Chest CT findings of coronavirus disease 2019 (COVID-19): A comprehensive meta-analysis of 9907 confirmed patients.

OBJECTIVES: We performed a systematic review and meta-analysis of the prevalence of chest CT findings in patients with confirmed COVID-19 infection.
METHODS: Systematic review of the literature was performed using PubMed, Scopus, Embase, and Google Scholar to retrieve original studies on chest CT findings of patients with confirmed COVID-19, available up to 10 May 2020. Data on frequency and distribution of chest CT findings were extracted from eligible studies, pooled and meta-analyzed using random-effects model to calculate the prevalence of chest CT findings.
RESULTS: Overall, 103 studies (pooled population: 9907 confirmed COVID-19 patients) were meta-analyzed. The most common CT findings were ground-glass opacities (GGOs) (77.18%, 95%CI = 72.23-81.47), reticulations (46.24%, 95%CI = 38.51-54.14), and air bronchogram (41.61%, 95%CI = 32.78-51.01). Pleural thickening (33.35%, 95%CI = 21.89-47.18) and bronchial wall thickening (15.48%, 95%CI = 8.54-26.43) were major atypical and airway findings. Lesions were predominantly distributed bilaterally (75.72%, 95%CI = 70.79-80.06) and peripherally (65.64%, 95%CI = 58.21-72.36), while 8.20% (95%CI = 6.30-10.61) of patients had no abnormal findings and pre-existing lung diseases were present in 6.01% (95%CI = 4.37-8.23).
CONCLUSIONS: The most common CT findings in COVID-19 are GGOs with/without consolidation, reticulations, and air bronchogram, which often involve both lungs with peripheral distribution. However, COVID-19 might present with atypical manifestations or no abnormal findings in chest CT, which deserve clinicians' notice.

Introduction

Since the outbreak of coronavirus disease 2019 (COVID-19) – primarily known as 2019 novel coronavirus or 2019-nCoV – in December 2019, a global effort has been undertaken to make timely diagnosis and prevention of the disease possible. On the 30th of January 2020, the world health organization announced COVID-19 as a global health emergency [1]. As of April 25, 2020, COVID-19 has infected over five million patients with more than 342,000 confirmed deaths worldwide, and the WHO has characterized the situation as a pandemic [2,3].

Although real-time reverse transcription polymerase chain reaction (RT-PCR) is the only definitive diagnostic method and the current modality of choice for detection of COVID-19 infection, rapid spread of the disease mandates the healthcare system to look for a faster, easier, and more available diagnostic tool. With the growing body of evidence on COVID-19 infection, chest CT is establishing its role in early diagnosis of the disease. Chest CT is now considered as an integral part in evaluation of suspected COVID-19 patients as a reliable, quick, and practical method for detection and assessment of COVID-19, particularly in the epidemic areas. Recent studies have shown superior sensitivity of CT scan in comparison with RT-PCR method for diagnosis of COVID-19 [4,5].

The typical chest CT finding of COVID-19 infection is reported to be multiple bilateral ground-glass opacities (GGOs) with a predominantly peripheral distribution. Although these imaging manifestations are believed to be the general picture of COVID-19, the disease may present with a variety of CT findings [[6], [7], [8]]. However, a dearth of comprehensive investigations to consolidate the emerging evidence is felt in this regard. There are already a few systematic reviews within the literature evaluating the same topic, but they have a limited sample size, are affected by publication bias, and have not included more recent publications. Moreover, many of the published studies have used nonstandard (non-Fleischner Society) terms for describing lung abnormalities, which are included in the previous meta-analyses and have made the reported estimates somewhat inaccurate. In March 2020, a consensus statement of reporting CT findings was published by the Radiologic Society of North America (RSNA), which categorized the CT findings into typical and atypical [9]. This categorization was based on the frequency of findings reported in the literature published by the time. Since then, some reports indicated the frequency of atypical findings to be higher than what was previously thought. Herein, we systematically reviewed the available literature on the chest CT findings in patients with confirmed molecular diagnosis of COVID-19 infection, and evaluated the frequency of typical and atypical findings in a large sample size of over 9900 patients.

