key: cord-0860340-9hdft1tc authors: Kim, M. S.; An, M. H.; Kim, W. J.; Hwang, T.-H. title: Comparative efficacy and safety of pharmacological interventions for the treatment of COVID-19: A systematic review and network meta-analysis of confounder-adjusted 20212 hospitalized patients date: 2020-06-19 journal: nan DOI: 10.1101/2020.06.15.20132407 sha: 792945cefe7a2f8bb4f6272fd1c4dfc067415d7e doc_id: 860340 cord_uid: 9hdft1tc Objective: To evaluate the comparative efficacy and safety of pharmacological interventions used in treating COVID-19 and form a basis for an evidence-based guideline of COVID-19 management by evaluating the level of evidence behind each treatment regimen in different clinical settings. Design: Systematic review and network meta-analysis Data Sources: PubMed, Google Scholar, MEDLINE, the Cochrane Library, medRxiv, SSRN, WHO International Clinical Trials Registry Platform, and ClinicalTrials.gov up to June 9th, 2020. Study Selection: Published and unpublished randomized controlled trials (RCTs) and baseline-adjusted observational studies which met our predefined eligibility criteria. Main Outcome Measures: The outcomes of interest were mortality, progression to severe disease (severe pneumonia or admission to intensive care unit (ICU)), time to viral clearance, QT prolongation, fatal cardiac complications, and non-cardiac serious adverse events. The level of evidence behind each outcome was also measured using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework. Results: 49 studies with a total of 20212 confounder-adjusted patients were included for analysis. The risk of progression to severe pneumonia or ICU admission was significantly reduced with tocilizumab (GRADE low), anakinra (GRADE very low), and remdesivir (GRADE high) compared to standard care. Tocilizumab was shown to reduce mortality rate for both moderate-severe patients in the non-ICU setting at admission (Odds ratio (OR) 0.31, 95% confidence interval (CI) 0.18 to 0.54, GRADE low) and critically ill patients in the ICU setting (OR 0.67, 95% CI 0.50 to 0.91, GRADE low). High dose IVIG reduced death rate (GRADE low) while corticosteroids increased mortality for critically ill patients (GRADE moderate). Convalescent plasma and hydroxychloroquine were shown to promote viral clearance (OR 11.39, 95% CI 3.91 to 33.18, GRADE low and OR 6.08, 95% CI 2.74 to 13.48, GRADE moderate, respectively) while not altering mortality or progression to the severe courses. The combination of hydroxychloroquine and azithromycin was shown to be associated with increased QT prolongation incidence (OR 1,85, 95% CI 1.05 to 3.26, GRADE low) and fatal cardiac complications in cardiac-impaired populations (OR 2.26, 95% CI 1.26 to 4.05, GRADE low). High-dose (>600mg/day) hydroxychloroquine monotherapy was significantly associated with increased non-cardiac serious adverse events (GRADE moderate). Conclusion: Anti-inflammatory agents (tocilizumab, anakinra, and IVIG) and remdesivir may safely and effectively improve outcomes of hospitalized COVID-19 patients. Widely used hydroxychloroquine provides marginal clinical benefit in improving viral clearance rates whilst posing both cardiac and non-cardiac safety risks, especially in the vulnerable population. Only 20% of current evidence on pharmacological management of COVID-19 is on moderate and high evidence certainty; remaining 80% are of low or very low certainty and warrant further studies to establish firm conclusions. Systematic Review Registration: PROSPERO 2020: CRD42020186527. Section 1: What is already known on this topic -Numerous clinical trials and observational studies have investigated various pharmacological agents as potential treatment for COVID-19. -Results from numerous studies are heterogeneous and sometimes even contradictory to one another, making it difficult for clinicians to determine which treatments are truly effective. -Level of evidence behind each outcome from diverse studies remains unknown. Section 2: What this study adds -Anti-inflammatory agents (tocilizumab, anakinra, and IVIG) and remdesivir may safely and effectively improve clinical outcomes of COVID-19. -Widely used hydroxychloroquine provides marginal clinical benefit in improving viral clearance rates whilst posing both cardiac and non-cardiac safety risks. -Only 20% of current evidence on pharmacological management of COVID-19 is on moderate/high evidence certainty and can be considered in practice and policy; remaining 80% are of low or very low certainty and warrant further studies to establish firm conclusions. 5 may safely and effectively improve outcomes of hospitalized COVID-19 patients. Widely used hydroxychloroquine provides marginal clinical benefit in improving viral clearance rates whilst posing both cardiac and non-cardiac safety risks, especially in the vulnerable population. Only 20% of current evidence on pharmacological management of COVID-19 is on moderate and high evidence certainty and can be considered in practice and policy; remaining 80% are of low or very low certainty and warrant further studies to establish firm conclusions. Systematic Review Registration: PROSPERO 2020: CRD42020186527. