Synthesis and systematic review of reported neonatal SARS-CoV-2 infections

A number of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infections have been reported in neonates. Here, we aim to clarify the transmission route, clinical features and outcomes of these infections. We present a meta-analysis of 176 published cases of neonatal SARS-CoV-2 infections that were defined by at least one positive nasopharyngeal swab and/or the presence of specific IgM. We report that 70% and 30% of infections are due to environmental and vertical transmission, respectively. Our analysis shows that 55% of infected neonates developed COVID-19; the most common symptoms were fever (44%), gastrointestinal (36%), respiratory (52%) and neurological manifestations (18%), and lung imaging was abnormal in 64% of cases. A lack of mother–neonate separation from birth is associated with late SARS-CoV-2 infection (OR 4.94 (95% CI: 1.98–13.08), p = 0.0002; adjusted OR 6.6 (95% CI: 2.6–16), p < 0.0001), while breastfeeding is not (OR 0.35 (95% CI: 0.09–1.18), p = 0.10; adjusted OR 2.2 (95% CI: 0.7–6.5), p = 0.148). Our findings add to the literature on neonatal SARS-CoV-2 infections.

-They should detail how many positive cases, how many cases of disease in the epidemiological background.
-All the lines should be supported by one or more scientifically sound references.
-They should include dermatological symptoms and clinical signs.
-They should better define the role played by the risk factors in the occurrence of the disease.
-They should define the prevalence of asymptomatic infection in children and adults.
-The aim is unclear: they should better define the primary objective and two/three secondary objectives.

Methods
-The serological criterion for the infection is disputable based on the reliability of the tests and on the gaps of knowledge in the immunopathogenesis.
-Case-reports or letters to the editors do not provide sufficient data to address the issues -Why did they select the date of the 1st December 2019? -Exclusion criteria should be defined.
-Preprint archives could be a biased source of information: they did not include peer-reviewed articles and, then, the reliability is poor.
-How did they define the timing of the infection? -Which outcomes did they record? -To assess the quality of the case-reports they could use an ad hoc tool published in the BMJ.
-The statistical plan is inappropriate. It is generic and not tailored on what they want to prove.

Results
-References should be used for every finding.
-Characteristics of the studies should be summarized.
-When they describe the results, proportions or summary estimates for continuous variables should be adopted (e.g., symptoms).

