COVID-19 in 7780 pediatric patients: A systematic review

Introduction

In December 2019, an unprecedented number of pneumonia cases presented in adult individuals from Wuhan, China [1]. Despite rapid action by the Chinese government and health officials, the number of similar presenting cases continued to rise at an alarming rate [2]. By January 2020 an emerging zoonotic agent, known as Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2), was identified in respiratory samples in patients diagnosed with pneumonia who subsequently developed respiratory failure [1]. The spread of SARS-CoV-2 from human to human, through respiratory droplets, has now resulted in a worldwide outbreak, now classified as a pandemic by the World Health Organization [3].

As of June 3rd, 2020, there has been more than 6·4 million confirmed cases worldwide and >380,000 fatalities [4] Most symptomatic cases have occurred in the adult population, characterized by fever, cough, malaise, and frequent hospitalization [1]. Accordingly, most of the published data is derived from adults with coronavirus disease 2019 (COVID-19) who were hospitalized in China [5]. As the pandemic continues, we are now observing numerous reports describing the clinical presentation and hospital course of children with confirmed COVID-19 [5].

What is currently known is that children have milder symptoms and are less likely to be hospitalized when compared to adults [6]. However, on May 14th, 2020 the United States Centers for Disease Control and Prevention (CDC) released a health advisory reporting a multisystem inflammatory syndrome in children (MIS-C) associated with COVID-19 [7]. This statement stemmed from a subset of pediatric patients manifesting with severe inflammation, multi-organ failure, and testing positive for SARS-CoV-2 [8,9].

Our goal was to conduct a systematic review: (i) to understand the clinical picture and presentation of pediatric patients with confirmed COVID-19, and (ii) to provide an initial observation of the signs, symptoms, and laboratory findings of pediatric patients who developed MIS-C.

Methods

2.1 Search strategy and selection criteria

Our methods adhere to the guidelines established by Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). Our study protocol was registered with PROSPERO (International Prospective Register of Systematic Reviews) under the following identifier # CRD42020182261.

We conducted a systematic search in the following databases: PubMed, LitCovid, Scopus, and the WHO COVID-19 database. Additionally, we searched for studies that included the following terms- “novel coronavirus, COVID-19, 2019-nCOV, SARS-CoV-2, pediatric, child, and neonate” into the freely accessible research domains of JAMA, Lancet, NEJM, CHEST, and Google Scholar. The last search was performed on May 14th, 2020 and was not limited by language (translation performed with Google Translate).

We included published or in press peer-reviewed articles reporting pediatric cases of confirmed COVID-19. We accepted the following types of studies: cross-sectional, cases series, case-control, case reports, review articles, opinion papers, and letters to journal editors that incorporated clinical, laboratory, imaging, and hospital course of pediatric patients. The pediatric population included neonates, children, and young adults up to 21 years of age. We set the upper limit of age to 21 years as several countries use this number to stratify their pediatric versus adult data. Patients were included if SARS-CoV-2 was detected by real time reverse transcription polymerase chain reaction (RT-PCR) in nasopharyngeal, throat, blood, or stool samples at any point of their clinical evaluation. Suspected cases of COVID-19 without positive RT-PCR were excluded in this study. Furthermore, we also excluded in vitro studies or manuscripts focusing on animal experiments.

Screening by title and abstract was conducted independently by at least two investigators (AH, KC, or AxM). A third investigator (AM) was consulted to resolve differences of opinion in either phase. Subsequent full-text review and data extraction was conducted by all investigators using a standardized online form shared among the authors. Data retrieved from each article was cross-checked by at least two independent investigators.

Our outcomes of interest were to describe the clinical signs, imaging findings, and laboratory results characteristic of pediatric patients with confirmed COVID-19. Also, we wanted to provide an initial description of children with confirmed diagnosis of SARS-CoV-2 who develop MIS-C. We used the definition by the CDC to define MIS-C

(e.g., fever, laboratory evidence of inflammation, and evidence of clinically severe illness requiring hospitalization, with multisystem (≥2) organ involvement with no alternative diagnosis, and positive for SARS-CoV-2 infection) [7]. Control cases were patients from the same case series who did not meet criteria for MIS-C or studies that presented individual patient data where MIS-C could be definitively ruled out.

