key: cord-0781061-tsmqssjm
authors: Thibault, F.; Guihur, A.; Rebeaud, M.; Mulot, M.; Mahamat-Saleh, Y.
title: Hydroxychloroquine and mortality risk of patients with COVID-19: a systematic review and meta-analysis of human comparative studies
date: 2020-06-19
journal: nan
DOI: 10.1101/2020.06.17.20133884
sha: a1ca9feecb195188d8ad02f7f30f4ee6b384736a
doc_id: 781061
cord_uid: tsmqssjm

Background: Global COVID-19 deaths reached at least 400,000 fatalities. Hydroxychloroquine is an antimalarial drug that elicit immunomodulatory effects and had shown in vitro antiviral effects against SRAS-CoV-2. This drug divided opinion worldwide in the medical community but also in the press, the general public and in public health policies. The aim of this systematic review and this meta-analysis was to bring a new overview on this controversial drug and to assess whether hydroxychloroquine could reduce COVID-19 mortality risk in hospitalized patients. Methods and Findings: Pubmed, Web of Science, Cochrane Library, MedRxiv and grey literature were searched until 10 June 2020. Only studies of COVID-19 patients treated with hydroxychloroquine (with or without azithromycin) compared with a comparative standard care group and with full-text articles in English were included. Studies reporting effect sizes as Odds Ratios, Hazard Ratio and Relative Risk for mortality risk and the number of deaths per groups were included. This meta-analysis was conducted following PRISMA guidelines and registered on PROSPERO (Registration number: CRD42020190801). Independent extraction has been performed by two independent reviewers. Effect sizes were pooled using a random-effects model. The initial search leaded to 112 articles, from which 16 articles met our inclusion criteria. 15 studies were retained for association between hydroxychloroquine and COVID-19 survival including 15,081 patients (8,072 patients in the hydroxychloroquine arm and 7,009 patients in the standard care arm with respectively, 1,578 deaths and 1,423 deaths). 6 studies were retained for hydroxychloroquine with azithromycin. Hydroxychloroquine was not significantly associated with mortality risk (pooled Relative Risk RR=0.82 (95% Confidence Interval: 0.62-1.07, I2=82, Pheterogeneity<0.01, n=15)) within hospitalized patients, nor in association with azithromycin (pooled Relative Risk RR=1.33 (95% CI: 0.92-1.92, I2=75%, Pheterogeneity<0.01, n=6)), nor in the numerous subgroup analysis by study design, median age population, published studies (vs unpublished articles), level of bias risk. However, stratified analysis by continents, we found a significant decreased risk of mortality associated with hydroxychlroquine alone but not with azithromycin among European (RR= 0.62 (95%CI: 0.41-0.93, n=7)) and Asian studies (RR=0.36 (95%CI:0.18-0.73, n=1)), with heterogeneity detected across continent (Pheterogeneity between=0.003). These finding should be interpreted with caution since several included studies had a low quality of evidence with a small sample size, a lack of adjustment on potential confounders or selection and intervention biases. Conclusion: Our meta-analysis does not support the use of hydroxychloroquine with or without azithromycin to reduce COVID-19 mortality in hospitalized patients. It raises the question of the hydroxychloroquine use outside of clinical trial. Additional results from larger randomised controlled trials are needed

Inclusion criteria were 1) reports must contain original data with available risk estimates (Hazard 160

Ration, Odds Ratios, Relative Risk and/or with data on the number of death in HCQ and control 161

groups 2) all publication dates will be considered 3) publications in English language 4) comparative 162 studies with a control group without hydroxychloroquine and 5) COVID-19 confirmed cases by PCR. Reviews and meta-analysis, commentaries, in vitro and in vivo studies were excluded. 164 165

Data extraction 166 Data extraction was performed by two investigators (Mr. T. Fiolet and Mr. Y. Mahamat-Saleh) who 167 screened the titles and abstracts. Discrepancies were resolved by a third investigator (Dr. Anthony 168

Guihur). 169

The following data were extracted from each study: study design, publication date, location, number 170

of participants (total, in treatment and control groups, doses when available, effect size (Hazard Ratio, 171

Odds Ratio or Relative Risk) and 95% confidence intervals for reported risk estimates. Hazard Ratio 172