Reviewer #2 (Remarks to the Author):
This manuscript entitled "Neonatal SARS-CoV-2 infections: systematic review, synthesis and metaanalysis of reported cases" by Dr De Luca et al is a descriptive meta-analysis of the published case series and case reports of neonates with COVID-19. They identified 117 infected newborns.
Overall, it is a well-written study that tries to describe clinical presentation and the risks factors to acquire the infection in the neonatal period Comments: In the Result section Page 7 and Table 2 It would be helpful if the authors could provide additional information, if available, regarding the 36 newborns admitted to the NICU in gestational age sub-categories, e.g., moderate preterm (32 to 33 weeks) and late preterm (34 to 36 weeks). Similarly, it would be helpful if the authors could provide data on the number of neonates in birthweight sub-categories, e.g., normal (≥2500g), low (1500-2499g), very low (1000-1499g), and extremely low (<1000g) and the admission diagnoses and if COVID-19 related or due to prematurity or other causes Page 8, paragraph 2 It would be important to understand if correct infection control precautions were undertaken by mothers (mask, hand hygiene, etc) when near the newborn or infection occurred mainly because of lack of the above practices.
In the Discussion section Page 9 Even though from these reported cases there is a high incidence of post-natal transmission, there is no description if precautions were undertaken. There is now new published data regarding the safety of rooming-in and against mother-newborn dyad separation if precautions are undertaken. The new American Academy of Pediatrics Guidelines and a recent study published online July 23, 2020 https://doi.org/10.1016/S2352-4642(20)30235-2 in Lancet Child and Adolescent Health are supporting not to separate the mother and newborns. I would revise this section of the discussion according to these new findings and recommendations and add the 2 references. As it is now, the discussion seems favoring the separation of the dyad and could give the wrong message Reviewer #3 (Remarks to the Author): I agree with the first reviewer regarding the assessment of study quality. The authors have ignored major components of the systematic review. The systematic reviews aim to assess the quality of included articles to disclose the risk of bias and conclude the level of evidence. The concluded level of evidence in systematic reviews is an important source for both future research and clinical recommendations. Therefore, I suggest assessing the study quality and report the results inside the text. The statistical analysis is performed well, but the authors did not include the Forest plot. I highly recommended the authors to include the forest plot. C4.-All the lines should be supported by one or more scientifically sound references. A4. We especially thank the Reviewer for this comment. Originally, we did not insert a citation for every single important assumption in the intro for brevity. However, the Reviewer is totally right and now we have inserted several references, paying attention to slightly rephrase according to the very last knowledge on COVID-19 physiopathology and giving preferences to large manuscripts and meta-analysis published in Nature or other major journals. C7.-They should define the prevalence of asymptomatic infection in children and adults. A7. We thank again the Reviewer as this gives us the possibility to update the intro with the most recent and high-quality data on the subject. We have now rephrased this part highlighting that, according to recent studies in NEJM and Lancet the incidence of asymptomatic infections seems lower in children (16-22%, references 12 and 13) than in adults (40-45% as we reported above, see references 3-4). However, these data must be considered preliminary as the knowledge on pediatric SARS-CoV-2 infection is still limited and we lack of data coming from longitudinal large cohorts (as our EPICENTRE that is still ongoing -see reference 10). These are available only in adults (and have been meta-analyzed in reference 4). As such, the limited pediatric knowledge contributed to the uncertainty about neonatal infections as well, and represents the background for our work. Thanks to the Reviewer, we hope that this rephrased part is now guiding better the reader directly towards our study aims.
C8. The aim is unclear: they should better define the primary objective and two/three secondary objectives. A8. Thanks to the Reviewer's suggestions and consistently with point C7, this part has been rephrased and is now more understandable. We had two objectives: the primary was to describe the clinical characteristics and route of transmission of neonatal SARS-CoV-2 infections (this has never been done, since the beginning of pandemics). Our secondary objective was to clarify the effect of mother-neonate separation and breastfeeding on the incidence of neonatal infections acquires postnatally. Added in the Intro, pag.4,1 st par.

Methods
C9. -The serological criterion for the infection is disputable based on the reliability of the tests and on the gaps of knowledge in the immunopathogenesis. A9. We agree in principle with the Reviewer: in fact, the uncertainty on neonatal COVID-19 constitutes the background for our work (see also point C7). However, for the purpose of our analysis, this point has no relevance since only 1 neonate was diagnosed by having positive IgM only and negative PCR on nasopharyngeal swabs (Dong et al, JAMA 2020, reference 32). According to the classification of neonatal and perinatal SARS-CoV-2 infections, this case fully qualifies as "possible congenital infection in a live born neonate" as there was "No detection of the virus by PCR in nasopharyngeal swab at birth (collected after cleaning baby) BUT presence of anti-SARS-CoV-2 IgM antibodies in umbilical cord blood or neonatal blood collected within first 12 hours of birth or placental tissue" (classification in reference 17, pag.566) All the other cases were diagnosed as infected based on at least one positive PCR.
Thus, following Reviewer's suggestion, we specifically described this case with this peculiarity in results, pag.8, 3 rd par and also recalled this in discussion, study limitations, pag.12. We are willing to totally exclude the unique patient with positive IgM if required by the Reviewer or the Editor.
C10.-Case-reports or letters to the editors do not provide sufficient data to address the issues. A10. We respectfully disagree and we believe that there is a misunderstanding here. While the vast majority of meta-analyses are based on RCTs or large observational studies, these studies cannot be used for the purpose of our work for several reasons: 1. There is no RCT and no large observational study focused on neonatal SARS-CoV-2 infection. This is because pandemics is a quickly evolving situation and because COVID-19 is a disease affecting much more older adults than children (thus there are few cases in newborn infants). As such, all our knowledge about neonatal SARS-CoV-2 infections and COVID-19 comes from case reports and series. 2. Case reports and series, if well prepared, do provide all the relevant informations to address our objectives (that is, primarily to describe clinical characteristics and route of transmission of neonatal SARS-CoV-2 infections and, secondarily, to clarify the effect of mother-neonate separation and breastfeeding on the incidence of late, environmentally-acquired neonatal infections). In fact, similar works have been already performed with similar primary aims in older children ( intervention, compared to a control, on the outcome in a given population (PICO questions methodology). This is an objective completely different from ours, as there is no intervention possible for infected neonates, because we are in a much earlier knowledge phase and we need to understand their clinical characteristics and transmission route first (please see our study purpose, point C8 and also point C16). C12. -Exclusion criteria should be defined. A12. We are extremely grateful to the Reviewer as this was a forgotten point! We thought to have described these implicitly in Fig.1, but we have now clearly added all the exclusion criteria, in methods, pag. 5, 1 st par.

