Typical chest CT features can determine the severity of Coronavirus Disease 2019 (COVID-19): a systematic review and meta-analysis of observational studies

Nahid Hashemimadani Endocrine Research Center, Institute of Endocrinology and Metabolism, Iran University of Medical Sciences (IUMS), Tehran, Iran. https://orcid.org/0000-0002-6134-9906 Zahra Emami Endocrine Research Center, Institute of Endocrinology and Metabolism, Iran University of Medical Sciences (IUMS), Tehran, Iran. https://orcid.org/0000-0001-6011-1915 Leila Janani Department of Biostatistics, School of Public Health, Iran University of Medical Sciences (IUMS), Tehran, Iran https://orcid.org/00000002-3370-9310 Mohammad E. Khamseh (  Khamseh.m@iums.ac.ir ) Endocrine Research Center, Institute of Endocrinology and Metabolism, Iran University of Medical Sciences (IUMS), Tehran, Iran. https://orcid.org/0000-0003-4313-8440


Introduction
Corona Virus Disease 2019 , caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), has been spreading throughout the world threatening the global health condition. COVID-19 infection causes a spectrum of respiratory illness ranges from a mild pneumonia to the critically ill and fatal cases of Acute Respiratory Distress Syndrome (ARDS) (1)(2)(3). The timely identi cation and successful treatment of severe and critical cases of COVID-19 are essential to reduce the complications and mortality of the disease. Many studies aimed to nd the disease severity-related factors in the clinical practice resulting in identi cation of some clinical, laboratory, and imaging factors as the predictors of disease severity (3,4).
Chest CT scan is a frequently used imaging tool for the early detection and monitoring of patients affected by COVID-19 due to the facts that it is relatively easy to perform and can produce fast diagnosis. Typical CT features of con rmed COVID-19 pneumonia have been shown to be helpful in evaluation of the severity and extent of disease (5,6). Multilobular in ltration, accounted for the highest score in MuLBSTA system (multilobular in ltration, hypo-lymphocytosis, bacterial co-infection, smoking history, hypertension, and age) for predicting the mortality in patients with viral pneumonia (7), has been more frequently reported in patients with severe and critical types of COVID-19 infection compared to the common types (3,5,6). The detailed imaging features in various clinical types of COVID-19 infection have been investigated in several studies indicating signi cant differences between common and severe cases (3,6,8).
Exploring the chest CT features in the spectrum of clinical presentation of the COVID-19 infection can help us to predict outcome and support clinical decision making. Although several studies indicated some CT features are associated with severe types of COVID-19 infection, as far as we know, there is no meta-analysis of these ndings. Thus, we aimed to quantitatively summarize results from published studies to date to identify the CT features which determine the severity of COVID-19 infection.

Search strategy and selection criteria
We searched for studies of any design and in any setting that included patients with COVID-19 infection and reported disease severity as well as chest CT characteristics of the patients. We initially did not limit our study by language.
The aim of this systematic review was to quantitatively assess the association of clinical severity of COVID-19 infection with the chest CT characteristics. The clinical severity of the disease de ned as common type, which included all patients with fever, cough, or other respiratory symptoms with or without CT manifestation of pneumonia, and severe type, which included the patients with any of the following condition:1) respiratory rate ≥30 breaths per minute; 2) nger of oxygen saturation ≤ 93% in a resting state; 3) arterial oxygen tension (PaO2)/inspiratory oxygen fraction (FiO2) ≤ 300 mmHg (1mmHg=0.133kPa); 4) respiratory failure and mechanical ventilation requirement; 5) shock; 6) patients with other organ failure needed intensive care unit (ICU) monitoring treatment (9).
We searched PubMed, Embase, Scopus, web of science databases (WOS), Cochrane library, and Google scholar from January 1, 2020 up to May 19, 2020 applying the following search terms: Coronavirus OR covid-19 OR 2019-CoV OR CoV OR severe acute respiratory syndrome coronavirus 2 OR SARS-CoV-2 AND CT scan OR Computed tomography OR CT image* OR CT nd* OR CT manifestation* OR CT Feature* OR Chest CT AND Severity OR Refractory OR Critical OR Emergency OR ICU admission.

