Chloroquine or hydroxychloroquine for prevention and treatment of COVID‐19

Summary of findings 1. Hydroxychloroquine (HCQ) compared to standard care or placebo for the treatment of people with COVID‐19

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Patients or population: adults with mild to severe COVID‐19 Settings: hospital inpatients and ambulatory care in the community Intervention: HCQ Comparison: standard care or placebo (no HCQ)
Outcomes	Risk with standard care or placebo	Risk with HCQ	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Death due to any cause	18 per 100	19 per 100 (18 to 21)	RR 1.09 (0.99 to 1.19)	8208 (9 RCTs)^a	⨁⨁⨁⨁ HIGH ^b,c	HCQ results in little or no difference to death due to any cause.
Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment^d	83 per 100	83 per 100 (76 to 91)	RR 1.00 (0.91 to 1.10)	213 (3 RCTs)^e	⨁⨁◯◯ LOW ^f,g	HCQ may make little or no difference to proportion of people having negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment.
Progression to mechanical ventilation	8 per 100	9 per 100 (7 to 11)	RR 1.11 (0.91 to 1.37)	4521 (3 RCTs)^h	⨁⨁⨁◯ MODERATE ^i,j	HCQ probably results in little to no difference in progression to mechanical ventilation.
Time to clinical improvement	28 per 100	28 per 100 (18 to 44)	HR 1.01 (0.59 to 1.74)	119 (1 RCT)^k	⨁◯◯◯ VERY LOW ^f,l,m	We are uncertain whether HCQ increases or decreases the proportion of people with clinical improvement at day 28 from enrolment.
Participants with any adverse events	16 per 100	46 per 100 (24 to 90)	RR 2.90 (1.49 to 5.64)	1394 (6 RCTs)ⁿ	⨁⨁⨁◯ MODERATE ^o,p	HCQ probably increases the risk of developing adverse events.
Participants with serious adverse events	36 per 1000	30 per 1000 (13 to 64)	RR 0.82 (0.37 to 1.79)	1004 (6 RCTs)^q	⨁⨁◯◯ LOW ^r	HCQ may result in little or no difference to risk of serious adverse events.
Participants with prolongation of QT‐interval on ECG	2 per 100	17 per 100 (2 to 100)	RR 8.47 (1.14 to 63.03)	147 (1 RCT)^s	⨁◯◯◯ VERY LOW ^t,u,v	The evidence is very uncertain about the effect of HCQ on prolongation of QT‐interval on ECG.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; ECG: electrocardiogram; HCQ: hydroxychloroquine; HR: hazard ratio; PCR: polymerase chain reaction RCT: randomized controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aAbd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020. Of these, no participants died in Chen 2020a; Chen 2020c; Mitjà 2020a; Tang 2020. ^bNot downgraded for risk of bias, as most of the evidence comes from Horby 2020 and Pan 2020, which have low risk of bias for this outcome. ^cNot downgraded for indirectness, but it is noted that the population in the largest trial, Horby 2020, was mostly severely/critically unwell. ^dThis was selected as the most relevant of three related outcomes reported by trials in this review. Analyses for the other outcomes (time to negative PCR for SARS‐CoV‐2 on respiratory samples; negative PCR for SARS‐CoV‐2 at day 7 from enrolment) did not demonstrate an important benefit/harm. ^eChen 2020a; Chen 2020c; Tang 2020. ^fDowngraded by one level for serious indirectness: almost all people had mild or moderate COVID‐19; all were hospitalized; and all were from one region. ^gNot downgraded for imprecision: narrow confidence interval, not including appreciable benefit nor harm. The sample size has approximately 80% power to detect an absolute difference of 13%, or 90% power to detect an absolute difference of 15%, in this outcome for the group receiving HCQ versus those receiving standard care. ^hCavalcanti 2020; Horby 2020; Tang 2020. ⁱNot downgraded for indirectness: the three trials all recruited participants admitted to hospital. ^jDowngraded by one level for serious imprecision: lower confidence interval bound represents no benefit nor harm from HCQ, whereas the upper bound suggests appreciable harm. ^kTang 2020. ^lDowngraded by one level for serious risk of bias: unclear risk of attrition and reporting bias, and high risk of other bias. ^mDowngraded by one level for serious imprecision: lower confidence interval bound represents appreciable harm from HCQ, whereas the upper bound suggests no appreciable benefit. ⁿCavalcanti 2020; Chen 2020a; Chen 2020b; Mitjà 2020a; Skipper 2020; Tang 2020. ^oDowngraded by one level for serious risk of bias: all trials except Skipper 2020 were open‐label. Chen 2020a had a high risk of selection and reporting bias; Chen 2020b a high risk of performance, detection, and reporting bias and unclear risk of selection bias; Mitjà 2020a a high risk of performance, detection, attrition, and reporting bias for this outcome, and unclear risk of selection bias; Skipper 2020 a high risk of reporting bias and unclear risk of attrition bias; and Tang 2020 an unclear risk of attrition and reporting bias. We deemed Skipper 2020, Mitjà 2020a, and Tang 2020 as at high risk of other bias. ^pNot downgraded for inconsistency: despite high statistical heterogeneity (I² = 87%), all of the effect estimates were above a risk ratio of 1, with only one trial having a confidence interval that crossed 1. ^qCavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Skipper 2020; Tang 2020. ^rDowngraded by two levels for very serious imprecision: low number of events, and lower confidence interval bound represents appreciable harm from HCQ, whereas the upper bound includes appreciable benefit. ^sCavalcanti 2020. ^tDowngraded by one level for risk of bias: Cavalcanti 2020 was unblinded, which could have led to detection bias, meaning more participants with QT prolongation were identified in the HCQ group. ^uDowngraded by one level for indirectness: Cavalcanti 2020 included only hospitalized patients, and excluded participants with severe disease, in whom problems with drug interactions and cardiac arrhythmia are more likely. ^vDowngraded by one level for imprecision: Cavalcanti 2020 had a low event rate for this outcome, and a small sample size leading to a very broad confidence interval.

