta‐analysis of chest CT features of patients with COVID‐19 pneumonia

Abstract The objective of this paper is to perform a meta‐analysis regarding the chest computed tomography (CT) manifestations of coronavirus disease‐2019 (COVID‐19) pneumonia patients. PubMed, Embase, and Cochrane Library databases were searched from 1 December 2019 to 1 May 2020 using the keywords of “COVID‐19 virus,” “the 2019 novel coronavirus,” “novel coronavirus,” and “COVID‐19.” Studies that evaluated the CT manifestations of common and severe COVID‐19 pneumonia were included. Among the 9736 searched results, 15 articles describing 1453 common patients and 697 severe patients met the inclusion criteria. Based on the CT images, the common patients were less frequent to exhibit consolidation (odds ratio [OR] = 0.31), pleural effusion (OR = 0.19), lymphadenopathy (OR = 0.17), crazy‐paving pattern (OR = 0.22), interlobular septal thickening (OR = 0.27), reticulation (OR = 0.20), traction bronchiectasis (OR = 0.40) with over two lobes involved (OR = 0.07) and central distribution (OR = 0.18) while more frequent to bear unilateral pneumonia (OR = 4.65) involving one lobe (OR = 13.84) or two lobes (OR = 6.95) when compared with severe patients. Other CT features including ground‐glass opacities (P = .404), air bronchogram (P = .070), nodule (P = .093), bronchial wall thickening (P = .15), subpleural band (P = .983), vascular enlargement (P = .207), and peripheral distribution (P = .668) did not have a significant association with the severity of the disease. No publication bias among the selected studies was suggested (Harbord's tests, P > .05 for all.) We obtained reliable estimates of the chest CT manifestations of COVID‐19 pneumonia patients, which might provide an important clue for the diagnosis and classification of COVID‐19 pneumonia.


| Data extraction
We reviewed the titles, abstracts, and full texts of manuscripts by duplicate removal based on the above-mentioned selection criteria.
Abstracts of identified articles were separately reviewed by two readers. After we confirmed the inclusion of associated documents, we independently extracted following variables, including the name of the first author, publication year, age of patients, number of patients, and study area. All included literatures were evaluated using the Newcastle-Ottawa Scale. 4 The highest quality of the literature is 10 scores and the lowest is 0 score. Data extraction and quality assessment were carried out independently by two reviewers. In case of disagreement, consensus was reached by discussing with a third reviewer.

| Statistical analysis
All the statistical analyses were carried out using Stata statistical software version 12.0. The proportions of various CT features in each group were analyzed as follows: original data were transformed by double arcsine method in Stata at first and the final conclusions were drawn using restoring formula (P = (sin(tp/2)) 2 ). The association between the CT features and the severity of COVID-19 pneumonia was assessed in the form of odds ratio (OR) at a 95% confidence interval (95% CI). Heterogeneity among each study was evaluated using Cochran's Q test and Inconsistency index (I 2 ) test. 5 I 2 > 50% indicates the apparent heterogeneity between the studies and the randomeffects model (Der Simonian and Laird method) would be adopted.
Otherwise, the fixed-effect model (Mantel-Haenszel model) would be used. Publication bias was assessed for CT characteristics that included more than 10 studies using funnel plots and Harbord's tests.
Deviation from the funnel-shaped distribution of eligible research works suggested the presence of publication bias.

| Inclusion of studies
From the databases mentioned above, we retrieved 9736 articles.
After removing 1435 duplicated articles, 8301 articles remained. After reading the titles and abstracts, 8022 papers were excluded. After the reading the full text, we kept 15 descriptive studies including 2451 COVID-19 pneumonia patients in this meta-analysis. [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20] The entire process is shown in Figure 1. All the included studies were retrospective studies. The primary characteristics of the literature are F I G U R E 1 Summary of article selection process 242 | exhibited in Table 1. Generally speaking, these articles were considered to be of good quality. All the 15 articles were over 5 scores. (95% CI, 0.00, 0.07), and lymphadenopathy 0.01 (95% CI, 0.00, 0.03) were relatively rare in the common group. All the data above are shown in Table 2.

