Introduction

A novel strain of coronavirus (referred to as 2019-nCoV or SARS-CoV-2), which causes the sometimes severe respiratory infection COVID-19, was first identified in Wuhan city, China, toward the end of 2019 [1]. By 12 March 2020, COVID-19 was declared a global pandemic by the World Health Organization (WHO), and at the time of writing it had spread to 187 countries, with almost 3.5 million confirmed cases, and had claimed more than 244,000 lives [2].

Epidemiological studies originating from China have shown that children are less likely to be clinically affected than older adults, with one study finding only 0.9% of those affected being younger than 15 years old [3]. A different Chinese study, which included 731 confirmed paediatric cases, found that the majority (84.1%) sustained either mild or moderate symptoms, with <3% described as being severely or critically affected [4]. Only one study reported a death in a child, a 10-month-old with multiorgan failure and intussusception [5]. In the United States, data published on 2 April 2020 [6] showed that only 1.7% (n=2,572) of people affected by COVID-19 were younger than 18 years, with 0.58–2% of children requiring intensive care admission. Given the low number of paediatric cases, kee** abreast of the latest information and assimilating the combined knowledge of radiographic findings in infected children is challenging. Whilst several systematic reviews of imaging findings in COVID-19 cases have been performed for adults [7, 8], none has specifically focused on children.

The overall objective of this study was therefore to assimilate the available information on imaging features of COVID-19 disease in children. Particular points of interest include identifying typical imaging findings during diagnosis and follow-up stages of the infection, and whether any features might be used as prognostic markers to determine patient outcome. Where knowledge gaps exist, we intend to highlight these and suggest potential future avenues for research.

Materials and methods

Ethics approval was not required for this retrospective review of published data. We followed Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines for transparent reporting of systematic reviews. This study was registered in PROSPERO, an international prospective register of systematic reviews (Registration ID: CRD42020175945 [9]).

Literature review

A systematic literature search was performed of Medline (Ovid), Embase and the Cochrane Library databases for the latest articles published between 1 January 2015 and 17 March 2020 (5-year range), using database-specific Boolean search strategies with key terms and word variations relating to all three categories:

  1. (1)

    “coronavirus”, “COVID-19”, “SARS-CoV-2” or “2019-nCoV”;

  2. (2)

    “paediatrics”, “children”, “neonate”, “infant” or “adolescent”;

  3. (3)

    “radiology”, “imaging”, “ultrasound”, “CT”, “MRI” or “radiography.”

Full search terms are shown in the supplementary material (Tables S1, S2 and S3). To include as many recent articles as possible, we also performed a grey literature search (i.e. literature not formally published in sources such as books or journal articles [10], such as government white paper articles and guidelines) using the same keywords on Google Scholar and for any WHO Global Library publications. Additional articles were retrieved by manual screening of the reference lists of included studies and relevant review articles/editorial pieces. The initial search was conducted on 17 March 2020. A repeat search was conducted on 23 March 2020, and again on 30 April 2020 for any further eligible manuscripts.

Eligibility criteria

Inclusion criteria encompassed all studies investigating and describing imaging findings of confirmed COVID-19 infection in children, using reverse transcriptase polymerase chain reaction (RT-PCR) testing. Studies were limited to human subjects, including foetuses (any gestation) and children (ages 0–18 years). No restrictions were placed on type of imaging modality, number of cases described or type of clinical setting. To widen our search and include as many cases as possible, we included case reports. No language restrictions were used given that many early articles on imaging findings have been published in Chinese. Where this was the case, we sought online translation services and advice from native-speaking colleagues to interpret reported findings.

Exclusion criteria included studies reported as editorials, opinion articles or multimedia files (online videos, podcasts). Suspected but unconfirmed cases of COVID-19 were not included. Studies relating to other coronavirus-related illnesses, such as Middle East respiratory syndrome (MERS) or severe acute respiratory syndrome (SARS) were excluded. We also excluded any articles reporting on a mixed adult and paediatric cohort where imaging results for the paediatric cohort could not be extracted.

Data extraction and quantitative data synthesis

All articles were independently searched by two reviewers (S.C.S. and S.T. with 7 years and 10 years of paediatric radiology experience, respectively). They examined abstracts of suitable studies and obtained full papers according to the eligibility criteria. Disagreements were resolved by consensus.

The same two reviewers (S.C.S., S.T.) independently extracted data from the full articles into a database (Excel; Microsoft, Redmond, WA), which included the following factors: study design, study setting/country, population demographics (e.g., gender, age, underlying comorbidities), sample size, patient outcomes (number of mortalities), imaging modality and imaging findings (pattern and location of involvement of disease) and results of any follow-up imaging.