Summary of findings 2. Hydroxychloroquine (HCQ) compared to placebo for the prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Patients or population: people who have been exposed to SARS‐CoV‐2 Settings: community Intervention: HCQ Comparison: placebo
Outcomes	Risk with placebo	Risk with HCQ	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Development of confirmed COVID‐19 at 14 days from enrolment	2 per 100	2 per 100 (1 to 6)	RR 1.20 (0.50 to 2.87)	821 (1 RCT)	⨁◯◯◯ VERY LOW ^a,b	The evidence is very uncertain about the effect of HCQ on development of confirmed COVID‐19 at 14 days from enrolment.
Hospitalized due to COVID‐19^c	2 per 1000	2 per 1000 (0 to 31)	RR 0.98 (0.06 to 15.66)	821 (1 RCT)	⨁◯◯◯ VERY LOW ^a,b	The evidence is very uncertain about the effect of HCQ on risk of being hospitalized due to COVID‐19.
Participants with any adverse events	17 per 100	41 per 100 (31 to 53)	RR 2.39 (1.83 to 3.11)	700 (1 RCT)	⨁⨁⨁◯ MODERATE ^a	HCQ probably increases the risk of adverse events when compared with placebo.
Participants with serious adverse events	0 per 1000	0 per 1000 (0 to 0)	Not estimable	700 (1 RCT)	⨁⨁◯◯ LOW ^a,d	HCQ may result in little or no difference in serious adverse events when compared with placebo.
The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect* of the intervention (and its 95% CI). CI: confidence interval; HCQ: hydroxychloroquine; RCT: randomized controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level for serious indirectness: one trial, limited to North America; few older and comorbid participants, possibly due to social media‐based recruitment and internet‐based data collection (Boulware 2020). ^bDowngraded by two levels for very serious imprecision: confidence interval around effect estimate includes appreciable benefit and appreciable harm. ^cThis outcome, as reported by Boulware 2020, was closest to our predefined outcome of 'disease severity of participants who develop COVID‐19, as defined by study authors'. ^dDowngraded by one level for imprecision: no events in either group, therefore risk ratio is not estimable. The optimal information size to be confident that this is a true reflection of risk of serious adverse events would be larger than the total number of participants in this trial. Risk difference = 0% (95% CI −1% to 1%).

