key: cord-352557-l7sahv5t
authors: Takla, Michael; Jeevaratnam, Kamalan
title: Chloroquine, hydroxychloroquine, and COVID-19: systematic review and narrative synthesis of efficacy and safety
date: 2020-11-13
journal: Saudi Pharm J
DOI: 10.1016/j.jsps.2020.11.003
sha: 
doc_id: 352557
cord_uid: l7sahv5t

The COVID-19 pandemic has required clinicians to urgently identify new treatment options or the re-purposing of existing drugs. Of particular interest are chloroquine (CQ) and hydroxychloroquine (HCQ). The aims of this systematic review are to systematically identify and collate 24 studies describing the use of CQ and HCQ in human clinical trials and to provide a detailed synthesis of evidence of its efficacy and safety. Of clinical trials, 100% showed no significant difference in the probability of viral transmission or clearance in prophylaxis or therapy, respectively, compared to the control group. Among observational studies employing an endpoint specific to efficacy, 58% concurred with the finding of no significant difference in the attainment of outcomes. Three-fifths of clinical trials and half of observational studies examining an indicator unique to drug safety discovered a higher probability of adverse events in those treated patients suspected of, and diagnosed with, COVID-19. Of the total papers focusing on cardiac side-effects, 44% found a greater incidence of QTc prolongation and/or arrhythmias, 44% found no evidence of a significant difference, and 11% mixed results. The strongest available evidence points towards the inefficacy of CQ and HCQ in prophylaxis or in the treatment of hospitalised COVID-19 patients.

In line with the PICOT format (Riva et al., 2012) of framing subjects for clinical research, this study centres on answering the question: 'In patients suspected of, and diagnosed with, COVID-19, How efficacious, relative to standard symptomatic care, and safe are CQ and HCQ in patients at risk or suspected of, and diagnosed with, COVID-19?'

The subsequent elaboration of the systematic review and narrative synthesis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (L. et al., 2015) for evidence-based assessment of published research.

In light of the current global health emergency, and the requisite rapid turnover of publications to meet the consequently urgent need to obtain and analyse the results that they present, several authors have resorted to the use of preprint servers to disseminate their findings. Despite the evident shortfalls inherent in referring to data, whose quality has not been peer-reviewed, the present paucity of published original research on the efficacy and safety of CQ and HCQ in COVID-19 patients demands exceptional measures. As such, this systematic review will take into account non-peer-reviewed work, provided that they have been submitted to a preprint server, where they are available in open-access form.

Nevertheless, given its focus on data quality, this review will make unambiguous every instance in which data from such sources are used.

Therefore, on August 8 2020, MEDRXIV and BIORXIV, along with PubMed and Web of Science, acted as the databases for the initial search of items relevant to the PICOTformatted question. The preliminary use of the search terms "COVID", "chloroquine", and "hydroxychloroquine" yielded a large number of results that bore little relevance to the research topic. Combining such terms into phrases -"COVID" AND "chloroquine", and "COVID" AND "hydroxychloroquine" in the title or abstract -and requiring the term "outcome" or "outcomes" anywhere in a free full text appearing within the last year considerably focused the responses. The subsequent application of identical phase stenography to each database ensured internal consistency.

As this review aims to establish the weight of evidence for the use of a therapy in patients, data able to answer such a question must derive from primary research. Moreover, owing to the international extent of the present health crisis, any imposition of an original language requirement would exclude useful and otherwise rare resources. As such, following the collation of items in Mendeley and the removal of duplicates, application of these criteria excluded unique items for which there was either no English version or no original data.

Screening of the resulting papers against the criteria established by the PICOTformatted question -namely, the requirement that data be collected from COVID-19 patients treated with CQ and/or HCQ -included only controlled trials and observational studies. Of all case series considered, only those directly assessing CQ or HCQ safety addressed each facet of the question and were thus included.

Having applied the exclusion and inclusion criteria to all search results and removing duplicates at all stages where necessary, two investigators independently reviewed the final repertoire of studies.

Rather than merely verifying the relevance and scope of the material in the final library, holistic analysis of each item of research in line with the framework set out in the Checklist of Review Criteria provided by the Task Force of Academic Medicine and GEA-RIME committee (Bordage et al., 2001 ) ensured stringent appraisal of study quality.

Indeed, the identification of -among other facets of robust research -appropriate study design, statistical analysis, and quality control (Table 1) permitted only papers with sufficient scientific merit to pass onto the data extraction stage.

Among the research items constituting the final library for analysis, there exists a wide variation in study design, results, and, crucially, the extent to which each distinct aspect of the PICOT-formatted question is answered. As such, a specialised data extraction table collates and summarises the most important information in every paper (Table 2 ). In particular, emphasis on the different sample sizes and structures, symptomatic stages, doses of drug used, primary and/or secondary outcomes and overall design limitations, facilitates both clarity and caution when making comparisons between the sets of results presented.

The results of each search, and the number failing and passing exclusion and inclusion criteria, respectively, have been summarised in a flowchart ( Figure 1 ). proposals; 11 systematic reviews and/or meta-analyses; and three models of COVID-19 spread, pathological dynamics, and/or therapy. Likewise, of those unique items found on the preprint servers, 28 were excluded by the same criteria, distributed categorically as follows:

17 systematic reviews or meta-analyses; five study proposals; three commentaries; two models; and one literature review.