Therefore, the synthesis/meta-analysis of case reports and series, although less common than the meta-analysis of RCTs and observational studies, is a particularly interesting technique for a relatively rare condition lacking of other studies (such as neonatal SARS-CoV-2 infection). An example of the same situation may be found in the recently published meta-analysis of case reports on the epidemiology and clinical manifestations of mucormycosis (Jeong W et al, Clin
C13. -Preprint archives could be a biased source of information: they did not include peer-reviewed articles and, then, the reliability is poor. A13. We fully understand the Reviewer's point of view and we are grateful as this is a relevant issue!. This was considered important by the Editor as well and we want to immediately say that we did some additional analyses and added their results and related comments to the text following Reviewer's point of view (see below).
However, we believe that the role of preprint has rapidly changed during the pandemics. Please consider the following: 1. Preprints facilitate the spread of potentially important information that may be extremely useful during a rapidly evolving situation such as a pandemic new disease for which clinicians are unarmed and lacking knowledge. The use of preprint has exploded in recent times and this has been acknowledged by the general media (see for instance: https://cen.acs.org/policy/publishing/Pandemic-puts-preprints-first/98/i22) and by academic forums (see: https://rapidreviewscovid19.mitpress.mit.edu/pub/k8h2xox0/release/1). 2. This has also been recognized by a specific analysis of preprints on COVID19 (20)30113-3 -reference 97). 3. Nature Communications, as all Nature family journals, has its own preprint server and authors are encouraged to post submitted articles there (see: https://www.nature.com/nature-research/editorialpolicies/preprints-and-conference-proceedings). We did so for the present work and a previous one recently published. Obviously, the existence of a preprint server coupled with a high impact journal is a further guarantee of quality. 4. Beside Nature family journals, other major scientific publishers facilitate journals editors and preprint archives to work together in order to "complement traditional journal publishing, adding speed, openness, and faster feedback for researchers" (see for instance: John Wiley and Sons Publishing: https://www.wiley.com/network/archive/preprints-publishing-and-a-pandemic-your-questionsanswered). 5. According to the 2016 Statement on data sharing in public health emergencies, (https://wellcome.ac.uk/what-we-do/our-work/statement-data-sharing-public-health-emergencies) many healthcare systems, publishers and other bodies agreed to share still unpublished, under review data, and currently all the major journals are sending manuscripts on COVID-19 to the WHO in form of preprint in order to diffuse the knowledge as quicker as possible. 6. For the sake of our work, we should consider that we provided the first intention peer review for cases series/reports available only as preprint at the time of writing. In fact, they have been independently evaluated by two investigators (see methods, study selection and data collection process, pag.5-6) using a specifically dedicated scoring system (Mayo Evidence-Based Practice Centre -Murad MH et al, BMJ Evid Based Med 2018 -reference 22) and following the CARE guidelines (reference 21). The same tool and criteria have also been used for the already published articles (thus all articles have been evaluated in the same way).
Anyway, this is no more a very relevant issue because, since our first submission, many preprint articles have been already peer reviewed and finally published (we have updated the reference list); only two articles remain on preprint archives.
Having said that, we believe that this is an important issue and we acknowledged it as study limitation (Discussion, study limitations, end of pag.12) and cited the preprint analysis published in Lancet Global Health 2020 (reference 97). We also explicitly declared which manuscript was a preprint and which one has been peer reviewed and published (Tab.1).