2.2 Data collection and risk of bias assessment

Data extraction was performed by all investigators and compared by at least two investigators for consistency. Data collected included the type of article (e.g., case series), country of origin, number of pediatric patients, demographic information, and all clinical symptoms (e.g., fever, cough), laboratory values (e.g., CBC, LFTs, BMP), imaging studies (e.g., chest x-ray, CT, MRI), clinical outcomes (e.g., ICU admission), and treatments provided (e.g. antivirals).

The risk of bias for observational studies was appraised through the quality assessment tool published by the National Institutes of Health [10]. We opted to use this guide as the development of the assessment tool was conducted rigorously by researchers in the Agency for Healthcare Research and Quality Evidence-Based Practice Centers, the Cochrane Collaboration, the United States Preventive Services Task Force, the Scottish Intercollegiate Guidelines Network, the National Health Service Centre for Reviews and Disseminations, and consulting epidemiologists. Moreover, it was a preferred tool in a systematic review on risk of bias assessments used in PROSPERO-registered protocols [11]. Risk of bias was assessed independently by at least two investigators and disagreements were resolved by a third researcher (AM). Furthermore, the level of evidence was assessed according to Sackett [12].

2.3 Data analysis

All laboratory data were converted to similar units and presented as mean with standard deviation (SD). Laboratory information presented as median (IQR) were converted to mean (SD), and denoted when unable to convert [13]. Publications that provided multiple timepoints (e.g., hospital course of individuals) for laboratory results were gathered and averaged. If the symptom was present anytime during the hospitalization, it was considered positive and characterized as a count with percent. A similar approach was taken for imaging information. Means, standard deviations, and proportion ratios were calculated using Microsoft Excel.

Statistical analyses between COVID-19 pediatric patients with/without MIS-C was conducted on STATA v·13. All statistical tests were two-sided, and significance was defined as a p value <0·05. Continuous data was summarized as mean (standard deviation) or median (interquartile range) and assessed by Student's t-test or Wilcoxon rank sum. Categorical data was summarized as counts (percent) and analyzed by Fisher's exact test.

2.4 Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

The search yielded 1,142 studies. After removing 237 duplicates, 905 articles were reviewed by abstract and title. After initial screening, only 319 articles met inclusion criteria and underwent full text evaluation. Publications that were retracted, or consisted of editorials, reviews, or commentaries that did not meet our criteria were removed, generating a final list of 131 articles (see Fig. 1).

Fig. 1

PRISMA flow diagram.

Studies included in this review were published between January 24th to May 11th, 2020. Eight studies were cross sectional, 75 were case series, and 48 were case reports (refer to Table 1). Twenty-six countries were represented with the largest data derived from 2572 children from the United States. China comprised 64·1% of the studies included in this review. Appendix 1 displays publications by the country of origin.

Twenty of the publications pertained to the neonatal population and the ages extended from an extremely premature neonate at 26 weeks gestation to 20 years of age. The level of evidence for all of the studies was 5 (1 is highest, 5 is lowest) and the risk of bias scores were between 2 to 7 (1 is lowest, 9 is highest, refer to Appendix 2).

A total of 7780 COVID-19 positive children were included. Fifty six percent of the individuals were male (Table 2).

The mean age was 8·9 years (SD 0·5) and 75·6% of patients were exposed to a family member who was diagnosed with COVID-19. The most common method for detection of the virus was through nasopharyngeal or throat swab (86·5%). Need for intensive care unit observation or treatment was low (3·3%). Twenty studies (n=655 individuals) reported an underlying medical condition; COVID-19 positive children who were immunosuppressed or had a history of a respiratory or cardiac condition comprised the majority (65·%). Moreover, influenza and Mycoplasma were the most common co-infections (see Table 3).