(HR) refers to the ratio of hazards in the intervention group divided by those occurring in the control 173

group. Hazard represents the instantaneous event rate, which means the probability that an individual 174

would experience an event (e.g. death) at a particular given point in time after the intervention, 175

assuming that this individual has survived to that particular point of time without experiencing any 176 event. In contrast, Relative Risk (RR) and Odds Ratio (OR) does not take account of the timing of 177 each event. RR and OR are similar when the event (death) is rare. The most adjusted effect size 178

reflecting the greatest control of potential confounders was extracted. 179

Three included studies did not report effect size for mortality risk (15,20,21). Thus we used the 180 number of death per groups to calculate an unadjusted relative risk using metabin function in meta 181 package in R Software (22) . RR calculation is based on Cochrane Handbook for Systematic Reviews 182 of Interventions formula RR = number of deaths in treatment group number of participants in treatment group number of deaths in control group number of participants in control group

For all the other studies, reported adjusted OR, RR or HR were used. The quality of each study was 184 assessed with ROBIN-I tool following Cochrane guidelines for non-randomized studies and with Rob2 185

for randomized studies (24,25).

Outcome 188

The outcome is COVID-19 mortality.

Statistical analysis 191

Effect of HCQ alone and HCQ + AZ 192

A primary meta-analysis was performed to assess the association between hydroxychloroquine alone 193 (vs standard care) and risk of death. In a second time, the relationship between hydroxychloroquine 194 associated with azithromycin and mortality was assessed. HRs, ORs and RRs were treated as 195 equivalent measures of mortality risk. Pooled RRs were determined by using a random effect model 196 with inverse variance weighting (DerSimonian-Laird method) (26). Significance was checked by Z-197 test (p<0.05 was considered as significant).

Heterogeneity was assessed by the Chi-square test and I² test. 30%<I²<60% was interpreted as 200 moderate heterogeneity and I²>60 as high heterogeneity. Funnel plot was constructed to assess the 201 publication bias. Begg's and Egger's test were conducted to assess the publication bias (7,27). RR or 202

HR and their 95% confidence interval were used to assessed mortality risk. 203 204 205 206 . 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. . https://doi.org/10.1101/2020.06.17.20133884 doi: medRxiv preprint

Subgroup analyses were further conducted according to the quality assessment to explore the source of 208 heterogeneity among observational studies. We performed stratified analyses by continents, the type 209 of article (peer-reviewed vs unpublished), the use of an adjustment on confounding factors (studies 210

with RRunadjusted vs RRadjusted), the mean daily dose of hydroxychloroquine (continuous), the median 211 population age across the studies (median age>63 years) and the level of bias risk identified with 212 ROBIN-I (moderate/serious/critical) (24), the exclusion of studies with cancer and dialysis patients. 213

Mean daily dose of hydroxychloroquine is a daily average between the loading dose and the 214 maintenance doses. Additionally, influence analysis was conducted by omitting each study to find 215 potential outliers (28). It is used to detect studies which influence the overall estimate of our meta-216

analysis the most, omitting one study at a time (leave-one-out method).

A two-sided p-value <0.05 was considered statistically significant. All analysis were conducted using 220 R version 3.6.1 with meta package and robvis package (29 Study characteristics 236 . 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. Characteristics of studies included in the meta-analysis for COVID-19 mortality IQ=Interquartile range, SD=Standard Deviation, HCQ=Hydroxychloroquine, AZ=Azithromycine, NA=Not available a Some studies did not report mean or median age b HR and OR are the most adjusted effect size reported in each study. Some studies did not report effect size. RRcalculated were calculated using the number of death in the treatment and the control groups

Risk of bias was assessed with ROBIN-I for non-randomised studies (n=14) and Rob2 was not 252 applicable for RECOVERY RCT because data were not available ( Figure S1 ). Details on the 253 assessment of studies quality are provided in Fig S2. Among the non-randomized studies, the majority 254 of these observational studies had a high or critical risk of bias (10 out of 16) 255

(12,14,15,20,21,30,32,33,35,36). Five articles had a moderate risk of bias(10,11,13,16,31). Some 256 studies did not report adjusted effect sizes to control confusion and selection bias (15,20,21,32,33,35). 257

Studies quality was lowered by the lack of information about the assignment of treatment, the time 258

between start of follow-up and start of intervention), some unbalanced co-intervention with other 259 antiviral and antibiotic drugs.