Finally, to be extra sure about the quality of our results, in relation to the quality of reviewed articles, we performed multivariate logistic regressions, with enter method, adjusting for the quality of reviewed article (evaluated by the specific tool (Mayo Evidence-Based Practice Centre -Murad MH et al, BMJ Evid Based Med 2018 -reference 22). The regressions had the incidence of late onset neonatal infections as dependent variable, and the mother-neonate separation or breastfeeding as independent variables. Adjusted OR were consistent with crude OR and our results did not change. This is now described in Methods, additional analyses, pag.8, 1st par and at the end of results, end of pag.9.
C14. -Explanation of the categories of the likelihood of infection should be provided. A14. These criteria are published elsewhere (reference 17) but we have now better described this in Methods, data items, pag.6-7 and the detailed criteria for each likelihood of infection category are listed in a supplementary material. This allows the reader to understand the transmission route without the need to retrieve the original publication describing the classification system (see also point C15 below).
-C15. How did they define the timing of the infection? A15. The timing of infection has been defined using the strict criteria provided by the Classification system and case definition for SARS-CoV-2 infections in fetuses and neonates published in Shah, PS, et al. Acta Obstet Gynecol Scand 2020 (reference 17).
Therefore, the timing was considered as congenital (intrauterum transmission before the delivery), intrapartum (transmission at the delivery) or postpartum (after at least 48h from birth); all infections had to be apparent within the first 30 days of life (neonatal period) as detailed in methods, eligibility and exclusion criteria, end of pag.4). All these criteria are now described in details in the supplementary material, reporting the original criteria listed from reference 17. This allows the reader to understand the transmission route and timing without the need to retrieve the original publication describing the classification system (see also point C14 above).
C16.-Which outcomes did they record? A16. There is a misunderstanding here. As also explained above (please see point C10), we did a synthesis and meta-analysis of case reports and series. This is very different from the meta-analysis of RCTs or observational studies. RCTs and observational studies report an "intervention" in a given patients' population (following randomization or not) and the effect of that intervention is evaluated by a given outcome, which is measured both in treated patients and in comparators (i.e.: in an untreated, or differently treated, patients' population). This is resumed in the so-called PICO (patient-intervention-comparator-outcome) questions acronym.
A meta-analysis of case reports/series is different because these studies usually do not have an intervention and, by definition, cannot have an external comparator. Moreover, in our case, we could not have an intervention since neonatal SARS-CoV-2 infection does not have any specifically validated or even proposed therapy.
In fact, all meta-analyzed case reports/series did not focus on a therapy but rather on the clinical characteristics and route of transmission of neonatal infections, whose description represents our primary objective (please see point C8). Thus, we did not analyze an intervention and could not have an intervention, nor an outcome potentially modified by that intervention. An example of the same situation may be found in the recently published meta-analysis of case reports on the epidemiology and clinical manifestations of mucormycosis (Jeong W et al, Clin Microbiol Infect 2019doi.org/10.1016/j.cmi.2018.07.011). However, we wanted to increase readability of our manuscript, especially because the message should be understandable to any reader during this worldwide emergency situation. Therefore, following the Reviewer's suggestion we better specified our outcome, which correspond to our study objectives. This is now better specified in Methods, data items, pag.7, 1 st par.
C17.-To assess the quality of the case-reports they could use an ad hoc tool published in the BMJ. A17. We already did so. We used the Mayo Evidence-Based Practice Centre tool, which is the specific tool dedicated to the evaluation of case report/series quality (published in Murad MH et al, BMJ Evid Based Med 2018 -reference 22). This is exactly the same tool suggested by the Editor. This tool has been already used in similar meta-analyses of case reports/series (for ex.: Bazerbachi F, et al. Gastroenterol Rep 2017).