Introduction

The novel Coronavirus disease 2019 (COVID-19) is to date a global pandemic, affecting approximately 1,600,000 individuals and with nearly 100,000 deaths and 215 countries, areas, or territories implicated.1 These data were updated to April 12, 2020.1 The COVID-19 disease can affect all age groups, although children and/or adolescents seem to be less susceptible to this infection.2 Furthermore, the reports of severe COVID-19 forms in pediatric populations are rare.3 At present, however, data regarding COVID-19 in children and adolescents are scarce and, specifically, a systematic description of the clinical spectrum of such disease in this age group is lacking.

Therefore, the aim of this meta-analysis was to assess the overall prevalence of clinical signs, symptoms, and radiological findings in children and/or adolescents with COVID-19 disease by meta-analyzing data from observational studies available so far.

Methods

Data sources and searches

We searched in PubMed, Scopus and Web of Science databases in order to identify all observational studies, published until April 11, 2020, evaluating COVID-19 in children and/or adolescents. Only studies with a sample size ≥4 were considered. No limitations in terms of race or ethnicity were used. Exclusion criteria were: (i) abstracts, editorials, comments, reviews, guidelines; and (ii) studies that have involved adult populations. Nineteen observational studies were identified with the terms “COVID19” (OR “COVID-19” OR “Coronavirus disease 2019”) with filter “Humans” and “English language”. Indeed, papers in non-English language were excluded. Given that the eligible studies were observational, we followed the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines for the meta-analysis.4

Data extraction and quality assessment

Three investigators (A.M., C.Z. and E.R.) evaluated the abstracts independently and then they achieved the full texts of relevant papers. They read the articles independently and decided if they met the inclusion criteria. Discrepancies were solved by discussion with a fourth investigator (A.D.). For all eligible studies, data on sample size, population, clinical and radiological features, and outcomes were extracted from all eligible studies and meta-analysis was carried out using random-effects modeling.

Two authors (A.M. and A.D.) evaluated the risk of bias independently. Given that the eligible studies were observational, we used the Newcastle−Ottawa Scale (NOS) adapted for cross-sectional study.

Data synthesis and analysis

The primary outcomes of the study were the percentage of children and/or adolescents with fever, cough, nasal congestion, dyspnea, diarrhea, or abdominal pain as well as the percentage of children and/or adolescents with mild or severe symptoms. In addition, we assessed the percentage of children and/or adolescents with evidence of pneumonia at computed tomography (CT) scan. The pooled prevalence of children and/or adolescents with clinical signs and symptoms was calculated with a random-effects model.5 The confidence intervals (CIs) were calculated with the method described by Wilson et al.6,7 Seeing that some eligible studies had prevalences of various clinical signs and symptoms and radiological findings equal to zero, we computed the pooled estimates also after the Freeman−Tukey double arcsine transformation.6,7

In order to assess the heterogeneity, we firstly used visual examination of the forest plots. The heterogeneity among eligible studies was also investigated by the I2-statistics. According to Higgins and Thompson,8 interpretation of I2-statistics is as follows: I2 values of nearly 25% show low heterogeneity; I2 values of roughly 50% show medium heterogeneity; and I2 values of roughly 75% show high heterogeneity. The publication bias was evaluated by the funnel plots and the Egger’s regression asymmetry tests for each primary outcome measure.9

In order to investigate the possible sources of heterogeneity among eligible studies, we also performed subgroup analyses. Specifically, in relation to the data observed in the eligible studies, the pooled prevalences of various signs and symptoms were evaluated stratifying the studies by study country (i.e., Asian countries vs. non-Asian countries). All statistical analyses were made with STATA® 14.2 (Stata, College Station, TX).

Results

Supplementary Fig. S1 reports the flow diagram of the study selection. As summarized in Table 1, we considered 19 eligible observational studies,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28 including a total of 2855 (mainly Chinese) children and/or adolescents (mean age 6.9 ± 7.0 years; 50.3% males) with confirmed diagnosis of COVID-19 disease by nasopharyngeal swabs. Eighteen of the eligible studies provided data on major adverse events,10,11,12,13,14,15,16,18,19,20,21,22,23,24,25,26,27,28 whereas one described only the severity of the disease.17 Overall, as reported from Supplementary Tables S1–S6, ∼47% of children had fever (95%CI 22−72%; I2 = 98.6%), 37% cough (95%CI 15−63%; I2 = 98.6%), 4% diarrhea (95%CI 0−12%; I2 = 92.2%), 2% nasal congestion (95%CI 0−7%; I2 = 87.7%), 1% dyspnea (95%CI 0−7%; I2 = 91.5%) and 0% abdominal pain (95%CI 0−1%; I2 = 76.3%). Children and adolescents presented mild symptoms in 79% (95%CI 65−91%; I2 = 93.5%) of cases, whereas only 4% (95%CI 1−9%; I2 = 76.4%) of patients were considered critical. Fifteen studies10,11,12,13,14,15,16,17,18,20,21,22,23,24,27,28 described also the radiological findings at chest CT (Table 2). As reported in Supplementary Table S7, no evidence of pneumonia at CT scan was observed in 26% (95%CI 8−48%; I2 = 92.4%) of children and/or adolescents. Among those with pneumonia, confirmed by CT, 26.4% (95%CI 13−41%; I2 = 80.8%) presented a unilateral involvement (Supplementary Table S8), 16% (95%CI 5−29%, I2 = 81.2%) had bilateral involvement (Supplementary Table S9) and 9% (95%CI 0−24%; I2 = 88.7%) had interstitial pneumonia (Supplementary Table S10).