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 6 A large registry-based study investigating the effect of hydroxychloroquine on COVID-19 patients reported surprisingly high mortality rate and ventricular arrhythmia incidence with hydroxychloroquine use 1 , and this paper contributed to the decision of the World Health Organization (WHO) to pause all ongoing trials on hydroxychloroquine due to safety concerns. Although this paper was retracted after concerns raised by clinicians and scientists, the delay of trials and confusion caused by the paper were unavoidable. This was an alarming event that due to the sudden advent of COVID-19 and global urgency to find treatments, many clinical and observational studies are of suboptimal quality 2 ; therefore, it may be unreliable to make decisions grounded on the evidence of a single paper. Prospective metaanalyses synthesizing multiple studies using predefined eligibility criteria can be an interim solution to generate reliable conclusions 2 and protect from ill-informed changes in practice and policy. Numerous COVID-19 clinical trials are underway, and over 31 pharmacological agents and combinations have been investigated as potential treatments of COVID-19 to date. Network meta-analysis (NMA) is an analytical tool that enables the a single coherent ranking of such numerous interventions, and it can thus aid decision-makers who must choose amongst an array of treatment options 3 . We conducted the first network meta-analysis with selective predefined eligibility criteria for both published and unpublished data, and investigated 31 treatment regimens for comparative efficacy and safety. We incorporated 49 studies (16 randomized controlled trials (RCT) and 33 baseline-adjusted observational studies) including a total of 20212 confounder-adjusted patients. The level of certainty behind the evidence for each outcome was evaluated to assist the decision-making of clinicians and policy makers. This study will serve as the basis for an individual patient data (IPD) network meta-analysis that we are designing as a future study. We searched PubMed, Google Scholar, MEDLINE, the Cochrane Library, medRxiv, SSRN, WHO International Clinical Trials Registry Platform, and ClinicalTrials.gov for RCTs and observational studies that evaluated treatment responses to pharmacological management in COVID-19 patients, from inception to June 9 th , 2020. Reference lists of review articles were also reviewed to search for additional articles that may not have been retrieved by the 7 prespecified searching strategy. We had no restriction on language, but all included studies were written in English. We contacted principal investigators of unpublished studies identified in trial registries and regulatory submissions to obtain unpublished data. Inclusion of unpublished data in NMAs is not uncommon 4-11 and reduces risk of selection and publication bias while increasing the density of study data. This is especially beneficial in the study of COVID-19 as all data were generated relatively recently, and the bulk of the relevant data are still in the unpublished stages. Pre-prints have been used in meta-analysis relatively frequently for the urgent topic of COVID-19 [9] [10] [11] [12] , and American Gastroenterological Association (AGA) recently published a management guideline for the gastrointestinal manifestation of COVID-19 patients based on the result of meta-analysis incorporating pre-prints 11 . We contacted authors of included preprints from medRxiv and SSRN, and any change in the results was updated. We included both RCTs and baseline-adjusted observational studies; the rationale is that inclusion of real-world data from non-randomized studies has the potential to improve precision of findings from RCTs if appropriately integrated 13 and that the volume of information provided by these studies is necessary to assess adverse events of low to moderate incidence [14] [15] [16] . As observational studies are more vulnerable to bias, we included only the studies that adjusted for relevant confounding variables through methods such as propensity score matching (PSM), inverse probability treatment weighting (IPTW), or regression model adjustment. Studies providing evidence that the risk for such confounding was low by establishing baseline similarity between the groups also met inclusion criteria (Appendix p5). Following studies were excluded: studies without a proper control group; studies of children or adolescents (<18 years); observational studies with significant differences in baseline characteristics between groups and did not perform adequate adjustments; studies investigating the effect of medication initiated prior to the diagnosis of COVID-19 (e.g. ACEi/ARB for hypertensive patients). The study search and data extraction were independently conducted by 3 authors (MS Kim, MH An, and WJ Kim). Manuscript and supplementary materials of the included studies were reviewed for relevant information which was extracted according to a pre-specified protocol. Any discrepancy or ambiguity in this process was resolved by discussion. Authors of