Table 4 summarizes clinical symptoms and imaging findings in COVID-19 confirmed pediatric patients. No symptoms were described in 456 of 2367 patients (19·3%), while the two most common symptoms were fever (59·1%), and cough (55·9%). While upper respiratory symptoms were characteristic of COVID-19, some patients presented with mild or often overlooked symptoms such as fatigue, abdominal pain, or decreased appetite [[14], [15]–16]. Table 4 also summates imaging findings. According to chest x-ray and computed tomography (CT), 23·6% and 18·9% had normal results, respectively. Patchy lesions were observed in 105 of 501 patients on chest radiography and bilateral ground glass opacities were the most frequent CT abnormality.

Complete blood counts were the most common laboratory results described (see Table 5). Overall, leukocytes were within normal values (7·1 × 10^{/μL), whereas neutrophils were mildly decreased (44·4%) while lymphocytes were marginally elevated (39·9%). Markers of liver and renal function were normal. Four serum inflammatory markers were above the mean: D-dimer, procalcitonin, creatine kinase, and interleukin-6.}

Sixty-six studies (n=614 individuals) provided information regarding treatments. Interferon was the most commonly administered drug (41·0%), followed by empiric antibiotics (20·2%). Of note, glucocorticoids, and intravenous immunoglobulin was used in 4·1% and 3·1% of patients, respectively. Complications we evaluated were rare and only described in 21 studies. There were 7 cases of kidney failure (0·09%), 19 cases of shock (0·24%), and 42 children were intubated (0·54%). More details on treatments provided and complications can be found in Table 6.

Eleven patients (0·14%) met the CDC's criteria for MIS-C [7]. Compared to control (n=14), children with severe inflammation were more likely to present with dyspnea (72·7% vs 28·6%), vomiting (45·5% vs. 7·1%), and diarrhea (45·5% vs. 21·4%). White blood cell counts were comparable between the groups; however, patients with MIS-C have significant lymphopenia (11·1% vs. 41·8%). No difference was noted in platelets or liver function markers. Serum lactate dehydrogenase and D-dimer were higher in children with MIS-C (p<0·05, details provided in Table 7). Also, patients with MIS-C had lower expression of circulating CD16^{CD56 natural killer cells. Imaging findings and treatments were comparable in MIS-C and non-MIS-C patients.}

Discussion

Over the last 6 months, there have been over 6·4 million worldwide cases of SARS-CoV-2 infection and our knowledge of the disease and its epidemiologic and clinical characteristics continue to evolve [4]. However, since it was first reported in Wuhan city in December 2019, most studies have focused on symptomatic adults. In the presence of this rapidly emerging, novel infection, identification of clinical and laboratory characteristics in the pediatric population is essential to guide clinical care, predict disease severity, and determine prognosis. In this context, we performed the largest and most comprehensive systematic review of published studies involving pediatric patients with known COVID-19. Our systematic review summarized the clinical, laboratory and radiologic features of COVID-19 in neonates, children, and adolescents.

Our review also supports the findings by a recent systematic review by Castagnoli et al. [17] Their study included a total of 1,065 COVID-19 infected children and concluded that, by and large, the prognosis for children was excellent, demonstrated by only one death. Compared to that review and other COVID-19 pediatric systematic reviews, [[18], [19], [20], [21]–21] this manuscript has several key advantages: (1) we summarize 131 studies that includes 7780 children from 26 different countries, (2) this report synthesizes underlying pediatric medical conditions and delineates bacterial and viral coinfections, (3) we quantitatively describe clinical symptoms and imaging findings, (4) herein, we conglomerate the mean and standard deviation of frequently used laboratory analytes in COVID-19 positive children, (5) our report presents antiviral therapies by specific agents, and (6) our systematic review offers a preliminary comparison of patients with/without MIS-C.