Methods

Data sources and search strategy

A computerized search was conducted to identify and to screen studies available online up to May 10, 2020, reporting the chest CT manifestations in patients with confirmed diagnosis of COVID-19. Three independent investigators (A.Z., R.R.O., and M.G.N.) searched the PubMed, Embase, and Scopus databases as well as Google Scholar search engine. We searched the literature applying various combinations of Boolean operators and relevant keywords, using the following search strategy:

(“COVID-19” OR “coronavirus disease 2019” OR “2019 novel coronavirus” OR “2019-nCoV” OR “coronavirus” OR “SARS-CoV-2” OR “SARS 2”) AND (“Computed tomography” OR “computerized tomography” OR “chest computed” OR “chest computerized” OR “chest CT” OR “CT scan” OR “CT imaging”).

All published studies in English language with an abstract or a full-text version containing any of the search terms relevant to our review were retrieved from the searched databases for further evaluation. Moreover, after selection of the eligible studies for inclusion in the review, their reference lists were manually checked for any possible missing relevant article.

Eligibility criteria

All studies containing data on chest CT findings of 10 or more patients with confirmed diagnosis of COVID-19 using RT-PCR were included in our study. We excluded articles of the following types: review, editorial, letter to editor, and reply to authors, unless they provided data on chest CT findings of 10 or more patients with confirmed COVID-19. Studies without sufficient information or those with no quantitative report of the frequency for chest CT findings were also excluded from the systematic review.

Moreover, studies that used nonstandard terminology for description of all their chest CT findings (those not compliant with Fleischner Society standards for thoracic imaging [10]) were excluded. However, for studies in which only some of the reported findings were nonstandard, we only included the standard terminology.

Study selection

Two investigators (A.Z. and M.G.N.) independently reviewed the titles and abstracts of all obtained studies for eligibility screening. After excluding the duplicates and irrelevant studies, all remaining studies underwent full-text review.

Four investigators (A.Z., M.G.N., R.R.O., and A.A.R.) independently reviewed the full texts of all eligible studies and reached a joint consensus to determine final inclusion. The flow diagram of the review process is presented in Fig. 1 .

Fig. 1

PRISMA flow diagram of the study.

If two studies were performed on similar patients from a single hospital with mutual authors and adjacent publication dates, the study with smaller sample size was excluded to reduce the chance of duplicate data. In order to reduce the risk of bias and duplicate data, only studies reporting chest CT findings on 10 or more patients were included in our meta-analysis.

Data extraction

Four investigators (A.Z., A.A.R., M.G.N., and R.R.O.) read the included full text articles to extract the study characteristics, demographic information, and chest CT findings. Extracted study characteristics were publication/acceptance date, first author, report origin (city and country from which the work originated), study design, and sample size. Extracted demographics of the patients included sex and age.

The frequency of each chest CT finding that was reported using standard terminology in patients with positive RT-PCR for COVID-19 was also extracted. Findings of the chest CT were organized according to the standard terminology for thoracic imaging provided by the Fleischner Society [10]. Accordingly, in an attempt to standardize the terminology we organized the reticular patterns including interlobular septal thickening and crazy paving in one single finding named “reticulation with/without GGO”. We also included nonspecific linear opacities (i.e. linear opacities, fibrotic stripes or streaks, and subpleural lines) and reverse halo sign in one single finding named “organizing pneumonia patterns”. We incorporated all pre-existing lung diseases reported by the studies, including chronic obstructive pulmonary disease, tuberculosis, bronchiectasis, malignancies, and asthma into one group of findings named “pre-existing lung diseases”. Besides, all chest CT patterns representing chronic lung diseases (e.g. usual interstitial pneumonia in idiopathic pulmonary fibrosis, and other interstitial lung diseases) were also included in this group as a single finding named “pulmonary fibrosis”.