certain included studies were contacted in case of missing or unclear information. Non-randomized studies were qualitatively assessed using the Newcastle-Ottawa Scale (NOS) 17 , and RCTs were assessed with the Jadad scale 18 . All studies were assessed for risk of bias using the RoB 2 tool for randomized studies and ROBINS-I tool for nonrandomized studies 19 . The quality of evidence of collective outcomes were estimated using the Grading of Recommendations 8 Assessment, Development, and Evaluation (GRADE) framework 20 . A comparison-adjusted funnel plot with Egger's test was constructed to assess for publication bias 21 . Control groups consisted of patients who received standard care or placebo. Patients who received hydroxychloroquine or corticosteroids were subdivided according to the dosage they received. For hydroxychloroquine, most studies reported 400mg hydroxychloroquine daily for maintenance, and this was considered the standard prescription; patients who received daily maintenance dosage of over 600mg hydroxychloroquine were classified into a separate high-dose hydroxychloroquine group. For corticosteroids, average daily dosage of 40mg methylprednisolone (or equivalent) was regarded as the standard dosage, while 1-2mg/kg/day methylprednisolone (or equivalent) was regard as high dose. 1mg methylprednisolone was considered equivalent to 0.1875mg dexamethasone and 5mg hydrocortisone. A critically ill patient was defined as a patient who received invasive mechanical ventilation or needed intensive care in the ICU before or soon after beginning the treatment of interest, while moderate-severe patients were defined as patients hospitalized in a non-ICU setting at admission. The mortality rate of patients included in our mortality analyses were 11.7% for moderate-severe (non-ICU) patients and 38.6% for critically ill (ICU) patients on average. We conducted a random-effects network meta-analysis within a frequentist framework using STATA (Stata Corp, College Station, TX, US, version 15.0) and R (version 3.6.0) software 22 . Direct and indirect (and mixed) comparison were accomplished through the self-programmed routines of STATA 21 23 and the netmeta package of R 24 , as done in our previous work 25 . The effect estimation was in odds ratios (OR) for dichotomous variables and mean difference (MD) for continuous variables, both with 95% confidential intervals (CI). When median (interquartile range) was presented for continuous variables of interest, it was converted to mean (standard deviation) by calculation 26 27 . A two-sided p-value of less than 0.05 was regarded as statistically significant. Statistical heterogeneity was estimated using restricted maximum likelihood method 28 and expressed with Higgins I 2 statistics and the Cochran Q test 29 . The net heat plot was constructed to visualize the inconsistency matrix and detect specific comparisons which introduced large inconsistencies 30 . The rank of effect estimation for each treatment was investigated using the surface under the cumulative rank curve (SUCRA) of P rank score of R 31 . Prespecified subgroup and sensitivity analyses were performed to determine whether the results were affected by the patient severity, treatment protocol, and study design. The primary outcomes were separately analyzed for moderate to severe patients (non-ICU at admission) and critically ill patients (ICU) as these patients may respond differently to 9 treatments. Sensitivity analyses were conducted by restricting the analyses to only RCTs, only published studies, excluding studies with high/serious risk of bias, and excluding studies in which initiation of treatment was over 14 days after symptom onset. Neither any patients nor the public were involved in the design, conduct, and reporting of the research. The study protocol is publicly available on PROSPERO (CRD42020186527) and medRxiv. The initial search identified 5970 articles. These studies were assessed for inclusion using the prespecified inclusion and exclusion criteria described in methods. Title and abstract of 3,626 articles were assessed, and 251 studies were found suitable for full-text review. After excluding 202 studies, 16 RCTs and 33 baseline-adjusted observations studies were finally included in our network meta-analysis ( Figure 1 ). Total of 20212 confounder-adjusted COVID-19 patients were included. Background characteristics and reference list of included studies are presented in the supplementary appendix pp113-170. The risk of bias in included studies were generally low to moderate (Supplementary appendix pp 63-108). For both pairwise meta-analysis and network meta-analysis, the primary outcomes presented no evidence of heterogeneity (Appendix pp . Inconsistency, which represents discordance of direct and indirect comparisons, was also evaluated for outcomes, but none were subject to global inconsistency. The network of eligible comparisons for clinically relevant outcomes are presented in Figure 2 . Detailed information of studies included in the analysis for cardiac adverse events are presented in Table 1 , and the certainty of evidence (GRADE) for each outcome is summarized