Hydroxychloroquine and mortality 262

The pooled RR for COVID-19 mortality was 0.82 (95% CI: 0.62-1.07, I²=82, Pheterogeneity<0.01, n=15) 263

( Figure 2 ) indicating no significant association between hydroxychloroquine and COVID-19 survival 264 or increased mortality. There was significant high heterogeneity across the included studies (I² =83%, 265

p<0.01). Egger's test (p= 0.42) and Begg's test (P=0.88) were not significant for asymmetry of the 266 funnel plot indicating that there is not a major publication bias ( Figure S3 ). In our separated analysis 267

by study design, we found a positive but not significant association between hydroxychloroquine alone 268

and mortality among interventional studies (RR: 1.10, 95%CI: 0.97-1.25, I²=0%, Pheterogeneity within=0.5, 269 n=2); however an inverse but not significant association was found among observational studies (RR: 270 0.78, 95%CI: 0.58-1.05, I²=82%, Pheterogeneity within <0.01), with heterogeneity observed across the study 271 design (Pheterogeneity between = 0.03) .  272  273  274  275  276  277  278  279  280  281  282  283  284  285  286  287  288  289  290  291  292  293  294  295  296  297  298  299  300 . 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. . 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. 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. n=1) for the RECOVERY randomized controlled trial (Figure 2 ). 336 337

After stratification by the level of bias from ROBIN-I evaluation, the association between 338

hydroxychloroquine and COVID-19 mortality remained non-significant. The broadness of 95% CI and 339 . 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. n=3)) ( Figure S4 ).

In our stratified analysis by continents ( Figure S5) Figure S7 ). Removing these studies make heterogeneity decrease at I²=0% but the results 363 remained non-significant (RR=1.00 (95% CI: 0.0-1.13, I²=29%, n=11) ( Table 2) . 364 365

All the results remained similar after exclusion of the two interventional studies (Table S1) .

Hydroxychloroquine with azithromycin and mortality 368

The pooled RR for COVID-19 mortality was 1.33 (95% CI: 0.91-1.921, n=6) ( Figure 3 ) indicating no 369 significant association between hydroxychloroquine with azithromycin and survival. There was 370 significant high heterogeneity across the included studies (I² =75%, p<0.01). Egger's test (p= 0.9) and 371

Begg's test (p=0.6) were not significant but the asymmetry in the funnel plot indicates that there could 372 be a publication bias. However, the number of included studies is small. RR=Risk ratio. 95%CI= 95% Confidence Interval 378 379 . 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. . https://doi.org/10.1101/2020.06.17.20133884 doi: medRxiv preprint

In all the subgroup analysis (type of article, effect size, risk of bias, continent, mean daily dose, age, 381 exclusion of cancer and haemodialysis patients, influence analysis), no significant association between 382 hydroxychloroquine with azithromycin and mortality was found (Table 2) . Nevertheless, in our 383 stratified analysis by continents, we found no significant association with COVID-19 survival risk 384

among American studies (RR=1.10, 95%CI: 0.91-1.32, I²=0%, Pheterogeneity within=0.48, n=3) and 385

European studies (RR=0.24 (95%CI: 0.00-13.43, I²=80%, Pheterogeneity within <0.02, n=2)) but there was a 386 significant increased risk of mortality in the multiple countries (RR=2.93, 95%CI: 1.79-4.79, n=1), 387

with heterogeneity detected across continent (Pheterogeneity between=0.0009). 388 389 390

Discussion 391

This meta-analysis summarized the results of 14 observational studies, 1 non-randomised study and 1 392

unpublished randomised controlled trial on hydroxychloroquine with or without azithromycin and 393

COVID-19 survival ( Table 1 ). The results indicated that hydroxychloroquine with or without 394

azithromycin is ineffective to reduce COVID-19 mortality risk in hospitalized patients (Figure 2 and  395 3 Figure S1 ). This significant relationship could be explained by a high risk of confusion bias since 407 these articles did not reported adjusted effect size. These studies also have several biases, such as a 408

selection bias Gautret et al, control and treatment groups did not come from the same hospital. In 3 409

Spanish studies (14,15,33), there was no information when treatment were administrated and when the 410 follow-up began which may lead to a bias in selection. Studies with an adjusted HR in figure S5 and 411

with a higher quality reported a non-significant higher RR than the other studies. In this meta-analysis, 412 the majority of the included studies had a high or critical risk of bias (10 out of 16) ( Figure S1 and S2). 413

Most of them do not always report the concomitant use of antiviral or antibacterial drugs. In our 414 subgroup analysis by study design, we found inconsistent results with a positive but not significant 415 association between hydroxychloroquine alone and mortality among interventional studies and an 416 inverse but not significant association among observational studies ( . 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. 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. 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. . https://doi.org/10.1101/2020.06.17.20133884 doi: medRxiv preprint Figure S2 : Assessment of quality of studies using ROBIN-I for non-randomised studies .   729   730  731  732  733  734  735  736  737  738  739  740  741  742  743  744  745  746  747  748  749  750  751  752  753  754  755  756  757  758  759  760 . 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. . https://doi.org/10.1101/2020.06.17.20133884 doi: medRxiv preprint 798   799  800  801  802  803  804  805  806  807  808  809  810  811  812  813  814  815  816 . 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. 820  821  822  823  824  825  826  827  828  829  830  831  832  833  834  835  836  837  838  839  840 . 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. 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. . 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. Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, metaregression), if done, indicating which were prespecified.