We categorized the findings into typical and atypical in an attempt to be in accordance with the Radiological Society of North America Expert consensus statement on reporting chest CT findings related to COVID-19 [11]. Typical findings included GGO with/without consolidation, reticulation with/without GGO, round GGO/consolidation, air bronchogram, consolidation, organizing pneumonia patterns, and halo sign. Atypical findings included pleural thickening, pulmonary nodules, pleural effusion, lymphadenopathy, pneumothorax, and cavitation. Bronchial wall thickening and airway secretions (mucoid impaction) were included in central bronchial findings category.

The frequency of normal CT (no findings) was also recorded for each study. Furthermore, the distribution of findings in terms of the involved lungs, respective lobes, and the dispersion of the lesions in the lung field were extracted.

Quality assessment

All studies included in the meta-analysis underwent quality assessment. Quality appraisal of case series studies was performed using the institute of health economics (IHE) checklist and the critical appraisal tool to assess the quality of cross-sectional studies (AXIS) was used to assess the quality of the included cross-sectional studies [12,13]. In order to avoid score discrepancy between investigators, a single investigator (M.G.N.) evaluated all the included articles and scored them on a scale of 0–20.

The IHE quality appraisal tool for the evaluation of case series studies originally includes 20 items assessing the methodologic quality of case series studies to generate a score of 0 to 20. It provides an effective means to assess the quality of non-controlled studies, which constitute a major part of our study. In cases without repeated measurements, the study was awarded the scores for questions 11, 13 and 14, which are specific to studies with a repeated-measures design [13]. The AXIS is also a 20-item checklist developed for critical appraisal of the cross-sectional studies, scoring the quality of studies on a scale of 0 to 20 [12].

Publication bias

In an attempt to minimize the publication bias, we searched several databases and search engines. We also included gray literature and preprints in the systematic review. Moreover, we assessed publication bias using the trim and fill method, Egger's test, and funnel plots.

Statistical analysis

The prevalence data were pooled across included the studies using a random effects model. Data were summarized as frequency and 95% confidence interval (95%CI). Moreover, we assessed heterogeneity of the findings and provided its indicators including the I-squared (I2) index.

All analyses were performed using comprehensive meta-analysis software, version 3.3.070 (Biostat, Englewood). We followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) checklist.

Results

Characteristics of the studies

Our primary search strategy identified 251 studies from PubMed, 168 studies from Embase, and 183 studies from Scopus databases. Google Scholar search engine was also searched and 1442 studies were found. Another 124 studies were obtained through hand searching of the references of included articles and unpublished or preprint studies.

After removing duplicates, 1337 studies underwent title and abstract review, of which 392 proved to be eligible for full-text review. Some full-text articles were excluded due to improper study design, nonstandard terminology for reporting thoracic findings, or lack of prevalence data on chest CT findings (289 records). Finally, 103 articles with a pooled sample size of 9907 patients with confirmed diagnosis of COVID-19 who underwent chest CT were included in the meta-analysis.

The PRISMA flow diagram is shown in Fig. 1. Characteristics and patient demographics for all studies included in the meta-analysis are shown in Supplementary Table 1.

Meta-analysis

Table 1 summarizes the pooled prevalence of all chest CT findings in patients with confirmed diagnosis of COVID-19 infection, according to the results of meta-analysis. A forest plot for all CT findings is presented in Fig. 2 .

Table 1

Pooled prevalence of chest CT findings in COVID-19 pneumonia.

