Dear Editor,
Currently research on COVID-19 has been prioritised globally with a high frequency of the articles being published in the literature. The incidence of false-negative tests for ‘happy hypoxia’ in asymptomatic patients, evolving epidemiological characteristics, and the universal risk of infection in every single age group with unpredictable capricious outcomes, explains the need to explore the risk factors associated with the novel coronavirus. The systematic review and meta-analysis on COVID-19 by Zheng et al. contribute to this global research output, and attempts to inform clinical decision making during this crisis.1 Nevertheless, there are a few recommendations that we would like to put forth in order to make the study robust enough to inform future decision making.
Few additional risk factors missing in the study
The epidemiology and clinical pattern of paediatric COVID 19 has a unique spectrum with infants and young children ≤ 5 years more likely to succumb to severe clinical symptoms of COVID19 than older children (i.e., ≥ 6 years). The immaturity of the immune system is cited as a plausible explanation.1 Children can swiftly progress to acute respiratory distress syndrome (ARDS). They may also have shock, encephalopathy, myocardial injury or heart failure, coagulation dysfunction, and acute kidney injury.
The forgotten “Cancer”
The unprecedented outbreak of COVID-19 has caused a substantial risk for cancer patients who are immunocompromised due to the disease and its treatment.2
Fixed effect model of meta-analysis
We noted that the authors used the fixed-effect model to perform the meta-analysis. The fixed-effects model is practically applied when the studies across which the data is being pooled have, similar study parameters, and are devoid of any major heterogeneity. Hence, this model is often used to assess cohorts within a singular larger study. Nevertheless, the meta-analysis by Zhang et al., where multiple different studies are used, are inherently predisposed to heterogeneity due to the differential protocols and parameters of individual study. We propose the “random effects model” for this study.3 , 4
Analysis of heterogeneity
Furthermore, although the analysis of heterogeneity was conducted using the commonly used I2 statistic, it may not have sufficient power by itself to determine between-study heterogeneity. The I2 statistic has shown to be limited in its application; nonetheless we recommend the authors also asses heterogeneity via the Cochran's Q statistic and the Tau2 statistic, which would add redundancy and robustness in the analysis of heterogeneity in this study.5
Publication bias
We also note that the authors fail to perform an analysis of publication bias as a mandatory application towards a mutable research subject where the bias of publication is likely to exist. Furthermore, systematic review and meta-analysis guidelines mandate the assessment of publication bias. Hence an analysis of publication bias also proposed for this study in order to achieve concrete results that can impact the clinical decision making.6 We recommend the Eggers bias indicator test for lucid graphical assessment of publication bias.7
Acknowledgments
Competing Interest
The authors confirmed that they have no competing interests.
Funding
Any external source did not fund this study.
Authors’ contributions
RJ predominantly conceived this review and led the development of the letter to the editor. RJ, SSS and CK wrote the first draft of the letter, and GR, SSS, RRM, and PS critically revised and edited successive drafts of the manuscript. All authors read and approved the final version of the manuscript.
References
- 1.Zheng Z., Peng F., Xu B. Risk factors of critical & mortal COVID-19 cases: A systematic literature review and meta-analysis. Journal of Infection. 2020 doi: 10.1016/j.jinf.2020.04.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dong Y., Mo X., Hu Y. Epidemiological characteristics of 2143 pediatric patients with 2019 coronavirus disease in China. Pediatrics. 2020 [Google Scholar]
- 3.Hedges L.V., Vevea J.L. Fixed-and random-effects models in meta-analysis. Psychological methods. 1998;3(4):486. [Google Scholar]
- 4.Kumarasamy C., Madhav M.R., Sabarimurugan S. Prognostic Value of miRNAs in Head and Neck Cancers: A Comprehensive Systematic and Meta-Analysis. Cells. 2019;8(8):772. doi: 10.3390/cells8080772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jayaraj R., Kumarasamy C., Sabarimurugan S., Madhav M.R. Meta-analysis of penile cancer: conceptual interpretations. The Lancet Oncology. 2019;20(3):e125. doi: 10.1016/S1470-2045(19)30023-3. [DOI] [PubMed] [Google Scholar]
- 6.Jayaraj R., Kumarasamy C. Conceptual, statistical and clinical interpretation of results from a systematic review and meta-analysis of prevalence of cervical HPV infection in women with SLE. Autoimmun Rev. 2019;18:433–434. doi: 10.1016/j.autrev.2018.12.003. [DOI] [PubMed] [Google Scholar]
- 7.Egger M., Smith G.D., Schneider M., Minder C. Bias in meta-analysis detected by a simple, graphical test. Bmj. 1997;315(7109):629–634. doi: 10.1136/bmj.315.7109.629. [DOI] [PMC free article] [PubMed] [Google Scholar]