As part of the search attrition methodology, parallel application of the inclusion criteria to each set of remaining unique results left 13 and sixteen items from the first two, and preprint, databases, respectively. Of these 29 studies, three were duplicates, culminating in a library of 26 papers from all four databases.

The use of 13 criteria created in accordance with the aforementioned Checklist of Review Criteria elaborated by the GERA-RIME committee, meeting 12 of which was judged to be indicative of scientific rigour, excluded two (Mehra et al., 2020), (Molina et al., 2020) of the remaining 26 items (Table 1) .

However, it is noteworthy that, of the 24 studies approved by an ethics committee, 11 explicitly waivered informed consent -another basic tenet of data collection from patient trials -citing the retrospective nature of their research. Confusingly, three studies claiming ethical board approval failed to explicitly mention consent (Sbidian et al., 2020) , (Arshad et al., 2020) , , while one investigation (Boulware et al., 2020) only sought consent after randomised allocation had already taken place.

As a result of the relatively broad scope of the research question, the authors of the 19 papers passing quality appraisal employed a variety of study types, therapeutic doses, and primary and/or secondary outcomes (Table 2) . It is thus essential to distinguish results by study design in order to prevent invalid inferences drawn from comparison of data sets.

Although the research question requires the use of CQ and/or HCQ in patients suspected of, and diagnosed with, COVID-19, there exists a range of possible approaches to the collection of data obtained from such patients.

The gold standard of primary clinical research into the efficacy and safety of drugs administered to humans is the randomised controlled trial (RCT), evidence from which may be further buttressed by masking of subjects, experimenters, or both, as well as the use of a placebo in the control group. However, only 21% of items passing quality appraisal were clinical trials, of which, though all were randomised, only one (Boulware et al., 2020) employed either masking or an exclusively placebo control. While unacceptable under normal circumstances, the absence of both masking and a placebo is admissible in light of the ethical violation that would otherwise result from the use of either in the context of patient consent being unlikely. Indeed, the paper reporting use of masking and a placebo was likely able to do so given the asymptomatic presentation of the uninfected patients constituting the sample from which the control group derived.

The remaining 79% of papers were observational studies, of which the vast majority retrospectively searched hospital databases to collect clinical data obtained by following up on cohorts of patients from the time they received CQ and/or HCQ, or the standard of care only, until a defined end-point. Despite 74% using a case-control structure, a minority (26%)

constituted case series focusing on the cardiac safety of drugs administered to hospitalised COVID-19 patients for a given duration of time.

The deliberate absence of a specified dose in the research question accounts for the diversity of administration regimens among the 24 papers. Indeed, the doses used largely reflect the studies being conducted on different dates, which, in turn, influences the relative sway of either federal healthcare guidelines or the results of prior clinical research on regimen selection.

Every clinical trial tested a distinct dosing scheme: a loading dose of 1200 mg HCQ per day for three days, followed by a maintenance dose of 800 mg per day for two or three weeks if symptoms are mild/moderate or severe, respectively (W. Tang et al., 2020); 800 mg HCQ once and 600 mg six to eight hours later for one day, then 600 mg per day for four days (Boulware et al., 2020) ; 800 mg HCQ twice in six hours, then 400 mg at 12 hrs on day one, followed by 400 mg twice a day for nine days, or until discharge (Horby et al., 2020) ; 800 mg HCQ once on day one, then 400 mg per day for six days (Mitja et al., 2020) ; and 400 mg twice a day on day one, followed by 200 mg twice a day for four days (kamran et al., 2020).

Likewise, each clinical trial treated its control group differently, from: a placebo folate tablet administered in a regimen identical to that of the treatment group (Boulware et al., 2020) , to only a standard of care, itself varying with clinical centre and national guidelines.

By contrast, the retrospective and often multi-centre nature of many observational studies has resulted in 21% using variable or undeclared doses, some collecting data from some patients taking azithromycin in combination with the HCQ. The remaining 79% of items relied on highly divergent dosing regimens (Table 2) only on day one, then 400 mg per day for nine days (Arshad et al., 2020) ; or twice per day on day one, followed by 400 mg per day for four days (Geleris et al., 2020) , with or without azithromycin. Only three sets of authors also analysed data for patients taking CQ variably (Hsia et al., 2020) , or, more commonly, at a full dose of 500 mg twice a day, either: for one day, followed by half-dose for four days, with or without 500 mg azithromycin per day for five days (Saleh et al., 2020) ; or at a half-dose for the duration of treatment (Huang et al., 2020a) . Similarly, the 74% of observational studies with a control treated the group differently, giving standard of care without, or with declaration of additional antivirals, antibiotics, or both.

Given the vast array of possible measures of CQ and HCQ efficacy and safety in COVID-19 patients, the different authors outlined distinct primary and/or secondary outcomes.

The primary outcome in 80% of clinical trials was C t > 40 for PCR amplification of SARS-CoV-2 RNA, indicating either prophylaxis or viral clearance. Likewise, 63% of observational studies directly measured a specific indicator of efficacy other than mortality rate, using similar outcomes to the clinical trials, as well as the duration of hospital stay, need for mechanical ventilation, and probability of transfer to an intensive care unit (ICU).

As regards direct measurement of safety, all clinical trials and 74% of observational studies explicitly recorded adverse events as an indicator of CQ and/or HCQ safety in COVID-19 patients, with 50% of the total actively focusing on cardiac pathology.