Chest CT finding		N studies	N patients	Prevalence (%)	95%CI	I2
Typical findings	GGO with/without consolidation	66	6224	77.18	72.23–81.47	89.50
	Reticulation with/without GGO	41	2667	46.24	38.51–54.14	90.21
	Air bronchogram	28	1953	41.61	32.78–51.01	90.99
	Rounded GGO/consolidation	7	385	38.02	22.14–56.97	78.76
	Organizing pneumonia patterns	33	2557	36.85	28.94–45.53	92.25
	Consolidation	44	4397	35.56	28.84–42.91	96.06
	Halo sign	7	601	25.63	13.86–42.46	94.82
Central bronchial findings	Bronchial wall thickening	10	873	15.48	8.54–26.43	83.19
	Airway secretions (mucoid impaction)	1	121	0.83	0.05–11.32	0.00
Atypical findings	Pleural thickening	12	1392	33.35	21.89–47.18	88.18
	Nodules	32	2582	13.11	9.26–18.25	84.60
	Pleural effusion	48	3963	6.96	5.09–9.45	78.25
	Lymphadenopathy	39	3197	5.19	3.53–7.57	88.05
	Cavitation	11	641	1.1	0.41–2.92	0.00
	Pneumothorax	2	167	0.89	0.11–6.92	0.00
Pre-existing lung diseases		45	6259	6.01	4.37–8.23	92.10
No findings (normal)		72	5936	8.20	6.30–10.61	83.57

GGO: ground-glass opacity, 95%CI: 95% confidence interval.

Fig. 2

Cumulative forest plot of CT findings.

Among the typical findings, GGOs with/without consolidation was the most common manifestation with a prevalence of 77.18% (95%CI = 72.23–81.47) reported from 66 studies with a pooled population of 6224 patients. Reticulation with/without GGO was reported in 41 studies with total population of 2667 patients and had the second highest prevalence (46.24%, 95%CI = 38.51–54.14). The third most common finding was air bronchogram (41.61%, 95%CI = 32.78–51.01), which was mentioned in 28 studies with a pooled population of 1953 patients.

The main atypical finding was pleural thickening which was reported in 12 studies with an overall population of 1392 patients (33.35%, 95%CI = 21.89–47.18). Bronchial wall thickening was the most prevalent airway abnormality on chest CT (17.27%, 95%CI = 10.65–26.75), which was reported in six studies including 625 patients.

We also observed that the lesions are most likely to involve the lower lobes (77.07%, 95%CI = 64.55–86.13). It is worth noting that 8.20% of patients with COVID-19 (95%CI = 6.30–10.61) had no abnormal findings in their chest CT scan, as reported by 72 studies with a pooled population of 5936 patients.

Of the 103 studies included, 86 had assessed pre-existing diseases and elaborated on the respective details, of which 41 reported no pre-existing diseases in any of the patients. In the remaining 45 studies with a cumulative population of 6259 patients, pre-existing lung diseases were present in 6.01% of patients (95%CI = 4.37–8.23).

According to our analysis, most of the lesions were distributed peripherally (65.64%, 95%CI = 58.21–72.36) and involved both lungs (75.72%, 95%CI = 70.79–80.06). Table 2 shows the distribution of chest CT findings in COVID-19 patients.

Table 2

Distribution of chest CT findings in COVID-19 pneumonia.

Distribution	N studies	N patients	Prevalence (%)	95%CI	I2
Bilateral involvement	70	5505	75.72	70.79–80.06	89.78
Peripheral involvement	43	3216	65.64	58.21–72.36	94.29
Diffuse involvement	28	2080	35.11	26.79–44.44	89.54
Central involvement	26	2160	6.11	3.89–9.47	91.73
Lower lobes	12	1265	77.07	64.55–86.13	95.46
Upper lobes	10	1155	47.70	32.37–63.48	95.57
Middle lobe	15	1487	47.87	35.40–60.60	94.44
1 lobe affected	23	1674	12.48	8.35–18.26	79.42
2 lobes affected	17	1154	10.01	6.15–15.88	68.57
3 lobes affected	17	1154	9.00	5.56–14.27	0.00
4 lobes affected	17	1154	13.81	8.86–20.89	28.38
5 lobes affected	19	1229	41.92	31.31–53.32	88.07

95%CI: 95% confidence interval.