Data extraction
We screened titles and abstracts, reviewed full texts, extracted data, and assessed risk of bias by two authors independently. We resolved disagreements by consensus. We included the studies met the following eligibility criteria: 1) the publications were original articles with full text; 2) the study population aged above 15 years with laboratory-con rmed COVID-19 infection; 3) the studies reported information on clinical severity of the disease; 4) at least one of the chest CT features of COVID-19 has been reported. Studies were excluded if 1) the full text was not in English; 2) the studied populations were children or pregnant women; 3) the study was case report or clinical trial.
We extracted data for study identi er, study sample size, mean or median age of the participants, gender distribution, disease severity classi cation, results of chest CT imaging features, the number of patients with each corresponding imaging features considering disease severity classi cation.

Data analysis
All Statistical analyses were performed using Stata version 14 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP.). We employed "metaprop" command to calculate the pooled prevalence estimates of chest CT features with 95% con dence. The associations between chest CT features and severity of disease, were expressed as odds ratios (ORs) and 95% con dence intervals (CIs). Heterogeneity among the primary studies was evaluated by the forest plots, Cochran's Q statistic, and I 2 statistic. A random-effects model was used if heterogeneity was high (I 2 > 50%); otherwise, a xed-effects model was applied. Also, the publication bias was formally tested with Egger's regression asymmetry tests to determine the asymmetry of the funnel plots, where p <0.10 was considered as evidence of bias.

Literature retrieval
A total of 1587 related records have been identi ed in the initial retrieval. Upon removal of duplication, 1158 records have been screened against the title and abstract. Full text of 60 studies has been assessed for eligibility. We identi ed 18 studies for our systematic review. Studies were all observational in nature. These 18 studies ful lled criteria for including in the meta-analysis ( Figure 1, Table 1).   Table 1). Risk of bias was generally low after considering the observational designs (Supplementary Table 1).

Pattern of lesions
Twelve studies were included to explore the association between the presence of GGO in CT images and clinical severity of the disease. Compared with the common type, the pooled OR for the presence of GGO in CT images of patients with severe disease was 1.37 (95% CI 1.08-1.73, I 2 = 0.0%, P for heterogeneity= 0.486) (Figure 2). Eight studies were included to explore the association between the presence of consolidation in CT images and clinical severity of the disease. The pooled OR for the presence of consolidation in CT images of patients with severe disease, compared to the common type, was 2.33 (95% CI 0.85-6.36, I 2 = 74.5%, P for heterogeneity= 0.000) (Figure 3). Three studies were included to explore the association between the presence of linear opacity in CT images and clinical severity of the disease.
Compared with the common type, the pooled OR for the presence of linear opacity in CT images of patients with severe disease was 3.27 (95% CI 1.10-9.70, I 2 = 73.8%, P for heterogeneity= 0.022) (Figure 4). Three studies were included to explore the association between the presence of bronchial wall thickening in CT images and clinical severity of the disease. The pooled OR for the presence of bronchial wall thickening in CT images of patients with severe disease, compared to the common type, was 11.64 (95% CI 1.81-74.66, I 2 = 81.8%, P for heterogeneity= 0.004) ( Figure 5). Substantial between-study heterogeneity was shown in results for three patterns of lesion namely, consolidation, linear opacity, and bronchial wall thickening.

Lesion distribution
Bilateral lung involvement has been explored in thirteen included studies. Compared with the common type, the polled OR for the presence of bilateral lung involvement in CT images of patients with severe disease was 3.44 (95% CI 1.74-6.79, I 2 = 63.9%, P for heterogeneity= 0.001) ( Figure 6). There was no evidence for substantial between study heterogeneity in results for bilateral lung involvement.

Assessment of publication bias
Begg's and Egger's regression tests provided no evidence of substantial publication bias except for GGO (P <0.1 for both tests)

Discussion
Our systematic review and meta-analysis of 18 studies involving 3323 patients provides comprehensive information on the chest CT features associated with the severe cases of COVID-19 infection. Bronchial wall thickening was more likely to be associated with severe cases of COVID-19 infection, followed by linear opacity, and GGO. However, there was no signi cant association between the presence of consolidation and severity of clinical presentation. Considering the lesion distribution, as expected, bilateral lung involvement was more frequently associated with severe clinical presentation.

GGO
GGO indicates an area of hazy increased lung opacity, less opaque than consolidation, through which vessels and bronchial structures may still be obscured. Although GGO is the most common CT feature of COVID-19 pneumonia, most studies did nd no statistically signi cant difference in the presence of GGO between severe and common cases of COVID-19. Nevertheless, the meta-analysis of these studies indicated a 37% increase in the risk of the presence of GGO in severe cases compared to the common type ( Figure 2). Since the GGO has been known as the typical early CT feature of COVID-19 pneumonia, it is advised that the small lesions, and especially new lesions, that contained an area of GGO requires follow-up to eliminate the possibility of COVID-19 pneumonia in high-risk subjects (21).