Notably, 42% of studies used mortality rate as a key end-point. In isolation, however, risk of death could be indicative of either safety or efficacy. As such, this review reports the findings on mortality rate separately from those pertaining to outcomes that are specific measures of one of efficacy or safety.

Of the clinical trials providing data on a specific indicator of CQ and/or HCQ efficacy in patients suspected of, and diagnosed with, COVID-19, all showed no significant difference in the probability of viral transmission or clearance in the prophylactic (Boulware et al., 2020) , (Mitja et al., 2020) (Horby et al., 2020) , and dosage across the range of papers.

In contrast, only 58% of observational studies employing an endpoint specific to efficacy recorded no significant difference in the attainment of outcomes, such as duration of hospital stay, need for mechanical ventilation, and probability of transfer to an intensive care unit (ICU), between COVID-19 patients given a range of CQ and/or HCQ doses, and the control groups. Indeed, of the remaining papers, 60% found evidence of a higher probability of discharge rate (Sbidian et al., 2020) , viral clearance and shorter symptom duration (Huang et al., 2020a) in a therapeutic context, and a lower incidence of SARS-CoV-2 infection in a prophylactic context (Bhattacharya et al., 2020 Although 60% of clinical trials found evidence of higher mild adverse drug-related events in the treatment group, none of those specifically focusing on cardiac-side effects discovered any significant difference relative to the control. Likewise, 50% of observational studies examining an indicator unique to drug safety discovered a higher probability of adverse events in those treated patients suspected of, and diagnosed with, COVID-19. Of those retrospective studies measuring cardiac side-effects, 44% failed to find a significantly different incidence between the treatment and control groups, while another 44% indicated a significantly greater probability, with QTc prolongation the most common finding, in addition to its potentially lethal consequences of VT and cardiac arrest. In contrast, one such paper presented mixed findings, with an elevated risk of cardiac events despite no apparent rise in the risk of QTc prolongation.

Of the total studies using mortality rate as a key end-point, 60% reported no significant change in the risk of death, while 30% showed a depression, in treated relative to control patients. Interestingly, one investigation (10%) yielded a mixed result of raised mortality risk in patients given HCQ only, but no significant difference in those coadministered azithromycin (Magagnoli et al., 2020) .

The absence of a pharmacological treatment tailored to COVID-19 has rendered urgent the search to find alternative therapies by repositioning drugs with the theoretical potential to alleviate symptoms. However, hypothetical plausibility is insufficient grounds for translation into clinical practice. Indeed, any therapeutic repurposing must only proceed in light of strong evidence for the pre-clinical basis, and clinical efficacy and safety, of the drug in question. This review finds that, while such evidence certainly exists for the former, it does not for the latter, calling into question any clinical use of CQ and/or HCQ in COVID-19 patients prior to the collection and analysis of high-quality RCT data.

Pre-clinical studies performed in vitro provide strong evidence for the theoretical utility of CQ and HCQ in inhibiting all stages of viral entry, maturation, and spread.

In vitro, CQ blocks infection both at, and after, entry of SARS-CoV-2 into Vero E6 cells, with an EC 50 of 1.13 µM . Indeed, although therapeutic doses of CQ do not seem to alter S protein glycosylation (Vincent et al., 2005) , whose pattern is distinct from that of SARS-CoV-1 (Kumar et al., 2020), they may inhibit biosynthesis of sialic acid (Kwiek et al., 2004) , N-glycosylation of ACE2, as well as downregulating the expression of PICALM (Wolfram et al., 2017) in the clathrin-dependent endocytosis machinery.

Furthermore, immunofluorescence analysis of the amount of nucleoprotein in distinct vesicular compartments of the host cell has demonstrated that treatment of infected cells with CQ and HCQ stalls transfer of viruses from early to late endosomes . In fact, by increasing the pH of the early endosome, CQ has the potential to reduce acid-dependent proteolytic cleavage of the S protein by cathepsin and TMPRSS2, thereby inhibiting viral uncoating, genomic replication and particle maturation (Wang et al., 2008) . Despite its similar effect on viral distribution, as well as its comparable cytotoxicity (Liu and Li, 2020), to CQ, HCQ appeared to amplify and enlarge the late endosomes, implying a slightly distinct mechanism of action. Furthermore, there exists conflicting evidence for the relative in vitro efficacy of the two drugs (Yao et al., 2020) .

It is nonetheless clear that the initial basis for investigating the translatability of CQ and/or HCQ to the treatment of hospitalised COVID-19 patients was predicated on highquality evidence for its pre-clinical antiviral efficacy.