Funnel plots regarding the publication bias for two major findings (GGOs and normal chest CT) are shown in Fig. 3, Fig. 4 . The results of Egger's test as well as trim and fill method for quantitative assessment of publication bias regarding several major findings of our study are illustrated in Table 3 .

Fig. 3

Funnel plot for ground-glass opacity.

Fig. 4

Funnel plot for normal CT findings.

Table 3

Quantitative results of assessment for publication bias.

Finding/distribution pattern	Egger's test		Trim and fill method
	Egger's intercept	p value	Number of trimmed studies	Point estimate after trim	Change after trim
GGO w/wo consolidation	2.61	0.001	15	070	0.06
Reticulation w/wo GGO	−0.64	0.652	0	0.50	0
Air bronchogram	−1.90	0.319	2	0.39	0.02
Consolidation	−0.41	0.825	0	0.40	0.04
Organizing pneumonia pattern	−0.14	0.912	5	0.43	0.05
Pleural effusion	−2.17	0.001	15	0.10	0.03
Lymphadenopathy	−3.09	0.001	14	0.08	0.36
Nodules	−2.53	0.001	9	0.18	0.05
Normal	−1.88	0.001	26	0.15	0.07
Pre-existing lung diseases	−2.88	0.002	11	0.08	0.03
Bilateral	0.41	0.602	12	0.70	0.06
Central	−4.36	0.001	6	0.07	0.03
Diffuse	−2.65	0.069	3	0.40	0.04
Peripheral	4.25	0.002	6	0.61	0.05
Middle lobe	7.17	0.004	0	0.48	0

GGO: Ground-glass opacities; COPD: chronic obstructive pulmonary disease; w/wo: with/without.

Discussion

This systematic review and meta-analysis was designed to analyze the reported chest CT findings in patients with COVID-19 pneumonia. Although chest radiograph is the first-line imaging modality in most patients with acute respiratory syndrome, lung CT is suggested as the subsequent study or even the imaging modality of choice in selected situations [117,118]. In spite of its relatively low specificity, chest CT has been reported to be a highly sensitive diagnostic tool for detection of COVID-19. CT scan has certain downsides like increased patient radiation, increased staff exposure, and time limitations in unstable patients. Taking these into account, CT scan is not considered the first diagnostic modality in many guidelines. The American College of Radiology has recommended chest CT only for hospitalized and symptomatic COVID-19 patients with specific clinical indications, not as the first-line diagnostic or screening tool. There has also been some issues regarding the risk for spreading infection with the use of imaging equipment and the latest guidelines have advised the clinicians to minimize visits of suspicious patients to healthcare facilities. Yet for a considerable number of patients, especially in the regions with limited PCR resources, some medical practitioners might require chest CT at the first visit to make decisions whether to test patients for COVID-19 infection, admit them or provide other treatments [119].

Accurate diagnosis and management of COVID-19 patients depend on knowledge of typical and atypical findings in chest CT scan and familiarity with the evolution of imaging manifestations. A comprehensive understanding of imaging manifestations of COVID-19 is particularly important since healthcare professionals are under increasing pressure to make rapid and reliable diagnoses [120]. Besides, radiologists should familiarize themselves with the CT appearance of COVID-19 infection in order to be able to identify findings consistent with infection in patients imaged for other reasons [121]. With respect to the call for studies on the clinical characteristics of COVID-19 from the WHO, in this systematic review and meta-analysis study, we have summarized all the available data on chest CT findings of around 10,000 patients with confirmed COVID-19 diagnosis derived from published and yet unpublished preprint reports from four countries. This is the first systematic review that evaluates the findings according to the RSNA categorization of typical and atypical CT features of COVID-19 infection.