Background

Description of the condition

Coronavirus disease 2019 (COVID‐19) is a viral infection transmitted by respiratory droplet spread. It is caused by severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2). COVID‐19 commonly presents as a mild respiratory tract illness, with fever and cough the most commonly reported symptoms; however, in some people this progresses to cause a life‐threatening respiratory syndrome (Guan 2020).

SARS‐CoV‐2 is a novel coronavirus that has caused a pandemic since December 2019. Over 27 million people have been diagnosed with COVID‐19, and as of 7 September 2020 over 890,000 people have died (JHU 2020). The World Health Organization (WHO) declared COVID‐19 a public health emergency of international concern on 30 January 2020, and a pandemic on 11 March 2020 (WHO 2020a).
National data from China and Italy describe severe disease in 14% to 20% of people with COVID‐19, and a further 2% to 5% are reported to have critical illness (ISS 2020; Wu 2020). Early mortality estimates ranged from around 2% to 12%, though this has varied considerably between countries and as the pandemic has progressed (ISS 2020; Wu 2020). Severe disease is characterized by hypoxia, and progressive acute respiratory distress syndrome appears to be the driver for mortality, although patients can experience a syndrome with clinical and laboratory features of severe systemic inflammation, termed a “cytokine storm” (Guan 2020; Mehta 2020).
At the other end of the spectrum, asymptomatic infection is not uncommon; national Italian data describe this in approximately 10% of all people with a confirmed COVID‐19 diagnosis (ISS 2020). More recently, wide‐ranging longer‐term morbidity has been described in the absence of a severe initial illness (Greenhalgh 2020).

Transmission is by direct contact with people with the infection, indirectly via contact with respiratory secretions on objects and surfaces, or from droplets generated by sneezing and coughing (WHO 2020b). Concerns have been raised about airborne transmission: viability of SARS‐CoV‐2 has been demonstrated for at least three hours when suspended in an aerosol (van Doremalen 2020). The amount of virus found in the respiratory tract appears to be higher in people with severe versus those with mild disease, with shedding of virus in the nasopharynx occurring for up to 25 days in people with severe disease (Liu 2020a). The virus has also been found in stools, with one study reporting live virus in non‐diarrhoeal stool, thus raising concerns about faecal‐oral transmission (Wang 2020a).
Multiple episodes of transmission by pre‐symptomatic or asymptomatic people have been described (Bai 2020; Rothe 2020).

The main method for diagnosis of COVID‐19 is by polymerase chain reaction (PCR) of respiratory tract samples, mostly from the nasopharynx or oropharynx. However, some guidelines advise nasal swabs (CDC 2020), and some evidence suggests lower respiratory samples, such as sputum, may have higher sensitivity (Wang 2020a). Serological tests are being used for detecting antibodies to SARS‐CoV‐2 for confirmation of past infection, although there are concerns regarding the evidence for their accuracy and value in certain populations and clinical situations (Deeks 2020).

Transmission is common in, though not limited to, households (Pung 2020). Self‐isolation, quarantine, and travel restrictions can limit community transmission (Kraemer 2020), but prevention measures within households can be more challenging. Healthcare workers are at high risk of being infected. Data from Italy show that 20% of frontline healthcare workers responding to the pandemic have developed COVID‐19 (Lancet 2020). There were widespread shortages of personal protective equipment (Lewis 2020). With established community transmission in many countries, healthcare workers are also at risk outside of health facilities. Despite vaccine roll‐out having commenced in some countries, achieving target coverage will take several months, and will not eliminate symptomatic infections in the near future. Consequently, there is great interest in using existing drugs as treatment for or prevention of COVID‐19.