|
As Table 3 shows, among severe patients, the predominant CT fea-
They were less likely to have abnormalities with over two lobes involved (OR = 0.07; 95% CI = 0.03-0.17; P = .000) (Figure 3).  Common patients were more frequent to have unilateral pneumonia:

| DISCUSSION
Coronaviridae (CoVs) are well-known single-stranded RNA viruses that are ubiquitous in many mammals including humans. 21 According to the antigenic criteria and phylogenetic analyses, they are categorized into three groups: alpha-CoVs, beta-CoVs, and gamma-CoVs. 22

| 245
In the present meta-analysis, we found that GGOs, vascular enlargement, interlobular septal thickening, and subpleural bands were the most common findings in either common or severe patients. GGOs or consolidations in peribronchovascular or subpleural distribution. 32 Compared with the immunocompromised population, small airway abnormalities such as airway thickening and dilatation, centrilobular nodules, and tree-in-bud sign are rare in immunocompetent patients. 32 What is more, some patterns including lymphadenopathy and pleural effusions are usually absent in H1N1 pneumonia. 32 For H7N9 pneumonia, the most common findings on CT are GGOs. 33 Diffused consolidations, air bronchograms, and interlobular septal thickening are the second most common imaging abnormalities. 33 Besides, H7N9 pneumonia F I G U R E 3 Forest plots of studies on association between lesion locations of common and severe patients ZHENG ET AL.
| 247 usually progresses rapidly and the right lower lobe is easier to be involved. 33 The most common radiographic abnormalities in adenovirus pneumonia are diffuse bilateral bronchopneumonia and lobar atelectasis. 34 Thickened interlobular septa and diffuse GGOs are infrequent in adenovirus pneumonia. 34 Different from COVID-19 pneumonia, adenovirus is much easier to infect pediatric patients. Right upper lobe atelectasis is common in infants, while in older children, atelectasis usually occurs at left lower lobe. 34 In the respiratory syncytial virus (RSV) infected patients, CT usually manifests as the pattern of nodules, tree-in-bud opacity, and bronchial wall thickening. 35 Compared with other viral pneumonia, consolidation and GGOs are rarely observed in RSV-infected pneumonia. 35 Similar with adenovirus, infants and immunocompromised adults are more susceptible to RSV-infected pneumonia. SARS-CoV, MERS-CoV, and COVID-19 virus all belong to Coronaviridae and they share a lot of similarities in CT manifestation. GGOs and consolidations that mainly distribute at the peripheral lower lung zone are also the predominant abnormalities on CT scanning of patients with SARS and MERS. [36][37][38] Interlobular septal thickening and intralobular lines are common as well.
What is more, opposite to patients with COVID-19, patients with SARS and MERS manifest unifocal involvement more often than multifocal involvement on chest CT. 39 In the early stage, the lesions mainly locate under the pleura, with the progression of illness, lesions become diffuse.
After recovery, the fibrotic changes may be irreversible. Some patterns such as mediastinal lymph nodes and substantial effusions are irregular. 37 Even though there are some traceable differences on chest CT between these viral pneumonias, it is still hard work to distinguish COVID-19 from other vial pneumonia. Real-time polymerase chain reaction is needed for a definitive diagnosis.
To the best of our knowledge, this article is the first to systematically assess the chest CT manifestations in different severity of COVID-19 pneumonia. The analysis is rigorous and the conclusions are convincing. This study also has limitations. First, all the studies are retrospective studies and significant heterogeneity are observed.
Second, some studies with small samples were also included in the analysis and the strength of the study may also be limited. Third, all the patients included are Chinese and the conclusions may be less representative.
In conclusion, our results indicate that vascular enlargement and GGOs are common chest CT findings in COVID-19 pneumonia.
Severe patients are more likely to have CT abnormalities with traction bronchiectasis, interlobular septal thickening, consolidation, crazy-paving pattern, reticulation, pleural effusion, and lymphadenopathy. All five lobes tend to be affected. However, because of the limitations mentioned above, studies with larger sample size and more rigorous design should be carried out.