That CQ and HCQ can reduce viral entry, trafficking, and budding in vitro constitutes evidence of translational potential relies on the underlying assumption that symptom severity Importantly, however, in COVID-19 patients, symptoms subsequent to the causal infection result firstly from the initial cytokine wave of the innate immune response , then a state of immunodeficiency and lymphopenia (Hadjadj et al., 2020) , and, finally, a potentially lethal cytokine storm (F. . The causal distinction between these symptomatic phases highlights not only the difficulty in repurposing a single drug for use at all time-points, but also the need to approach with caution the comparison of trial data collected from patients given drugs at different times post-infection. The sample of one of the first clinical trials performed on patients testing positive upon PCR amplification of SARS-CoV-2 RNA comprised asymptomatic patients (17%), as well as those with upper (61%) and lower (22%) respiratory tract infections, thereby capturing the range of symptom severity. After 6 days of treatment, patients given HCQ alone had a higher probability of viral clearance compared to those given the standard of care only (57% vs. 13%), rising to 100% in patients also given azithromycin (Gautret et al., 2020) . That the authors additionally discovered a greater drug effect on patients with upper and lower respiratory tract infections than on asymptomatic individuals raises the possibility that the potential therapeutic benefit of HCQ in COVID-19 patients lies in its capacity for immunomodulation. On a theoretical level, the anti-inflammatory effects of HCQ render such an effect possible. Indeed, through alkalinisation of early endosomes, CQ and HCQ could impair: PAMP-induced activation of TLR7 and TLR9 (Saitoh and Miyake, 2009) , and, by extension, MMP-9 expression (Lim et al., 2006) ; antigen presentation by major histocompatibility complexes (MHCs) (Roche and Furuta, 2015) , (Guerriero, 2019) ; prostaglandin and thromboxane production (Nosál' and Jančinová, 2002) ; and T and B cell activation (Goldman et al., 2000) , differentiation, and proliferation (Yang et al., 2018) . Importantly, both SARS-CoV-2 , (Conti et al., 2020) and related coronaviruses, such as SARS-CoV-1 (Wang et al., 2014) and MERS-Nevertheless, many observational studies support the absence of significant evidence CQ and/or HCQ efficacy compared to that of standard of care alone. Furthermore, the homogeneity of, and correction for, baseline characteristics in the case and control cohorts further buttresses the reliability of the evidence presented by studies with sample sizes ranging from 1,376 (Geleris et al., 2020) to 3,372 (S. . One such study, however -with a sample size of 2,512 -used a treatment cohort with a significantly lower age but greater symptomatic disease compared to the control (Ip et al., 2020) . Similarly, the vast majority of papers contending that CQ and/or HCQ delayed SARS-CoV-2 clearance in COVID-19 patients were severely underpowered and exhibited significant differences in sample structure and/or comorbidities, nullifying their influence on the conclusion of this review (Mahévas et al., 2020).

As such, it is reasonable to conclude that, at present, the highest quality evidence does not support the efficacy of either CQ or HCQ in the prophylaxis or treatment of patients at high risk of, or diagnosed with, COVID-19, relative to the standard in-hospital management of symptoms. Nevertheless, the completion of more RCTs will be necessary to further substantiate this claim.

Robust evidence for the safety of an otherwise efficacious drug is a prerequisite for its widespread application in any clinical setting. In light of the unsubstantiated benefit of CQ and/or HCQ administration in patients with COVID-19, there exists an even more compelling imperative to ensure that any compassionate use did, and does, not contribute to excess mortality.

Despite their adequate safety records in an anti-inflammatory context, as aminoquinolines, CQ, and its derivative, HCQ, are proarrhythmic (Khobragade et al., 2013) .

Arrythmias arise from an imbalance of the normal physiological variables influencing the activation and inactivation kinetics of the cardiac ion channels that permit the transmembrane currents forming the foundation of the cardiac action potential. In a healthy milieu intérieur, the waveform of this cardiac action potential is quadriphasic (Noble, 1962) . Electrical diastole (IV) precedes a rapid depolarisation (0), after which a refractory period is followed by a slow hyperpolarisation (I), a 300 ms plateau (II), and then a period of repolarisation (III)

to resting membrane potential (E m ). By blocking -in order of increasing potency -the delayed (I Kr ) and inwardly-rectifying (I K1 ) K + currents (Sánchez-Chapula et al., 2001) , and the latter preferentially at depolarised E m , CQ and HCQ significantly prolong the QT interval and slow ventricular conduction, thereby predisposing to early-after-depolarisation and, by extension, torsades de pointess. Combined with their tonic block of voltage-gated Na + and Ltype Ca 2+ currents at low channel opening frequencies, QT interval prolongation thus significantly increases the risk of potentially fatal VTs.

Indeed, despite the only randomised clinical trials examining prophylaxis having found no increases in adverse cardiac events upon HCQ administration to patients at risk of COVID-19 (Mitja et al., 2020) (Boulware et al., 2020) , there is some evidence that aminoquinoline treatment predisposes those diagnosed with confirmed infections to tachyarrhythmia. In fact, a significant association of high doses of CQ with lethality in patients with severe symptoms forced the premature termination of a RCT (Borba et al., 2020) . Despite this relationship with mortality risk disappearing upon correction for age, there remained a significantly higher proportion of patients in the high (19%) compared to the low (11%) dose group with QTcF > 500 ms, including three percent of patients who experienced VT before death. However, given the abortion of the study, as well as the coadministration of QT-prolonging oseltamivir (Hama and Bennett, 2017) confounding the causal link to cardiac side-effects, these data, alone, are insufficient to conclude that CQ is unsafe in COVID-19 patients.