p.5 lines 208-218

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

p.5 Fig. 1 Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

p. 6-10 Table 1 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Synthesis of results 21 Present results of each meta-analysis done, including confidence intervals and measures of consistency. . 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. . https://doi.org/10.1101/2020.06.17.20133884 doi: medRxiv preprint

Minimum costs to manufacture new treatments for 493 COVID-19

The Risks of Prescribing Hydroxychloroquine for Treatment of COVID-495 19-First, Do No Harm

Hydroxychloroquine in the treatment and prophylaxis of SARS-CoV-2 infection in 498 non-human primates

RoB 2: a revised tool 564 for assessing risk of bias in randomised trials

Meta-analysis in clinical trials

Operating characteristics of a rank correlation test for publication bias. 569 Biometrics

Outlier and influence diagnostics for meta-analysis. Res Synth 571 Method

Risk-of-bias VISualization (robvis): An R package and Shiny web app 573 for visualizing risk-of-bias assessments

Outcomes of 575 hydroxychloroquine usage in United States veterans hospitalized with COVID-19

Infectious 580 Diseases (except HIV/AIDS)

Renal COVID Task Force on the clinical characteristics and short-term outcome of hemodialysis 584 patients with SARS-CoV-2 infection

Status of 586 SARS-CoV-2 infection in patients on renal replacement therapy

RECOVERY trial: the UK covid-19 study resetting expectations for clinical trials

Covid-19 transmission, outcome and associated risk factors in cancer patients at the first 595 month of the pandemic in a Spanish hospital in Madrid

Outcomes of Hydroxychloroquine Treatment 597

Among Hospitalized COVID-19 Patients in the United States-Real-World Evidence From a 598

Federated Electronic Medical Record Network

No clinical benefit from use of hydroxychloroquine in hospitalised patients 602 with COVID-19 -RECOVERY Trial

Hydroxychloroquine in patients with 608 mainly mild to moderate coronavirus disease 2019: open label

A pilot study of hydroxychloroquine in treatment of patients with 612 moderate COVID-19

Hydroxychloroquine is 614 associated with slower viral clearance in clinical COVID-19 patients with mild to moderate 615 disease: A retrospective study. medRxiv

Hydroxychloroquine or

Chloroquine for Treatment or Prophylaxis of COVID-19: A Living Systematic Review

Safety of 620 hydroxychloroquine, alone and in combination with azithromycin, in light of rapid wide-spread 621 use for COVID-19: a multinational, network cohort and self-controlled case series study 622

Rating the quality of evidence-study limitations (risk of bias)

Does the inclusion of grey literature influence 628 estimates of intervention effectiveness reported in meta-analyses? The Lancet

Grey literature in meta-analyses of randomized 631 trials of health care interventions. Cochrane Database of Systematic Reviews

Solidarity" clinical trial for COVID-19 treatments

A 639 Randomized Trial of Hydroxychloroquine as Postexposure Prophylaxis for Covid-19

Assessment of QT 642 Intervals in a Case Series of Patients With Coronavirus Disease

Treated With Hydroxychloroquine Alone or in Combination With Azithromycin in an Intensive 644

Prolongation Associated With Use of Hydroxychloroquine With or Without Concomitant 648

COVID-19)

Coronavirus (COVID-19) Update: FDA Revokes Emergency Use 652 Authorization for Chloroquine and Hydroxychloroquine

COVID-19 : 656 l'ANSM souhaite suspendre par précaution les essais cliniques évaluant l'hydroxychloroquine 657 dans la prise en charge des patients -Point d'Information -ANSM : Agence nationale de 658 sécurité du médicament et des produits de santé

DIMITROVA EK. COVID-19: reminder of risk serious side effects with chloroquine and 663 hydroxychloroquine

Hydroxychloroquine prophylaxis for COVID-19 contacts in 667

The Lancet Infectious Diseases

Figure S4: Forest plot for hydroxychloroquine alone and COVID-19 mortality risk

COVID-19 or SRAS-CoV-2) in Title Abstract 866 Keyword -(Word variations have been searched) 867 868 PubMed 869 Website

You searched for: TOPIC: (covid-19 OR SRAS-CoV-2) AND TOPIC: (hydroxychloroquine or HCQ) 880 AND TOPIC

Hydroxychloroquine COVID-19 mortality 886 Google