Although SARS-CoV-2 infection was first identified in China, the United States has now amassed the highest number of confirmed cases [18]. Calculations made on June 4th, 2020 from the COVID-19 Dashboard by the Center for

Systems Science and Engineering at Johns Hopkins University indicate that China has 4·5% of total confirmed COVID-19 cases compared to the United States [4]. As expected, the most common vector for childhood infection is close contact to an affected family member or residing in an area with a high population of cases. Our findings align with the results of an April 2020 report by Dong et al, in which there was a clear trend that the disease spread rapidly from a Chinese province to surrounding provinces and cities in children from December to February [22]. Furthermore, Qiu and colleagues studied 36 pediatric COVID-19 positive patients in which ten patients (28%) were asymptomatic latent cases identified secondary to an adult family member who was infected, symptomatic, or traveled to an endemic area [23]. This lends concern that children, who may be asymptomatic, may play a role in community transmission of the virus.

Results from this systematic review echo findings describing milder symptoms in pediatric cases of SARS-CoV-2 infection [17,21]. For instance, the most common clinical manifestations we found were fever (59·1%), cough (55·9%), rhinorrhea (20·0%) and myalgia/fatigue (18·7%). Unlike adults, children rarely progressed to severe upper respiratory symptoms requiring intensive care unit admission [24,25]. Although transmission rates for SARS-CoV-2 are high, symptoms are less severe than SARS/Middle East Respiratory Syndrome (MERS) infection [26].

Serum inflammatory markers, specifically D-dimer, procalcitonin, creatine kinase, and interleukin-6, were consistently abnormal in the studies included in this review. Alterations to acute-phase infection-related biomarkers are corroborated in adult case series and meta-analyses [27,28]. However, we must take caution when interpreting these outcomes and await more robust, longitudinal laboratory analyses. Again, these blood analyses are non-specific and may merely represent a pro-inflammatory state induced by the virus [26].

In terms of imaging findings, we found that most patients had normal chest x-rays, a finding that is not surprising as most pediatric patients did not present with respiratory symptoms. Paralleling this review, a meta-analysis of CT features for COVID-19, showed that diffuse bilateral ground-glass opacities were the most common finding at all stages of disease [29,30]. Despite these promising associations, it is important to consider that radiologic manifestations from various pathogens may have a similar impression and should be ruled out. Co-infections with other respiratory illnesses including influenza and mycoplasma were described in 72 patients. As elegantly described by Cox and colleagues, most fatalities from the 1918 influenza outbreak were secondary to bacterial infection [31]. Thus, future reports should not only describe coinfections but also detail pertinent negatives. At present, our study had a low rate of reporting the infectious workup (26·7) of patients. Illustrating the importance, one of two patients that died in the study by Shekerdemian et al was due to gram negative sepsis in a child with comorbidities who developed end organ failure [32].

Although most children have an uneventful course, a present concern is an inflammatory cascade in pediatric patients with COVID-19 [8,9]. Clinical presentation includes an unremitting high fever, and includes systemic signs such as rash, conjunctivitis, and/or gastrointestinal symptoms. The case series of eight children from London required respiratory assistance, whether it was oxygen support (n=1), noninvasive ventilation (n=2) or intubation and mechanical ventilation (n=4) [8]. One patient was so ill that he required mechanical ventilation and extracorporeal membrane oxygenation. In addition, all required vasopressor support and demonstrated elevated levels of ferritin, D-dimers, troponin, procalcitonin, and C-reactive protein (CRP). Additionally, cardiac imaging showed ventricular dysfunction in five children. In another article, Italian investigators describe ten patients with MIS-C. Correspondingly, they describe patients manifesting with fever, diarrhea (n=6), and abnormal echocardiograms (n=6). Laboratory specifics showed elevated CRP, lymphopenia, thrombocytopenia, and elevated ferritin levels [9].