Notably, the only paper failing this review's quality appraisal suggested a significant excess mortality in patients treated with CQ and/or HCQ relative to controls. A highlypowered observational study of 96,032 patients with no significant differences in comorbidities between groups, it claimed that patients given CQ (four percent) or HCQ (six percent) alone had a significantly augmented risk of de novo in-hospital ventricular arrhythmias, compared to controls (three-tenths of one percent) given standard therapy, including remdesivir (Mehra et al., 2020). However, within one day of investigative journalists' discovery of inconsistencies in the reporting of the data from international hospitals involved in the study, over 140 signatories penned an open letter to the authors of the paper and the journal in which it was published, also criticising the statistical analysis and lack of ethics review. Despite the study's consequent correction of the continental assignment of one hospital, which had no impact on its overall findings, questions remain over the lack of transparency of the company that managed the original databases. On June 4 2020, three of the four co-authors retracted the paper (Mehra et al., n.d.), thereby necessitating caution when interpreting its original findings. Nonetheless, a more recent observational study of 2,512 patients suspected of, or diagnosed with, COVID-19, has elucidated a significantly higher proportion of cardiac-attributable mortality in the HCQ treatment (21%) than the control (16%) group. Moreover, despite being individually underpowered, several smaller retrospective database searches focusing specifically on the QTc duration have consistently and independently supported a significant prolongation in hospitalised COVID-19 patients treated with a range of CQ and/or HCQ doses (Saleh et al., 2020) , (Mercuro et al., 2020) .

Regardless, drawing causal inferences from such observational studies is inadvisable given the lack of randomisation and absence of a placebo in the control groups, leaving the data susceptible to unmeasured confounders.

By contrast, the most highly powered RCT found no evidence of a significant excess of de novo cardiac arrhythmias in COVID-19 patients of varying severity given HCQ for 10 days, or until discharge (Horby et al., 2020) . Likewise, many of the largest retrospective case-

control studies passing quality appraisal demonstrated an absence of a statistically significant difference in the incidences of VTs, fibrillation, or sudden cardiac deaths (SCDs) between hospitalised patients administered HCQ and those provided with the standard of care alone (A. K. , (Arshad et al., 2020) .

In any case, from a theoretical standpoint, the cardiac safety risk of CQ and HCQ use is unlikely uniform among COVID-19 patients. Indeed, there is a significant positive correlation between baseline QTc and age (Reardon and Malik, 1996) , QTc prolongation and anti-dysrhythmic, antipsychotic or macrolide antibiotic co-administration (Al-Khatib et al., 2003) , and QT interval dispersion and mortality risk in type II diabetes mellitus (Giunti et al., 2012) . Given the relative risk conferred by both older age ("The epidemiological characteristics of an outbreak of 2019 novel coronavirus diseases in China," 2020) and type II diabetes mellitus (Richardson et al., 2020) on symptom severity and consequent probability of hospitalisation of COVID-19 patients, the data for in-hospital CQ/HCQ safety may not be extrapolable to many infected individuals in the population due to selection bias (Henderson and Page, 2007) . As such, the wide variation in average sample patient age, symptom severity and drug dosing regimen (Table 2) further complicates inferences of reliable agreement between the papers.

Therefore, as of Aug 8 2020, there is a lack of strong evidence that, relative to standard in-hospital management of symptoms, the use of CQ and HCQ to treat hospitalised COVID-19 patients has been either distinctly safe or unsafe. The substantiation, of presently unclear strength, for the tendency of the aminoquinolines to prolong QTc in this therapeutic context may still provide cause for concern in patients whose comorbidities predispose them to VTs. The ongoing inquiries into the validity of data sets indicating a lack of drug safety, however, should preclude definitive judgment on the matter until the completion of more high-quality RCTs.

Crucial to the understanding of the conclusions drawn in this systematic review is an appreciation of its many limitations, which relate to both the search methodology and data analysis.

Insofar as peer-reviewed publications are concerned, this review searched two databases to yield only 144 unique results, leaving the authors to also seek the findings of 69 papers on two preprint servers. Despite facilitating the collection of a more representative sample of current research on the subject in question, the absence of documented expert scrutiny ought to prevent their data from influencing clinical decisions. Nevertheless, to compensate for the lack of peer review, rigorous application of the quality appraisal criteria established by the GEA-RIME committee and the Task Force of Academic Medicine ensured that only data from adequately designed studies were taken into account. Importantly, however, that peer-reviewed journal material was no longer a prerequisite for inclusion may have slightly reduced the effects of positive publication bias (Mlinarić et al., 2017) on the results of this systematic review.

In light of the recent investigation into the integrity of the data collection and statistical analysis of the largest observational study to date (Mehra et al., 2020) , it is also necessary to advocate caution in the interpretation of results, even if the paper in which they are presented has undergone peer review. Indeed, worries concerning the decline in review duration and even scrutiny, as a result of the rush to publish, should elicit wariness in considering even the most fundamental of underlying assumptions; namely, that a data set is valid.

However, the predominant shortcoming of the review is its inability to completely disentangle the large differences in study design when making comparisons between different data sets from the included papers. Indeed, despite stressing the obvious invalidity of cutting across distinct sample sizes, baseline characteristics, drug doses, and individual limitations, a systematic review, by nature, does exactly that. The categorisation of the results and data analysis by randomisation, COVID-19 symptom severity, and HCQ/CQ dosage constitutes an attempt to reduce this problem of comparative inferences as greatly as possible.

On March 18 2020, the WHO announced the launch of an international phase III-IV RCT, with four arms, measuring the efficacy and safety of: (1) remdesivir; (2) lopinavir and ritonavir;

(3) lopinavir, ritonavir, and IFN-(beta); and (4) CQ or HCQ (Kupferschmidt, 2020).