Several potential antivirals have been suggested for use in treating people with COVID‐19. Remdesivir, a drug trialled for Ebola virus disease and Middle East respiratory syndrome (MERS), showed promising results in vitro (Wang 2020b). An early trial showed no benefit on time to clinical improvement, mortality, or clearance of the virus from the respiratory tract (Wang 2020c). Subsequently, two randomized trials have reported a beneficial effect of remdesivir on measures of clinical improvement in patients hospitalized with COVID‐19, but no significant effect on mortality (Beigel 2020; Spinner 2020). Other experimental antivirals being studied include the influenza treatments umifenovir (Arbidol), Deng 2020, and favipiravir, Cai 2020, and the antiretroviral protease inhibitor combination lopinavir/ritonavir (Cao 2020). Of the many other options being investigated, corticosteroids are now recommended by WHO for patients with COVID‐19 requiring oxygen or higher respiratory support therapy (WHO 2020d), having been reported to reduce mortality in this population in a systematic review (REACT 2020). Other options that have yet to show benefit in randomized trials are tocilizumab (Stone 2020), convalescent plasma (Agarwal 2020), and camostat mesylate (Hoffman 2020). Several studies have used novel methods to assess whether existing drugs can be repurposed for COVID‐19 treatment (Chandel 2020; Zhou 2020).

Description of the intervention

Chloroquine (CQ) and hydroxychloroquine (HCQ) are 4‐aminoquinoline compounds, derivatives of quinine, and have been used as antimalarial drugs since the 1940s (Ben‐Zvi 2012). HCQ is an analogue of CQ in which one of the N‐ethyl substituents of CQ is β‐hydroxylated. HCQ and CQ have similar pharmacokinetic properties, with high oral bioavailability and tissue penetrance, partial hepatic metabolism, and high volumes of distribution as they diffuse into adipose tissue (Ben‐Zvi 2012).

Both drugs have been used widely and for many years for the treatment and prevention of malaria (although they are now largely ineffective against falciparum malaria) and in the treatment of rheumatological conditions, such as systemic lupus erythematosus and rheumatoid arthritis (Fiehn 2020; Steinhardt 2011).

The mechanism of action in malaria is thought to result from inhibition of the biocrystallization of hemozoin, causing cytotoxic accumulation of heme (Schrezenmeier 2020). For rheumatological conditions, the mechanism of action is not fully delineated, but appears to arise from multiple effects. As weak bases, both CQ and HCQ accumulate in the acidic environment within lysosomes, and thus interfere with lysosomal activity and autophagy, which in turn may inhibit major histocompatibility complex (MHC) class II expression and antigen presentation, inhibiting immune activation (Schrezenmeier 2020). CQ and HCQ also interfere with Toll‐like receptor (TLR) signalling, again via changes to local pH but also through direct binding to nucleic acids. TLR signal pathways stimulate cytokine production, and CQ and HCQ have been demonstrated to inhibit production of various cytokines including interleukin (IL)‐1, IL‐6, tumour necrosis factor (TNF), and interferon gamma (IFNγ) by mononuclear cells (van den Borne 1997).

CQ and HCQ have well‐described adverse effect profiles. Common adverse effects include gastrointestinal upset and headache (Ben‐Zvi 2012). Several adverse effects are associated with chronic therapy, such as QT‐interval prolongation on electrocardiogram, other cardiac arrhythmia, and retinopathy (Fiehn 2020). CQ is generally less tolerable than HCQ, and can cause acute poisoning at a lower dose, as has been seen in reports from the USA and Nigeria of members of the public taking CQ without a prescription (CNN 2020; Owens 2020).