We found evidence of MIS-C features in 11 children who also presented with fever (n=11), dyspnea (n=8), and diarrhea (n=6). According to Riphagen and Verdoni, lymphopenia was marked in our cohort of patients, as well as increased levels of lactate dehydrogenase, CRP and D-dimer [8,9]. Despite low numbers we did observe an interesting lower level of CD16^{CD56 natural killer (NK) cells in patients with MIS-C. Both lymphopenia and a reduced number/activity of NK cells in adults has correlated with a more severe COVID-19 disease progression [}[33], [34], [35]–36].

Little is known about the perinatal aspects of COVID-19, and there have been several reported cases of neonatal infection, suggesting a possible perinatal or vertical transmission during pregnancy [37]. However, in a report by Chen et al., all nine neonates born to COVID-19 positive mothers tested negative for the virus after cesarean delivery [38]. In another study by Zhang et al., 10 neonates from COVID-19 positive mother all tested negative for the infection [39]. Moreover, this is further supported by analysis of breast milk and placental pathologic specimens from COVID-19 positive mothers, which have returned negative for the virus [40,41]. Lastly, vertical transmission was not observed with either SARS-CoV-1 or in MERS-CoV;[41] therefore, it is unlikely that maternal vertical transmission during third trimester occurs, or is likely very rare. However, from the limited data published, we cannot determine the consequences of SARS-CoV-2 infection in early pregnancy and if it can be transmitted to the fetus and hinder organ development, malformations, growth abnormalities, or even lead to premature labor or spontaneous abortions [42,43]. Also, Dong et al communicated an alarming finding in which the proportion of severe and critical cases were higher in neonates when compared to the >16-year-old age group (10·6% vs. 3·0%) [44]. As a community, we must stay vigilant, practice social distancing, hand wash frequently, and be especially careful with our children who are at potentially higher risk for critical disease (e.g. multiple comorbidities, weakened immune systems, etc.).

There are several limitations to this review. First, many of the included studies were case reports or cases with low patient numbers. Second, the level of evidence for all the studies was low. Next, we unified the laboratory data to mean and standard deviation. There are inherent issues when using averages including the impact of outliers. We did not include suspected cases, which would allow for a direct comparison of symptoms, labs, imaging, and outcome data. Of concern, many of the studies were incomplete and did not include a comprehensive picture of the patients. Future studies should not generalize data (“CBC was normal”), or categorize laboratory values (i.e., number of patients with elevated CRP), or group therapies (i.e., patient received “antiviral therapy”), or display aggregate data between adults and children. If feasible, divide the symptoms, laboratory markers, and imaging characteristics by children vs. adults. A better understanding of COVID-19 requires access to data, even if it is provided in the appendix or supplementary section of the article. In this way, we will be able to identify the best biomarkers that can stratify disease severity and potential short- and long-term outcomes. Another limitation, is that we had a small number of patients that fit the criteria for MIS-C. Reasons for the small number of patients includes a lack of reporting all of the signs, symptoms, and laboratory markers necessary to make the diagnosis (especially duration of fever). Missing information for laboratory markers (D-dimer, interleukins, and CD%) hinders our preliminary findings. Lastly, the literature focusing on COVID-19 is very dynamic and growing rapidly and we expect the rates, especially for MIS-C, of our outcomes to change.

Contributors

Ansel Hoang-literature search, study design, data collection, data analysis, data interpretation, manuscript writing, risk of bias, tables. Kevin Chorath-literature search, study design, data collection, data interpretation, manuscript writing, risk of bias. Axel Moreira-literature search, study design, data collection, manuscript writing, data interpretation, risk of bias. Mary Evans-data collection, verifying data integrity, risk of bias. Finn Burmeister-Morton-data collection, verifying data integrity. Fiona Burmeister-data collection, verifying data integrity, risk of bias. Rija Naqvi-data collection, verifying data integrity, risk of bias. Matthew Petershack-data collection, risk of bias. Alvaro Moreira-literature search, study design, data collection, data analysis, data interpretation, manuscript writing, figure, tables, oversight.

Appendix. Supplementary materials