Missing data were recorded as “not recorded” or “not stated”. Authors of published studies were not contacted because of the tight time constraints involved in the completion of the systematic review during the unprecedented time of need for this information.

Methodological quality

The quality for each included study was assessed using the National Institutes of Health Quality Assessment Tool for Case Series Studies [11] by two reviewers (S.T. and J.L, with 17 years of paediatric radiology experience). Disagreements were resolved by consensus review. Any that could not be resolved by consensus was arbitrated by a third reviewer (S.C.S.).

Statistical analysis

We planned a meta-analysis to assess association of imaging findings with patient outcomes and demographic data; however, we omitted this because of a lack of sufficient data, with many cases being incompletely reported. Therefore, we used descriptive statistics to determine frequency and percentages of imaging appearances across different articles.

Results

Included studies

During the initial literature search, after removing duplicates, we identified 146 unique records. After screening titles and abstracts, we excluded 100 studies and checked 46 full-text articles. Reasons for exclusions included insufficient description regarding imaging results (n=8), opinion pieces (n=7), adult population only (n=5), no confirmed (only suspected) COVID-19 cases (n=3), retracted article (n=2) and no full text available (n=1). After the second and third searches of the databases, we found eight more records that met our inclusion criteria and reviewed them.

Although it was not explicitly stated in the articles, we found that a small case series of 8 children [12] and a larger one of 115 children [54]. Nevertheless, given that severe disease in children is less likely, the extent to which POCUS might be helpful is questionable, although it has been recommended as one of several potential options for lung assessment by a Chinese expert consensus review for neonatal management in COVID-19 [55]. Two adult publications have reported the use of [F-18]2-fluoro-2-deoxyglucose (FDG) positron emission tomography (PET)/CT in COVID-19 [56, 57], and suggested that it could help monitor disease progression and treatment outcomes, particularly by detecting residual activity in mediastinal lymph nodes. This modality has not been reported in infected children, and adult studies have only included a small number of cases (4 patients or fewer). The added value is thus still undetermined [58] and should not be first studied in children, particularly given the increased radiation burden.

Our review also did not find many articles reporting imaging findings in immunocompromised children, so it is undetermined whether these features might differ from those without health conditions. It has been well documented that more severe infections are found in immunocompromised children with other strains of coronavirus [59] and that these infections can spread to other parts of the body, such as the brain, causing encephalitis [60]. Whether this also occurs with COVID-19 remains to be seen. It is interesting that the only child death reported in this review was also suffering from an intussusception. Whether this was triggered by the underlying viral infection is unclear, but it is worth noting that gastrointestinal complaints can be the first and more prominent symptom of COVID-19 in some people and has been reported in an adult series [61].

There are several limitations to this work. Given the widespread public health interest, several manuscripts on the topic of COVID-19 are emerging each week, many bypassing usual peer review processes. It is likely that some information might be missing but later included, or in certain cases articles retracted (as for two articles during our screening process). Nevertheless, where possible, we have tried to mitigate this by conducting our literature review thrice in order to provide the most up-to-date information from reliable sources. Whilst not all imaging findings in all cases were reported in the studies, we described all available findings to give a general overview of the imaging pathology. Future works on the study of COVID-19 imaging findings could be improved by the use of standardised detailed descriptors for imaging findings (i.e. stating both the pattern and localisation of findings), in line with RSNA guidance [44], with clearly stated indications for imaging where possible.

Second, because of the origin of the virus in China, some articles have been published in a language other than English, or in English by non-native speakers. This might have hampered our understanding and interpretation of the data, although we used online translation services where required. Whilst other systematic reviews on the topic of COVID-19 have excluded articles not written in English, we thought it was important to review as many foreign-language articles where possible to increase our collective knowledge base, particularly given the few reported paediatric cases.

Finally, the majority of articles have included children from China, in particular Wuhan city. It is unclear whether some of these paediatric cohorts overlap, although we did identify three papers where there was clear similarity in many of the patients described, and we avoided repetition of findings in summary results. It is also important to highlight that differences in indications for CT imaging in children might also exist (which could explain why Ma et al. [23] found slightly more abnormalities on CT in children than adults), but unfortunately these indications were not made clear in the publications.

Conclusion

Chest imaging findings in children with COVID-19 are frequently normal or mild, with unilateral changes that include patchy ground-glass opacification, commonly affecting the lower lobes. Imaging appearances at follow-up frequently remain normal or improve in the majority of children. Chest CT imaging adds little to the further management of the child and should be reserved for severe cases or for identifying alternative diagnoses. Further areas of research should include information on imaging and clinical characteristics in immunocompromised children with COVID-19, and information on long-term follow-up, particularly in the more severely affected children. We should therefore be prudent with the usage of CT, particularly at a time where resources are stretched, and only use it in the more vulnerable populations.