There are two types of CQ salts: CQ phosphate and CQ sulphate. Most dosing recommendations for CQ refer to the salt rather than the base compound. Usual doses for CQ are 250 mg to 500 mg CQ phosphate (155 mg to 310 mg CQ base) per dose, or CQ sulphate 200 mg (150 mg CQ base), with weekly dosing for malaria prophylaxis, and daily dosing for treatment of malaria and rheumatological conditions. HCQ is given at a dose of 400 mg weekly for malaria prophylaxis, and 200 mg to 400 mg daily for rheumatological disease (Ben‐Zvi 2012).

How the intervention might work

Some researchers have suggested that both CQ and HCQ could be clinically effective against COVID‐19. Studies have reported in vitro activity against SARS‐CoV‐2 (Liu 2020b; Wang 2020b; Yao 2020), and pharmacokinetic modelling suggests efficacy of a few postulated dosing regimens for treatment (Yao 2020).

Liu 2020b reported that CQ and HCQ appear to inhibit transport of SARS‐CoV‐2 virions from early endosomes to endolysosomes in Vero E6 cells, which may be a requirement for release of the viral genome and subsequent viral replication. Wang 2020b performed a "time‐to‐addition" assay using Vero E6 cells and found that CQ appeared to both inhibit entry of SARS‐CoV‐2 into cells and inhibit viral replication after cell entry. The authors of both studies also speculate that CQ and HCQ could impact on disease severity in COVID‐19 through modulating the excess cytokine release that appears to contribute to life‐threatening forms of the disease (Liu 2020b; Wang 2020b).

Why it is important to do this review

Given the pace of the pandemic, and the extraordinary impact on public health and society in many countries, there is high demand for effective prevention and treatment interventions for COVID‐19. CQ and HCQ are inexpensive drugs that are registered in most countries, and are included on the WHO essential medicines list (WHO 2019). They can be delivered orally, and both drugs have well‐described safety profiles in adults and children. Given the uncertain effects of antiviral drugs for treatment of COVID‐19, or the effectiveness of the newly developed vaccines, identifying existing medicines that may be of benefit is of high importance. Despite the small number of published studies, some governments have recommended using HCQ as prophylaxis for healthcare workers, and some prominent political figures have asserted that CQ or HCQ should be used as a first‐line treatment for COVID‐19. Sadly, there have already been instances of significant harm where individuals have misinterpreted news stories about the use of CQ and suffered toxicity as a result (CNN 2020).

CQ and HCQ for treatment of COVID‐19

Earlier national guidelines, mostly in February to April 2020, recommended CQ or HCQ for the treatment of individuals with COVID‐19. Belgian guidelines recommended HCQ for severe disease, and advised that it be considered for mild‐moderate disease (WIV‐ISP 2020). Chinese guidelines advised consideration of CQ in all hospitalized patients, although later iterations have expressed caution regarding dosing and special patient groups (Wong 2020). Italian guidelines recommended early use of CQ or HCQ, or lopinavir/ritonavir (Brescia‐COVID Group 2020). More recently, concerns about adverse effects have led to removal of recommendations to use CQ and HCQ from several national guidelines, alongside which the US Food and Drug Administration revoked their initial emergency use authorization provided for use of CQ and HCQ in the treatment of COVID‐19 (FDA 2020), and the UK Medicines and Healthcare products Regulatory Agency enforced suspension of recruitment to trial arms using CQ or HCQ as an intervention (Robinson 2020).