We found that the vast majority (75.72%) of COVID-19 cases had bilateral involvement of lungs and most of the lesions were peripherally distributed (65.64%). A higher percentage of patients appeared to have involvements in all five lobes (41.92%). Moreover, lower lobes were more commonly involved (77.07%). A recent meta-analysis of 4121 patients with COVID-19 from 34 retrospective studies revealed relatively similar results, indicating predominance of bilateral (73.8%) and multilobar (67.3%) lung involvement [122].

The most common typical finding in chest CT of patients with COVID-19 was GGO with/without consolidation (77.18%), followed by reticulation with/without GGO (46.24%), and air bronchogram (41.61%). Other typical findings in the order of prevalence were rounded GGO/consolidation (38.02%), organizing pneumonia patterns (36.85%), consolidation (35.56%) and halo sign (25.63%). Bronchial wall thickening was also the most common airway finding in the patients (15.48%). Other relatively uncommon atypical findings were nodules (13.11%), pleura effusion (6.96%), lymphadenopathy (5.19%), cavitation (1.1%), and pneumothorax (0.89%). Overall, airway secretion (mucoid impaction) was the rarest chest CT finding, with a prevalence of 0.83%. Pleural thickening, generally categorized as an uncommon finding, had a prevalence of 33.35%, which is much higher than expected. We found the prevalence of pre-existing respiratory diseases to be around 6% in patients with confirmed COVID-19 diagnosis, as reported in a large body of reviewed studies. This relatively low prevalence implies that the majority of reported abnormal chest CT findings in the patients can be mainly imputed to COVID-19 and not to the pre-existing conditions. However, as most of the included patients were symptomatic and of older age, some of the reported findings might inevitably be due to the aging-related respiratory conditions that were not accurately reported by the studies. This can be the reason behind the high prevalence of some atypical findings in our study. For instance, the relatively high prevalence for pleural thickening, which is considered an atypical finding for COVID-19, can be because it is a non-specific finding and can be present in a variety of abnormal respiratory conditions. A study evaluating the incidence of new pleural thickening in patients with multiple CT studies could be helpful in clarifying this topic.

Relatively similar results were reported by a recent meta-analysis of 4121 patients with COVID-19. Zhu et al. reported the most common findings to be GGO (68.1%), air bronchogram (44.7%), crazy-paving pattern (35.6%), and consolidation (32.0%), respectively. They found pleural thickening to be present in 27.1%, while lymphadenopathy and pleural effusion were present in 5.4% and 5.3% of patients, respectively [122].

The typical findings in chest CT scans of patients with confirmed COVID-19 were generally comparable to those found in patients with other viral pneumonias, mainly severe acute respiratory syndrome and Middle East respiratory syndrome [21,98,123,124]. However, atypical and uncommon presentations on chest CT are also reported by some studies. Therefore, clinicians, particularly radiologists, should be aware of the uncommon findings as well, especially in cases with pulmonary comorbidities such as lung cancer [[125], [126], [127], [128], [129]].

Our meta-analysis showed an approximate 91.8% rate of abnormal CT manifestations, indicating an 8.2% rate of normal CTs. This is in line with the 8.4% prevalence of normal chest CTs in a recent meta-analysis [122]. Inconsistently, a recent meta-analysis of eight studies on COVID-19 patients, which is not published yet (preprint) reported a higher rate of 95.6% for incidence of abnormal chest CT findings [130]. Another published meta-analysis of 10 studies also reported a higher rate (96.6%) of abnormal chest CT findings in COVID-19 patients [131]. This inconsistency might be because the mentioned studies included fewer articles that were published early at the beginning of the outbreak, and focused mainly on critically symptomatic patients, who are more likely to present abnormal chest CT findings. Besides, it can be due to the difference in the time of performing CT scan in different studies. As reported in several studies, patients with initially normal chest CT scans in the first days of the disease can develop pulmonary lesions in the following days [36,132].