in Table 2 . Tocilizumab (Odds ratio (OR) 0.31, 95% confidence interval (CI) 0.18 to 0.54, low certainty) and anakinra (OR 0.30, 95% CI 0.11 to 0.80, very low certainty) significantly reduced the mortality in moderate-to-severe patients hospitalized in a non-ICU setting compared to the control group ( Figure 3A ). This effect could not be confirmed in a parallel sensitivity analysis with only RCTs ( Figure 3B ) as no RCTs were conducted for either agent. In critically ill patients hospitalized in the ICU, high dose IVIG (OR 0.13, 95% CI 0.04 to 0.42, low certainty) and tocilizumab (OR 0.67, 95% CI 0.50 to 0.91, low certainty) were shown to lower morality while corticosteroid therapy was shown to increase mortality (OR 2.40, 95% CI 1.02 to 5.61, moderate certainty) ( Figure 3C ). . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . Tocilizumab (OR 0.32, 95% CI 0.16 to 065, low certainty), anakinra (OR 0.22, 95% CI 0.09 to 0.54, very low certainty), and remdesivir (OR 0.31, 95% CI 0.18 to 0.55, high certainty) showed effectiveness in preventing progression to severe courses ( Figure 3E ). Only remdesivir was shown to be effective in the analysis using only RCTs (OR 0.31, 95% CI 0.18 to 0.55) ( Figure 3F ). The use of convalescent plasma (OR 11.39, 95% CI 3.91 to 33.18, low certainty), hydroxychloroquine (OR 6.08, 95% CI 2.74 to 13.48, moderate certainty), and meplazumab (OR 8.67, 95% CI 1.53 to 49.22, very low certainty) showed significantly higher viral clearance rate compared to standard supportive therapy ( Figure 4A ). However, this effect of hydroxychloroquine and meplazumab was not replicated in the analysis of only RCTs ( Figure 4B ). The result was similar when using the continuous variable of time to viral clearance (days) as the outcome measure ( Figure 4C , D). The effect of the timing of hydroxychloroquine treatment initiation after the symptom onset ( Figure 4E ) was assessed. Treatment initiated after 14 days (MD -2.02, 95% CI -7.03 to 3.00) from symptom onset did not reduce the time to viral clearance compared to standard care. Compared to hydroxychloroquine monotherapy, the prolongation of QTc interval after treatment initiation was statistically significantly longer in the hydroxychloroquine plus azithromycin group (MD 20.79ms, 95% CI 12.60 to 28.98, low certainty) ( Figure 5A ). The proportion of patients experiencing QTc prolongation (defined by QTc interval >500ms or Δ QTc >60ms) was also significantly higher in the hydroxychloroquine plus azithromycin group compared to the control group (OR 1.85, 95% CI 1.05 to 3.26, very low certainty) but not in the hydroxychloroquine monotherapy group, azithromycin monotherapy group, or high-dose hydroxychloroquine group ( Figure 5B ). The associations between fatal cardiac complications and hydroxychloroquine, azithromycin, or hydroxychloroquine plus azithromycin therapy were analyzed ( Figure 5C ). Overall, treatment with hydroxychloroquine plus azithromycin showed a significant association (OR 2.25, 95% CI 1.27 to 3.99, low certainty) while others did not. We further subdivided the included studies based on prevalence of coronary artery disease (CAD) and congestive heart . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . 1 1 disease (CHD) at baseline. In studies in which >10% of the baseline population had CAD/CHD, the risk of fatal cardiac complication was statistically significantly higher in patients receiving hydroxychloroquine plus azithromycin. In studies in which <10% of the baseline population had CAD/CHD, no notable difference in incidence of fatal heart complication was observed in any treatment group. High dose (>600mg/day) hydroxychloroquine was associated with increased non-cardiac serious adverse events ( Figure 5D ). In contrast, there was a protective tendency with a decreased rate of adverse events with remdesivir compared to standard care (OR 0.71, 95% CI 0.55 to 0.92, high certainty). The results of our subgroup and sensitivity analysis are reported in the Appendix pp 48-62. The assessments of other specific complications such as nausea/vomiting, diarrhea, hypoalbuminemia, anemia, leukopenia, lymphopenia, elevated AST/ALT, elevated CK, and increase total bilirubin are also presented in Appendix pp 52-54. This is the first network meta-analysis (NMA) of pharmacological treatment for COVID-19. We comprehensively analyzed 31 active pharmacologic agents and their combinations in a large-scale analysis incorporating 20212 confounder-adjusted patients. Our study included unpublished data to integrate recent investigations and avoid selection and publication bias, as done in previous studies 4-7 . We did not limit our inclusions to RCTs and incorporated observational studies as we deemed that, in this analysis, the inclusion of real-world evidence from non-randomized studies has the potential to add validity to certain findings 13 , provide additional information regarding low-to-moderate incidence adverse events [14] [15] [16] , and improve the density of the network 14 . Many previous NMAs included observational studies with this rationale 14-16 32 33 , but inclusion of observational studies to an NMA requires careful integration to avoid biases from these observational studies pervading the meta-analysis 34 ; as such, we exclusively included cohort studies that adjusted for confounders through methods such as propensity score matching (PSM), inverse probability treatment weighting (IPTW), and regression model adjustment or established similarity in the baseline characteristics of the groups being compared so that such adjustments are not necessary or irrelevant. Our conclusions support the use of individualized treatment strategies based on clinical setting and severity. For moderate and severe patients hospitalized in non-ICU settings, . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . tocilizumab and anakinra were shown to reduce risk of progression to severe pneumonia or ICU admission. Both of these selective anti-inflammatory agents also showed survival benefit compared to standard care. Remdesivir was the only antiviral agent shown to prevent progression of disease to severe pneumonia or transfer to ICU, but it did not alter mortality rate for non-ICU patients. For ICU-based critically ill patients, high dose IVIG and tocilizumab may reduce mortality while corticosteroid was associated with increased mortality. Convalescent plasma and hydroxychloroquine, topics of much debate, were not shown to reduce mortality rate or prevent progression to severe disease in our analysis; however, they demonstrated benefit in promoting viral clearance. Our analysis showed that hydroxychloroquine was significantly associated with reduced time to viral clearance ( Figure 4C ). Although this result was not supported by a single RCT on this subject 35 , this RCT should be interpreted with caution due to the median 16 days of delay from symptom onset to the treatment; our analysis indicated that the effects of hydroxychloroquine may fade after 14 days of delay from symptom onset to treatment initiation ( Figure 4E ). It should be noted that hydroxychloroquine was not shown to reduce mortality rate or progression to severe courses. As the level of evidence (GRADE assessment) varies in certainty for these results, further prospective large randomized trials with early initiation of treatment may be warranted to establish firm conclusions. The potential cardiotoxicity of hydroxychloroquine and azithromycin is a widely shared concern in treating COVID-19 with these medications. According to our quantitative synthesis, incidence of QT prolongation was significantly higher in the patients who received hydroxychloroquine plus azithromycin compared to those who received standard care ( Figure 5B ). In addition, this combination of hydroxychloroquine and azithromycin was also associated with increased rate of fatal cardiac complications such as torsades de pointes, cardiac arrest, and severe ventricular arrhythmia in the cardiac-impaired population with a pooled incidence of 12.2%; in comparison, the pooled fatal cardiac complications rate in healthy populations with preserved cardiac function was about 0%. Therefore, the use of hydroxychloroquine/azithromycin should be limited to patients with healthy cardiac function, and monotherapy should be preferred to combination therapy for patients with poor cardiac function ( Figure 5C ). It should also be noted that non-cardiac adverse events were significantly more frequent in high dose (>600mg/day) hydroxychloroquine monotherapy compared to standard care ( Figure 5D ); nausea, vomiting, and diarrhea that required discontinuation of the treatment were more frequent with high dose hydroxychloroquine intake. Strict monitoring should be implemented in all patients receiving hydroxychloroquine with or without azithromycin to maintain a tolerable safety margin. The results of our study also showed the efficacy of remdesivir in reducing the progression of COVID-19 to more severe pneumonia or admission to the ICU ( Figure 3E-F) . This result was supported by a high certainty of evidence (Table 2) and was replicated in the sensitivity analysis that included only RCTs. Interestingly, non-cardiac serious adverse events occurred . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . significantly less in patients who received remdesivir compared to the control group ( Figure 5D ). This may be explained by the preventive effect of remdesivir against progression to severe diseases as numerous clinical trials reported possible consequences of severe disease such as septic shock and acute kidney injury as non-cardiac serious adverse events. Tocilizumab (monoclonal IL-6 receptor antibody), anakinra (IL-1 receptor antagonist), and IVIG were associated with significantly reduced mortality in COVID-19 patients (Figure 3A -D). These three agents are known anti-inflammatory agents that have been conventionally used in hyperimmune or autoimmune conditions; tocilizumab and anakinra have been used for the management of severe rheumatoid arthritis 36 37 and juvenile idiopathic arthritis 38-40 , and IVIG was used for management of Kawasaki's disease 41 42 , inflammatory muscle diseases 43 44 , and sepsis 45 . As there is accumulating evidence for an hyper-immune response characterized