Table 1 Main characteristics of children and/or adolescents with COVID-19.
Full size table

The relatively high heterogeneity (I2 > 80%) observed in the overall primary analyses of our study may reflect important differences among the main characteristics of study populations, including age (from some days to 13 years), country (i.e., China, USA, and Spain), and different care setting. In fact, heterogeneity was partly attenuated stratifying the studies by country.

The Egger’s regression tests showed a significant asymmetry of the funnel plots for some primary outcome measures (i.e., fever, cough, diarrhea, and nasal congestion), indicating a possible publication bias (Supplementary Table S11). For this reason, we then did the nonparametric trim and fill analysis29 that have indicated that the impact of publication bias was very little (data not shown).

Discussion

The principal results of our meta-analysis are as follows: (a) in children and/or adolescents with COVID-19, the most important symptoms are fever and cough, although a relatively small number of individuals also had diarrhea; (b) mild symptoms were observed in most cases; and (c) no children and/or adolescents died.

Subjects of any age can gain severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection (Table 2). In a systematic review, Ludvigsson30 also reported that in children and/or adolescents symptomatic infection appears to be uncommon, although severe cases have been reported. Our meta-analysis substantially confirmed the results by Ludvigsson,30 but it adds further information regarding the prevalence of various clinical signs and symptoms, as well as the percentage of children and/or adolescents with specific alterations on CT by the meta-analysis of data reported in 19 observational studies available so far. In particular, our meta-analysis suggests that fever and cough, but also diarrhea, are the most frequent clinical characteristics that can be observed in children and/or adolescents with COVID-19. In addition, very recently, in a systematic review of 18 studies involving pediatric cohorts, Castagnoli et al.31 also confirmed that children and/or adolescents undergo less severe COVID-19 infection when compared to adults, thus having mild symptoms and a good prognosis.

Table 2 Main chest computed tomography (CT) findings of children and/or adolescents with COVID-19.
Full size table

The number of cases with COVID-19 observed in children and/or adolescents is scarcely reported in the daily reports of most non-Asian countries, but are, instead, commonly described in the Asian reports. For this reason, in our meta-analysis, we found 17 studies conducted in Chinese children and/or adolescents, only one conducted in the US population and one performed in Spain. In addition, it is important to note that more recently in a large Chinese report, only 2% of infections were observed in individuals younger than 20 years old.32 Similarly, in South Korea, 6.3% of 8000 infections were observed in individuals under the age of 20 years.33 Importantly, the majority of the studies reported hospitalization and symptoms at the event. Our report also documented that mild symptoms were observed in approximately 80% of cases, whereas more than 4% of children were considered critical. However, no one has died from COVID-19 in eligible studies of our systematic review and meta-analysis.

Along with other authors30,31 we believe that children and/or adolescents may have a better prognosis for COVID-19, when compared to adults, for the following reasons: (a) seeing that the SARS-CoV-2 S protein attaches the angiotensin-converting enzyme (ACE) 230,31,34 children and/or adolescents may be protected, as this enzyme may be less developed at a younger age35; and (b) the percentage of children and/or adolescents affected by COVID-19 having an exaggerated inflammatory response against the virus are not commonly described in the reports available so far.30,36 However, it is important to underline that in the middle of May 2020, the Centers for Disease Control and Prevention (CDC)37 and World Health Organization (WHO)38 have communicated the existence of a hyperinflammatory syndrome in children, also defined as Multisystem Inflammatory Syndrome in Children (MIS-C). Indeed, this is an emerging syndrome characterized by fever, many signs and symptoms, such as conjunctivitis, rash, edema, shock, and/or gastrointestinal syndrome, and also multiorgan failure.37,38 By study design, however, we were unable to capture this syndrome across eligible studies.

The principal limitations are that all the studies were mainly conducted in Asian countries (i.e., China). Another limitation is the sample size of each study that had collected few cases. In addition, we were not able to include the current manuscripts published in non-English language. Thus, it is plausible to suppose that some important pediatric cohorts were not included in our meta-analysis, given that at present the main data from studies and clinical cases involving children and adolescents come from China. However, in this context, it is also important to underline that other authors have made a similar choice in conducting meta-analyses on various aspects of COVID-19 outbreak. Lastly, even though we used a random-effects model, the interpretation of the results may require caution, given the heterogeneity observed in the analyses. However, the heterogeneity observed in our meta-analysis may reflect a mix of patients from different countries (i.e., China, USA, and Spain), with different ages ranging from some days to 13 years, and treated in different settings.

In conclusion, our report shows that children and/or adolescents tend to have a mild COVID-19 course with a good prognosis.