Consolidation
Consolidation is often a middle-to-late stage feature in pulmonary infection. Although some studies indicated the association of consolidation with the severity of COVID-19 infection, others failed to show this association. The pooled analysis of these studies did not demonstrate any signi cant association between the presence of consolidation and the severity of clinical manifestations ( Figure 3). Consolidated tissue is more radio-opaque than normally aerated lung parenchyma, indicating the alveoli are completely lled by in ammatory exudation, hemorrhage, or pus. This feature could re ect the severity of lung involvement. The reason why this meta-analysis could not nd a signi cant association between the presence of consolidation in the chest CT and severity of clinical presentation might be explained by the fact that in the included studies the chest CT scan had been performed at the early stage of the infection while consolidation has been demonstrated to be noted on chest CT during disease progression (3). Thus, the time point from the onset of disease at which the chest CT is performed has a critical role in demonstrating the consolidation.

Linear opacity
The linear opacity, occurred in different interstitial lung disease, is typical for an evolution towards organizing pneumonia. Pooled analysis of studies evaluated the presence of linear opacity among patients with various disease severities indicated this nding is more likely to be observed in severe cases of COVID-19 than that observed in the common type ( Figure 4). Temporal analyzing of multiple chest CT scans demonstrated a speci c pattern during which linear opacity, observed 7-31 days after disease onset, indicates sub-segmental atelectasis or secondary organizing pneumonia. This pattern might be representative of irreversible brosis (21,22).

Bronchial wall thickening
Abnormal thickening of bronchial walls, arisen from various pathological entities, usually indicates in ammation of the airways. A few studies investigated the association of bronchial wall thickening with the severity of COVID-19 infection, resulting in the con icting ndings. Pooled analysis of these ndings indicated signi cantly higher odds of bronchial wall thickening in severe cases of COVID-19 compared to the common type ( Figure 5). Airway wall thickening, which has been attributed to the high level of Vascular endothelial growth factor (VEGF), is suggested to re ect the duration and severity of disease and the degree of air ow obstruction in asthma (23,24). On the other hand, recent evidence has revealed an increase in the levels of VEGF in COVID-19 patients (25)

Bilateral lung involvement
Many studies investigated the distribution of lung involvement in patients with various severity indicating the con icting results. However, analysis of the pooled data showed bilateral lung involvement is 3.44 times more likely to be observed in severe cases of COVID-19 compared to the common type ( Figure 6). Irrespective of the disease severity, some previous studies indicated that bilateral lung involvement is the most anatomic distribution of COVID-19 pneumonia (26,27).

Strength and Limitation
To the best of our knowledge, this meta-analysis is the rst one that evaluated chest CT features associated with clinical severity of COVID-19 infection. Moreover, high quality studies with a relatively large number of participants have been included in this study. In addition, all studies included in this meta-analysis followed the same classi cation for determination of disease severity.
However, this study has several limitations. First of all some features of chest CT imaging have not been included in the meta-analysis.
Furthermore, some patients might have comorbidities such as chronic obstructive pulmonary disease (COPD) that could affect imaging appearance. Moreover, all included studies were retrospective that the results might be in uenced by the confounding factors. In addition, because of different CT scanners and interpreting radiologists, the reported imaging features might be variable across sites. Lastly, this meta-analysis indicated a signi cant heterogeneity between the studies regarding some CT features. Moreover, results on GGO might be affected by the publication bias.

Conclusion
In summary, our meta-analysis of the observational studies shows bronchial wall thickening, linear opacity, and GGO are more likely to be associated with the severe type of COVID-19 compared to the common type. However, consolidation in chest CT was not associated with the severity of clinical presentation. Regarding the lesion distribution, as expected, bilateral lung involvement was more likely to be observed in severe cases rather than common ones. Identi cation of the chest CT images associated with severe cases of COVID-19 is useful for early diagnosis and timely treatment of the severe and fatal cases of COVID-19 infection.

Declarations
Funding This study was not supported by any funding.
Compliance with ethical standards Con ict of interest The authors declare that they have no con ict of interest Human and animals rights The present meta-analysis was performed using data extracted from published papers. All procedures were in accordance with the 1964 Helsinki declaration and its later amendments.
Informed consent The present meta-analysis was performed using data extracted from published papers. Figure 1 Flow diagram of literature search and study selection.     Forest plot showing the association between bilateral lung involvement in chest CT images and severity of clinical presentation.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.