Initial observational studies reported efficacy of CQ and HCQ. A widely publicized small, non‐randomized study from Marseille, France, reported that HCQ was associated with earlier negative PCR for SARS‐CoV‐2 among 20 patients given HCQ compared to those who had refused to take HCQ or who had presented to other hospitals (Gautret 2020a). Subgroup analyses reported quicker clearance of the virus for six participants who had azithromycin in combination with HCQ versus those who received neither drug (Gautret 2020a). There has been widespread criticism of the methods, reporting, and conclusions of this study (Machiels 2020). The same group then published two observational single‐arm cohorts of patients treated with HCQ plus azithromycin, reporting benefit of the combination (Gautret 2020b; Million 2020). Soon after this, another research group from France reported much poorer clinical and virological outcomes in 11 hospitalized patients treated with both drugs (Molina 2020). A quasi‐experimental study of patients admitted with moderate COVID‐19 in four French hospitals reported no difference in efficacy outcomes, but reported early discontinuation of HCQ in 9 of 84 participants due to abnormalities on electrocardiography (Mahévas 2020).

More recently, a number of larger non‐randomized studies have reported beneficial effects of HCQ. A retrospective cohort study in Michigan, USA compared four groups of a total of 2541 patients hospitalized with confirmed COVID‐19 according to physician‐directed treatment assignment: 1202 received HCQ; 147 azithromycin alone; 783 HCQ with azithromycin; and 409 received neither drug (Arshad 2020). A significant reduction in mortality was reported when HCQ was received (hazard ratio (HR) 0.49, 95% confidence interval (CI) 0.29 to 0.83). Differences in baseline characteristics suggested underlying confounding, although an underpowered propensity‐matching analysis reported persistence of the reported mortality benefit (Arshad 2020). The quantity of missing data and early patient exclusions were not reported (Arshad 2020). Another study retrospectively comparing 4542 patients in Belgian hospitals reported lower risk of death in the group who received HCQ as per national guidance (804/4542, 17.7%) versus 3533 patients who did not receive HCQ (957/3533, 27.1%) (Catteau 2020). After adjusting for multiple covariates, this difference was found to be statistically significant (adjusted HR 0.68, 95% CI 0.62 to 0.76) (Catteau 2020). Of note, nearly 50% of patients screened for eligibility were excluded, though some of these patients were found to have similar baseline characteristics to those included in the analysis (Catteau 2020).

At the time of writing the protocol for this review, China had reported two small randomized trials of HCQ, with mixed results (Chen 2020a; Chen 2020b). Several trials have since been reported and are included in this review.

CQ and HCQ for preventing COVID‐19

Despite no human data on prophylaxis early in the pandemic, the Indian Council of Medical Research (ICMR) recommended HCQ as pre‐exposure prophylaxis for frontline healthcare workers having “high‐risk” contact with patients with suspected or confirmed COVID‐19, and postexposure prophylaxis for household and healthcare worker contacts of patients with confirmed COVID‐19 (ICMR 2020). The background section of this recommendation referred to in vivo evidence for efficacy of HCQ for the treatment of COVID‐19, and inferred prophylactic efficacy from therapeutic efficacy (ICMR 2020). Concerns have been raised by multiple groups regarding this approach (Rathi 2020).

Since then, two comparative studies have reported the effect of use of CQ or HCQ for prophylaxis of COVID‐19, one of which is a randomized trial (Boulware 2020), and the other a case‐control study conducted by the ICMR (Chatterjee 2020). The former is included in this review. The case‐control study involved a telephone survey of healthcare workers tested for SARS‐CoV‐2 when suspected of having symptomatic COVID‐19: the 378 cases (172 of whom took HCQ) had a positive PCR test for SARS‐CoV‐2, whilst 373 controls (193 of whom used HCQ) had a negative test (Chatterjee 2020). Whilst use of HCQ versus no use of HCQ was not found to be significantly associated with confirmed COVID‐19, a dose‐response effect was reported, with lower odds of positive PCR the higher the number of weekly doses reported to have been taken: for four or five maintenance doses of HCQ after an initial loading dose, the adjusted odds ratio using multivariate regression analysis was 0.44 (95% CI 0.22 to 0.88) (Chatterjee 2020). Reported side effects were uncommon. Methods were reported incompletely, such as the sampling approach for cases and controls from the database of 21,402 healthcare workers, of whom 1073 has a positive PCR test (Chatterjee 2020). The target sample size was not met, though this was calculated for HCQ prophylaxis as a binary exposure variable, rather than the duration‐based groups used in the eventual analysis (Chatterjee 2020). Several trials exploring the use of CQ or HCQ for prophylaxis of COVID‐19 are ongoing (Cortegiani 2020).