In the meantime, amidst a dearth of high-quality evidence from completed randomised clinical trials, the U.S. Food and Drug Administration (FDA) issued an emergency use authorisation of CQ and HCQ in COVID-19 patients (Piller, 2020) .

Since then, data from the most robust of the completed case-control studies have failed to find any significant evidence of the efficacy of CQ and/or HCQ in the treatment of hospitalised COVID-19 patients. Meanwhile, on May 22 2020, in the midst of a large number of research groups finding evidence of significant QTc prolongation upon administration of aminoquinoline drugs, a recent large retrospective observational study indicated a possible lack of drug (particularly cardiac) safety in this clinical context. On May 25 2020, the WHO suspended the fourth arm of the Solidarity Trial, citing these safety concerns. In the UK, the Medicines and Healthcare products Regulatory Agency (MHRA) followed suit, calling the COPCOV trial to a halt. Less than one week later, however, an international coalition of scientists raised concerns regarding the integrity of the collection and analysis of the inaccessible data on which the suspension was based, and the responsible paper has since been retracted. The MHRA has authorised the COPCOV trial to recommence. Following the WHO's recontinuation of the Solidarity Trial arm, interim results provided little evidence of a reduction in mortality risk for hospitalised COVID-19 patients administered CQ and/or HCQ, ultimately leading to its indefinite discontinuation on July 4 2020.

The urge to begin all international and national clinical trials arose from the necessary desire to rapidly compensate for the prior and present scarcity of randomised data on the efficacy and safety of CQ and/or HCQ in patients infected with SARS-CoV-2. In the absence of evidence for the safety of CQ and HCQ in COVID-19 patients, the FDA's initial decision to authorise their use in hospitalised patients was of questionable scientific prudence.

Indeed, this review finds that, despite data from different in vivo studies conflicting and even contradicting each other, the strongest evidence does not support the efficacy of either CQ or HCQ in the prophylaxis or treatment of patients at high risk of, or diagnosed with, COVID-19, relative to the standard in-hospital management of symptoms. Likewise, there is a lack of strong evidence that the use of CQ and HCQ to treat hospitalised COVID-19 patients has had a distinct effect on safety outcomes.

Yet it is precisely because of this dearth of evidence that there still exists a pressing demand for RCTs. In fact, the saga ensuing from the hastiness in accepting the findings of the now-retracted observational study is a case in point of the need to collect randomised clinical data in order to substantiate or contend any claims of drug efficacy and/or safety. 

Long QT and hydroxychloroquine; a poorly recognised problem in rheumatology patients

What Clinicians Should Know about the QT Interval

Chloroquine and COVID-19: A Light at the End of the Tunnel, or is it Another Train? Electron

Treatment Response to Hydroxychloroquine and Antibiotics for mild to moderate COVID-19: a retrospective cohort study from South Korea

Treatment with hydroxychloroquine, azithromycin, and combination in patients hospitalized with COVID-19

Observational Study of the Efficiency of Treatments in Patients Hospitalized with Covid-19 in Madrid

Antiaggregation of chloroquine

Pre exposure Hydroxychloroquine use is associated with reduced COVID19 risk in healthcare workers -a Retrospective

Chloroquine Diphosphate as Adjunctive Therapy for Patients Hospitalized With Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Infection

Joint task force of academic medicine and the GEA-RIME committee

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

Prevention Of Postoperative Deep Venous Thrombosis In Legs By Orally Administered Hydroxychloroquine Sulphate

Chronic hydroxychloroquine use associated with QT prolongation and refractory ventricular arrhythmia

Clinical and immunologic features in severe and moderate Coronavirus Disease

Viral Replication in the Nasopharynx Is Associated with Diarrhea in Patients with Severe Acute Respiratory Syndrome

How to reduce the likelihood of coronavirus-19 (CoV-19 or SARS-CoV-2) infection and lung inflammation mediated by IL-1

The spike glycoprotein of the new coronavirus 2019-nCoV contains a furin-like cleavage site absent in CoV of the same clade

WHO declares COVID-19 a pandemic

An interactive web-based dashboard to track COVID-19 in real time

Structural and molecular modelling studies reveal a new mechanism of action of chloroquine and hydroxychloroquine against SARS-CoV-2 infection

Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial

Observational study of hydroxychloroquine in hospitalized patients with Covid-19

Increased QT nterval dispersion predicts 15-year cardiovascular mortality in type 2 diabetic subjects: The population-based Casale Monferrato Study

Hydroxychloroquine inhibits calcium signals in T cells: A new mechanism to explain its immunomodulatory properties

The species Severe acute respiratory syndrome-related coronavirus: classifying 2019-nCoV and naming it SARS-CoV-2

Macrophages: Their Untold Story in T Cell Activation and Function

Impaired type I interferon activity and exacerbated inflammatory responses in severe Covid-19 patients

The mechanisms of sudden-onset type adverse reactions to oseltamivir

Tissue distribution of ACE2 protein, the functional receptor for SARS coronavirus. A first step in understanding SARS pathogenesis

Temporal dynamics in viral shedding and transmissibility of COVID-19

Appraising the evidence: What is selection bias?