On the one hand, it is worth mentioning that although some patients with COVID-19 might not demonstrate abnormal findings in their initial chest CT studies during the first days of infection, they would probably develop pulmonary opacities in later stages. Therefore, it is recommended that the clinicians follow the patients, since normal chest CT does not unequivocally rule out COVID-19 [36]. On the other hand, initial RT-PCR testing might yield negative results in some patients who have presented with typical chest CT findings and clinical manifestations that raise the suspicion of COVID-19. Since repeated RT-PCR results might also be negative until the late stages of the disease, chest CT can be particularly helpful in such patients, in that it enables the clinicians to detect highly probable COVID-19 patients and take precautionary measures to isolate them and control the spread of the disease [126,133,134]. In patients who show negative specimen test, the clinicians should consider the possibility of false-negative results. Therefore, it can be recommended to use a combination of CT scan and RT-PCR test in patients with initially negative RT-PCR screening but high clinical suspicion of COVID-19 infection [5,135]. Early diagnosis of COVID-19 cases enables healthcare professionals to offer adequate clinical monitoring and necessary interventions for the patients, as well as taking precautionary actions to control the spread and prevent further transmission of the infection [[136], [137], [138], [139]].

Regarding the assessment of publication bias, funnel plots showed asymmetry and the results of Egger's test confirmed the existence of publication bias for some of our findings. Although publication bias has inevitably affected our findings, quantified assessment using trim and fill method showed that it did not have a significant effect on the estimated prevalence of our findings, except for lymphadenopathy, which is an atypical finding.

The results of this study can provide useful insights into the prevalence of CT findings in patients diagnosed with COVID-19 pneumonia. This can be helpful particularly in areas where routine RT-PCR testing is not available or for clinicians facing cases with high clinical suspicion of COVID-19 infection who still lack a confirmed molecular diagnosis.

This study has several limitations. For instance, there are still a limited number of studies available from outside China. In order to provide a more comprehensive understanding of chest CT findings of COVID-19, more studies from other parts of the world are required. Another limitation of our study is that although we incorporated reported respiratory comorbidities into our analyses, there is still an inevitable possibility that some findings might be related to chronic underlying conditions other than COVID-19, particularly because the disease severity tends to be higher in the older population who carry a higher rate of comorbidities. Possible selection bias in the included studies can also be a limitation, for it is quite likely that most of the imaged patients were selected from those with the worst clinical conditions. Besides, although we tried to minimize the duplicates by excluding small studies and those reporting similar patients from the same hospital in a limited period, our study inevitably carries a risk of including duplicate cases as most of the available studies are from a limited number of cities in China. Moreover, high levels of heterogeneity were observed regarding some of our findings, which can be possibly due to the heterogeneous characteristics of included populations and the difference in time of performing chest CT. The technical characteristics such as slice thickness and tube current of performed CT scans are not evaluated in this study due to the heterogeneity in the reviewed studies in this regard. This heterogeneity of the technical details may have some effect on the frequency of reported findings in different studies. Finally, only a few studies reported accurate time course of chest CT findings in COVID-19 patients and we could not incorporate this item in the meta-analysis. Further studies providing adequate data on accurate time of appearance for each chest CT finding in patients with COVID-19 infection are required to shed light on this important matter.

In conclusion, GGOs with or without consolidation, reticulation, and air bronchogram were the typical chest CT findings in COVID-19 pneumonia. Lesions often involve both lungs with peripheral distribution. In addition to these common findings, it should be considered that COVID-19 infection might present with a variety of manifestations in chest CT scanning. The alarm should sound for every first-line clinician when facing any of these CT manifestations, especially in areas with epidemic COVID-19 infection and limited access to RT-PCR. We recommend that future investigations focus on the changes in CT findings over time and the role of CT manifestations in the patients' prognosis.

The following is the supplementary data related to this article.

Supplementary Table 1

Characteristics of the studies included in the meta-analysis.

Declaration of competing interest

None.