Adverse events have been a particular concern with CQ and HCQ. Studies using data from pharmacovigilance databases prior to the use of these drugs, and azithromycin, have suggested caution regarding even short‐term use due to their association with cardiac adverse effects (Nguyen 2020; Singh 2020). A randomized trial comparing higher‐dose CQ (41 participants) versus lower‐dose CQ (40 participants) in patients hospitalized with severe COVID‐19 in northern Brazil was stopped early by the independent safety monitoring board due to higher death and cardiac serious adverse events, including QT‐interval prolongation on electrocardiogram, in the group receiving higher‐dose CQ (Borba 2020). An article published in The Lancet reporting higher incidence of death and serious adverse events in patients receiving CQ or HCQ with or without a macrolide drug (azithromycin or clarithromycin), as documented in a large international surgical registry. The Lancet later retracted this when the data and analysis were questioned, though regulatory authorities and trial steering groups had already decided to stop trials or trial arms investigating CQ and HCQ (Mehra 2020).

At the time of development of the protocol for this review, other systematic reviews had already been produced. Due to the intense interest in finding a therapeutic that is safe and effective for COVID‐19, many review papers have been published over the last six months. Reviews published early in the outbreak relied on pre‐clinical data, expert commentary, and small, mostly non‐randomized studies. A systematic review of CQ for the treatment of COVID‐19, which searched PubMed and Embase up to 1 March 2020, identified no published studies other than the aforementioned letter (Gao 2020), though 23 clinical trials of CQ or HCQ were found on registries (Cortegiani 2020). Another systematic review of CQ and HCQ for treating COVID‐19, published as a preprint on 30 March 2020, concluded: “There is theoretical, experimental, preclinical and clinical evidence of the effectiveness of chloroquine in patients affected with COVID‐19” (Kapoor 2020). A further review included one non‐randomized study and one randomized trial, and concluded: "Without further evidence, HCQ is not appropriate for patients with COVID‐19 in primary care" (McCormack 2020). A systematic review of antimalarials (CQ and HCQ) for the treatment of COVID‐19 was produced by the Epistemonikos Working Group, which synthesized the results of two small randomized trials and found low‐ to very low‐certainty evidence regarding efficacy and harms (Epistemonikos 2020).

We propose that, in this context, a systematic review of randomized controlled trials using standard Cochrane methods that provides summary estimates of effects for both treatment and prophylactic use of CQ and HCQ, with an appraisal of the certainty of the evidence using the GRADE approach, is important for the general public, clinicians, and policymakers. We plan to update this review in an expedited fashion as new data become available from the trials that are currently in progress on prevention.

Objectives

To evaluate the effects of chloroquine (CQ) or hydroxychloroquine (HCQ) as:

an antiviral treatment on death and time to clearance of the virus from clinical samples in people with COVID‐19;
a prophylactic treatment on prevention of COVID‐19 in people at risk of SARS‐CoV‐2 exposure;
a prophylactic treatment on prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2.

Methods

Criteria for considering studies for this review

Types of studies

Randomized controlled trials (RCTs).

Types of participants

Objective 1. People who have COVID‐19, as defined by study authors.

Objective 2. People who are at risk of SARS‐CoV‐2 exposure, as defined by study authors.

Objective 3. People who have been exposed to SARS‐CoV‐2, as defined by study authors.

Types of interventions

Intervention

Chloroquine (CQ) or hydroxychloroquine (HCQ) given by any route of administration and dose used alone or in combination with other treatments.