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Effect of Hydroxychloroquine in Hospitalized Patients with COVID-19: Preliminary results from a multi-centre, randomized

QT prolongation in a diverse, urban population of COVID-19 patients treated with hydroxychloroquine, chloroquine, or azithromycin

Preliminary evidence from a multicenter prospective observational study of the safety and efficacy of chloroquine for the treatment of COVID-19

Treating COVID-19 with

Reply to Gautret et al. 2020: A Bayesian reanalysis of the effects of hydroxychloroquine and azithromycin on viral carriage in patients with COVID-19

Hydroxychloroquine and Tocilizumab Therapy in COVID-19 Patients -An Observational Study

Phylogenetic Analysis and Structural Modeling of SARS-CoV-2 Spike Protein Reveals an Evolutionary Distinct and Proteolytically Sensitive Activation Loop

Chloroquine and hydroxychloroquine for COVID-19: implications for cardiac safety

Clearing the fog: Is HCQ effective in reducing COVID-19 progression: A randomized controlled trial. medRxiv 2020.07.30

Electrocardiographic safety of daily Hydroxychloroquine 400mg plus Azithromycin 250mg as an ambulatory treatment for COVID-19 patients in Cameroon

Dealing with the positive publication bias: Why you should really publish your negative results

No evidence of rapid antiviral clearance or clinical benefit with the combination of hydroxychloroquine and azithromycin in patients with severe COVID-19 infection

A modification of the Hodgkin-Huxley equations applicable to Purkinje fibre action and pacemaker potentials

Cationic amphiphilic drugs and platelet phospholipase A2 (cPLA2)

A short therapeutic regimen based on hydroxychloroquine plus azithromycin for the treatment of COVID-19 in patients with non-severe disease. A strategy associated with a reduction in hospital admissions and complications

Inhibition of Chikungunya virus infection in cultured human muscle cells by furin inhibitors: Impairment of the maturation of the E2 surface glycoprotein

Hydroxychloroquine sulfate in prevention of thromboembolic phenomena in surgical patients

Former FDA leaders decry emergency authorization of malaria drugs for coronavirus

Pulmonary Embolism in COVID-19 Patients: Awareness of an Increased Prevalence

The blood-retinal barrier in chloroquine retinopathy

Hydroxychloroquine Protects the Annexin A5 Anticoagulant Shield from Disruption by Antiphospholipid Antibodies

QT interval change with age in an overtly healthy older population

Presenting Characteristics, Comorbidities, and Outcomes among

Patients Hospitalized with COVID-19 in the New York City Area

What is your research question? An introduction to the PICOT format for clinicians

The ins and outs of MHC class II-mediated antigen processing and presentation

Regulatory molecules required for nucleotide-sensing Toll-like receptors

The Effect of

Hydroxychloroquine and Azithromycin on the Corrected QT Interval in Patients with SARS-CoV-2 Infection

Blockade of currents by the antimalarial drug chloroquine in feline ventricular myocytes

Hydroxychloroquine with or without azithromycin and in-hospital mortality or discharge in patients hospitalized for COVID-19 infection: a cohort study of 4,642 in-patients in France

Chloroquine and hydroxychloroquine in the treatment of COVID-19 with or without diabetes: A systematic search and a narrative review with a special reference to India and other developing countries

Outcomes of Hydroxychloroquine Treatment Among Hospitalized COVID-19 Patients in the United States-Real-World Evidence From a Federated Electronic Medical Record Network

Chloroquine is a potent inhibitor of SARS coronavirus infection and spread

SARS coronavirus entry into host cells through a novel clathrin-and caveolae-independent endocytic pathway

Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro

Antibody-dependent SARS coronavirus infection is mediated by antibodies against spike proteins

A chloroquine-induced macrophage-preconditioning strategy for improved nanodelivery

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China

Ocular toxicity of hydroxychloroquine

Hydroxychloroquine inhibits the differentiation of Th17 cells in systemic lupus erythematosus

Exuberant elevation of IP-10, MCP-3 and IL-1ra during SARS-CoV-2 infection is associated with disease severity and fatal outcome

In Vitro Antiviral Activity and Projection of Optimized Dosing Design of Hydroxychloroquine for the Treatment of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)

Low dose of hydroxychloroquine reduces fatality of critically ill patients with COVID-19

Low dose of hydroxychloroquine reduces fatality of critically ill patients with COVID-19

D-dimer levels on admission to predict in-hospital mortality in patients with Covid-19

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

No funding received for this study

MT wrote the first draft and edited all subsequent drafts of the manuscript. KJ designed the study and edited subsequent drafts of manuscript.

HorbyMention of ethical approval, but explicit waiver of informed consent No serious side-effects observed.Absence of a placebo. Standard of care unusually consisted of Zinc, and vitamins C and D.Overwhelmingly male sample prevents generalisation to both sexes.A subset of day 7 PCR negatives retested positive on day 14, suggesting either false negatives on day 7 due to variable kit sensitivity, or false positives on day 14 as non-replicable viral nucleic acid is detected by PCR after 10 days. No significant difference in mortality by 28 days between HCQ only and control.Significantly higher mortality risk for patients given HCQ + azithromycin compared to control.Significantly higher discharge rate in the group given HCQ only.