by the release of pro-inflammatory cytokines in severe and deceased Covid-19 patients 46-50 , suppression of the inflammatory response and potential cytokine storm with immune-modulatory therapies was proposed as a potential therapeutic target; the results of this network meta-analysis support the efficacy of these treatments. Effectiveness of antiinflammatory agents (tocilizumab, anakinra, IVIG) and ineffectiveness of antiviral agents, except for remdesivir, in hospitalized COVID-19 patients suggest that the management should focus more on the immune response rather than viral mechanism itself. Although numerous studies reported consistent results on beneficial effect of such agents, the certainty of evidence for these agents are either low or very low because conclusions on tocilizumab, anakinra, and IVIG to date are all based on observational studies. Randomized controlled trials on these anti-inflammatory agents are required to confirm these findings and increase the level of evidence. Corticosteroids, on the other hand, were associated with significantly increased risk of morality in critically ill COVID-19 patients ( Figure 3C ). It could be argued that the frequent use of corticosteroids on patients with more severe conditions may have skewed the results against corticosteroid use; however, the studies included in our synthesis have adjusted for confounders for mortality including severity of disease, implicating that the observed unsafe effect of corticosteroids in critically ill COVID-19 should not be neglected. The detrimental effects of corticosteroid on mortality is in line with previous studies that showed the higher mortality rate with adjuvant steroid use in HIN1 influenza-infected critically ill patients in the ICU due to increased rates of superinfection and non-selective suppression of immune responses [51] [52] [53] . Our study has several limitations. First, some of the results were derived from a single study (i.e., Anakinra) or studies with high risk of bias. To account for such weakness in evidence, we assessed the certainty of evidence for each outcome using the GRADE framework as summarized in Table 2 . Second, for certain treatment agents, many articles have been . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . published among which only one or few have been included in our analysis (e.g. convalescent plasma). This is because we prospectively collected studies adhere to predefined inclusion criteria, and studies that did not adequately account for confounding or those prone to significant bias were filtered out. The excluded studies are listed and described in the supplementary appendix pp 171-174 with reasons for exclusion. Third, we included observational studies and unpublished data. While such inclusions may introduce biases into the final analysis, we judged the benefits overweigh the risks for reasons we mentioned in methods. Furthermore, we attempted to minimize biases by exclusively including observational studies with confounder-adjustment and further conducted sensitivity analyses in which the same analysis was performed using only RCTs or only published studies (Appendix pp 55-58). Lastly, some of the results derived from this NMA lacks the support of pairwise meta-analysis. However, the methodological power of NMA is credible as empirical evidence supported that NMAs were 20% more likely to provide stronger evidence against the null hypothesis than conventional pairwise meta-analyses 54 . Accordingly, our NMA can offer meaningful implications for guiding management of COVID-19 until future studies build up stronger evidence. Anti-inflammatory agents (tocilizumab, anakinra, and IVIG) and remdesivir may safely and effectively improve clinical outcomes of COVID-19. Widely used hydroxychloroquine provides marginal clinical benefit in improving viral clearance rates whilst posing both cardiac and non-cardiac safety risks. Only 20% of current evidence on pharmacological management of COVID-19 is on moderate/high evidence certainty and can be considered in practice and policy; remaining 80% are of low or very low certainty and warrant further studies to establish firm conclusions. There was no funding source for this study. MS Kim and MH An had full access to all of the data in the study and MS Kim and T Hwang had final decision responsibility to submit for publication. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work. Dissemination declare: The manuscript's guarantor (MSK) affirms that this manuscript is an honest, accurate, and transparent account of study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. . Proportion of patients with cardiovascular comorbidities was high at baseline (approximately 30%), and proportions of cardiovascular comorbidities were significantly different between groups in crude analysis. However, adjustments were made for sex, age category (<65 vs≥65 years), diabetes, any chronic lung disease, cardiovascular disease, abnormal chest imaging, respiration rate >22/min, O2 saturation <90%, elevated creatinine, and AST >40 U/L. Adjusted odds ratios for prolonged QT interval were presented. Obesity was significantly higher in pharmacologic treatment groups (hydroxychloroquine, hydroxychloroquine, alone, and azithromycin alone) compared to control (neither drug). Obesity was not adjusted for. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. Only the adverse event-related data from this study was used in this network meta-analysis, as this study investigates the effect of hydroxychloroquine as a prophylactic measure which is not the focus of our meta-analysis. -Median age 41 -Cardiovascular disease (0.7%, not include hypertension) CAD/CHD: coronary artery disease / congestive heart disease. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 19, 2020. Records selected for full test review: n = 251 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . C Progression of disease to severe courses (i.e. progression to severe pneumonia and/or admission to ICU) D Time to viral clearance (days) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 19, 2020. . Pharmacological treatments for generalised anxiety disorder: a systematic review and network meta-analysis Reboxetine for acute treatment of major depression: systematic review and meta-analysis of published and unpublished placebo and selective serotonin reuptake inhibitor controlled trials Comparative efficacy and safety of reperfusion therapy with fibrinolytic agents in patients with ST-segment elevation myocardial infarction: a systematic review and network meta-analysis Organised inpatient (stroke unit) care for stroke: network meta analysis Prediction models for diagnosis and prognosis of covid-19 infection: systematic review and critical appraisal The impact of 2019 novel coronavirus on heart injury: A systemic review and Meta-analysis AGA Institute Rapid Review of the GI and Liver Manifestations of COVID-19, Meta-Analysis of International Data, and Recommendations for the Consultative Management of Patients with COVID-19 School closure and management practices during coronavirus outbreaks including COVID-19: a rapid systematic review Combining randomized and non randomized evidence in network meta analysis Efficacy and safety of pharmacological treatments for Patent Ductus Arteriosus closure: a systematic review and network metaanalysis of clinical trials and observational studies Efficacy, effectiveness, and safety of herpes zoster vaccines in adults aged 50 and older: systematic review and network meta-analysis Relative efficacy and safety of topical non-steroidal anti-inflammatory drugs for osteoarthritis: a systematic review and network meta-analysis of randomised controlled trials and observational studies The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses Assessing the quality of randomized trials: reliability of the Jadad scale ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions A GRADE Working Group approach for rating the quality of treatment effect estimates from network meta-analysis ):e76654. systemic-onset juvenile idiopathic arthritis: a randomised, double-blind, placebo-controlled, withdrawal phase III trial Kawasaki-like multisystem inflammatory syndrome in children during the covid-19 pandemic in Paris, France: prospective observational study Intravenous immunoglobulin for the treatment of Kawasaki disease in children Cochrane Review: Intravenous immunoglobulin for Guillain Barré syndrome. Cochrane Database Syst Rev Anti-inflammatory actions of intravenous immunoglobulin Intravenous immunoglobulin for treating sepsis, severe sepsis and septic shock A systematic review of Anakinra, Tocilizumab, Sarilumab and Siltuximab for coronavirus-related infections Dysregulation of immune response in patients with COVID Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study COVID-19: consider cytokine storm syndromes and immunosuppression COVID-19: immunopathology and its implications for therapy Corticosteroid treatment in critically ill patients with pandemic influenza A/H1N1 2009 infection: analytic strategy using propensity scores Clinical features of human influenza A (H5N1) infection in Vietnam Rationale: Study design: If randomized trials form the evidence base, the quality rating starts at "high". If observational studies form the evidence, base the quality rating starts at "low". Risk of bias: Downgraded for failure to conceal random allocation or blind participants in randomized controlled trials or failure to adequately control for confounding in observational studies. Inconsistency: Downgraded if heterogeneity represented by I 2 statistics or global inconsistency (Q statistic to assess consistency under the assumption of a full design-by-treatment interaction random effects model) was high. Indirectness. Downgraded when assumption of transitivity is challenged, or the result is solely derived from indirect comparisons. Imprecision: Downgraded when confidential interval (CI) is too large A GRADE Working Group approach for rating the quality of treatment effect estimates from network meta-analysis"): High quality: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate quality: We are moderately confident in the effect estimate i.e. the true effect is likely to be close to the estimate of the effect