Lack of randomisation.Limited sample size for the HCQ + azithromycin subgroup necessitates caution in interpreting higher mortality risk compared to control.Dosage may have varied from guidelines due to physician discretion.Significantly higher proportion of younger, male and smoking patients in the treatment cohorts than the control, though no significant differences in biological parameters. No significant difference in mortality.Significantly higher proportion of mortality attributable to cardiac causes in patients treated with HCQ (21%) compared to the control group (16%).Similar incidence of arrhythmias and cardiomyopathy in treated (5% and 1%) and control (4% and 1%) patients.Lack of randomisation.Significantly lower age, but later presentation in clinical course and greater symptomatic disease in the treatment group.Possible misclassification due to manual abstraction of HER data.Possible sampling bias due to use of a convenience sample for data collection. Considerable imbalance in the sample size of the treatment and control groups.Interferon application reached 11% in the control, but 0% in the treatment group.Among patients who died, significantly longer hospital stay for the HCQ (15 days) compared to the control (8 days) group.Significant reduction in plasma IL-6 in HCQ, but not control patients. Among patients with IL-6 > 60 pg ml -1 , HCQ treatment, but not control treatment reversed the trend after 10 days, and significantly reduced fatality. Significantly higher probability of viral clearance by days 10 and 14 in the treatment group.Significantly shorter duration of fever symptoms in the treatment group.No serious adverse events observed in the treatment group. However, higher incidence of GI disturbances.Significantly lower incidence of adverse events in patients in the treatment group on halfdose, compared to those given the full dose.Lack of randomisation.Only 1 patient in the treatment, compared to 9 patients in the control, group experienced aggravated (i.e. moderate to severe) symptoms.Impossibility of dissociating the effects of CQ from those of other antiviral drugs used prior to its administration. However, patients given CQ within 3 days of symptom onset still exhibit faster viral clearance. Higher probability of prescribing HCQ ± azithromycin to patients with more severe metabolic, haematological, and ventilatory symptoms of COVID-19.Significant differences in demography, vital signs, prescription drug use, comorbidities, and disease severity. However, all adjusted by propensity score. Lack of randomisation.Small sample size, with large imbalance between treatment (31) and control (195) groups limiting the number of factors included in the propensity score model.A few patients in the treatment group (7.5% of total) also received treatment with lopinavir/ritonavir, though duration of use included in multivariate Cox analysis.Baseline systolic ABP and haematocrit varied significantly between treatment and control group, even after matching. 10% of HCQ patients experienced averse ECG modifications requiring cessation of treatment after a median of 4 days. Among them, 88% had a QTc prolongation > 60 ms (including > 500 ms in one patient). One of these patients presented with a 1 st degree AV block after 2 days despite a lack of concomitant proarrhythmic medication.

Small sample size.Lower probability of co-administration of azithromycin in the HCQ (18%) compared to the control (29%) group. Higher probability of co-administration of amoxicillin and clavulanic acid in the HCQ (52%) compared to the control (28%) group.HCQ patients had lower prevalence of comorbidities, except hepatic cirrhosis.The 4 covariates exceeding the standardised difference threshold were excluded from the final propensity score model. Imbalance in the number of HCQ patients between centres not taken into account by the propensity score model. in the treatment group. as GI disturbances (19%), skin rash (6%), and headache (4%).No serious adverse effects observed.Atypically low prevalence of comorbidities in atypically young demographic.No taking into account of differences in behaviourassociated risk. Significantly higher probability of QTc prolongation in any treatment group as compared to baseline.Significantly higher mortality risk with QTc prolongation > 60 ms.No significant differences in the prevalence of specific arrhythmic events, such as atrial fibrillation, tachycardia or torsade de pointes.Lack of randomisation.Small sample size.Inclusion only of patients requiring hospitalisation increases risk of selection bias.Higher probability of ECG-monitored patients being in a specialised care unit with more testing availability.3% and 10% of patients were coadministered methadone and ondansetron, respectively, both of which prolong QTc, creating a confounder effect. 11% of patients ceased HCQ before day 5 due to arrhythmic and GI adverse events, as well as a case of hypoglycaemia.Lack of randomisation.Small sample size.Possible underestimation of QTc due to short followup period. Possible role of COVID-19-associated myocarditis and/or stress cardiomyopathy in observed adverse events.Baseline QTc was shorter in HCQ + azithromycin compared to HCQ only patients.Most patients had at least 1 cardiac comorbidity, and were taking 2 or more drugs prolonging QTc. Significantly longer time to viral clearance than patients given lopinavirritonavir 400 and 100 mg 2 day -1 , respectively.No significant difference in delay until improvement of disease symptoms. No significant difference in the incidence of adverse events.Significantly lower incidence of lymphopenia and hyperbilirubinaemia than in patients given lopinavirritonavir.Lack of randomisation.Small sample size.Significantly greater probability of confirmed pneumonia on CT scans in patients given lopinavirritonavir than those given HCQ. No observed side-effects of HCQ.No significant changes in plasma counts of leukocytes, lymphocytes, or concentrations of CRP and ferritin in either HCQ or control groups.Lack of randomisation.Small sample size.No direct measurement of QTc prolongation.HCQ group had both significantly higher comorbidities and D-dimer levels.Abbreviations: RCT = RCT; OS = observational study; SCD = sudden cardiac death; MCOT = mobile cardiac outpatient telemetry (2) acquisition of submitted, but not yet peerreviewed, items from a preprint server; and (3) application of exclusion and inclusion criteria.The removal of duplicates at each step, where necessary, left 26 unique items to pass onto the quality appraisal stage.