Control

No treatment, supportive treatment, or other experimental antiviral treatment (i.e. any other treatment that does not contain CQ or HCQ).

Types of outcome measures

Objective 1. For treatment of COVID‐19 disease

Primary outcomes

Death
Time to negative PCR for SARS‐CoV‐2 on respiratory samples

Secondary outcomes

Number of participants admitted to hospital (if receiving ambulatory treatment)
Number of participants requiring mechanical ventilation
Length of hospital admission
Time to clinical improvement, as defined by study authors
Duration of mechanical ventilation postenrolment in survivors of COVID‐19

Objectives 2 and 3. For prevention of COVID‐19 disease in people at risk of exposure/who have been exposed to SARS‐CoV‐2

Primary outcomes

Development of confirmed COVID‐19, as defined by study authors
Production of antibodies to SARS‐CoV‐2

Secondary outcomes

Development of COVID‐19 in household contacts of the recipient of the prophylaxis
Disease severity of participants who develop COVID‐19, as defined by study authors

Adverse events (relating to objectives 1, 2, and 3)

All adverse events
All serious adverse events attributed to study drug (i.e. serious adverse effects)
QT‐interval prolongation

Search methods for identification of studies

We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press, and in progress) up to 15 September 2020.

Electronic searches

We searched the following databases on 15 September 2020 using the search terms and strategy described in Appendix 1: the Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library, up to Issue 9 of 12, September 2020; MEDLINE (PubMed) (1966 to 15 September 2020); and Embase (1974 to 15 September 2020). We also searched Current Controlled Trials (www.controlled-trials.com) and the World Health Organization International Clinical Trials Registry Platform (www.who.int/clinical-trials-registry-platform) using 'chloroquine', 'hydroxychloroquine', 'coronavirus', and 'COVID‐19' as search terms on 15 September 2020. We also searched COVID‐specific resources COVID‐NMA (www.covid-nma.com) and the Cochrane COVID‐19 Study Register (covid-19.cochrane.org/), which are updated daily with lists of ongoing and published trials, using 'chloroquine' and 'hydroxychloroquine' on 15 September 2020.

Searching other resources

We contacted researchers in the field to identify any unpublished or ongoing studies.

Data collection and analysis

Two review authors (BS and HR, MC, or TK) independently conducted each step of study selection and data extraction. Any disagreements were resolved through discussion.

Selection of studies

Two review authors (BS and HR or MC) independently screened the search results using Covidence (Covidence), and retrieved the full‐text articles of all potentially relevant trials. We examined each trial report to ensure that we included multiple publications from the same trial only once. We planned to contact trial authors for clarification if eligibility of a trial was unclear. Any disagreements were resolved through discussion. We listed the excluded studies and the reasons for their exclusion in the 'Characteristics of excluded studies' table. The study selection process is illustrated in a PRISMA diagram (Figure 1).

Figure 1

Study flow diagram.

Data extraction and management

Two review authors (BS and HR, MC, or TK) used a piloted data extraction form to extract data on participant characteristics, diagnostic criteria, disease severity, comorbidity, CQ or HCQ dose and administration, other treatments given, and outcome measures. Any disagreements were resolved through discussion. We contacted the corresponding trial author in the case of unclear or missing data.

For dichotomous outcomes, we recorded the number of participants that experienced the event and the number of participants randomized to each treatment group. We recorded the number of participants analysed in each treatment/prophylaxis arm, and used the discrepancy between the figures to calculate the number of participants lost to follow‐up, which would allow us to perform sensitivity analyses to investigate the effect of missing data if necessary. For continuous outcomes, we planned to extract means for the outcome in each group; we also recorded medians for narrative comparisons where means were unavailable.

Assessment of risk of bias in included studies

Two review authors (BS and HR, MC, or TK) assessed the methodological quality of studies using the Cochrane 'Risk of bias' tool, and reported the results in a 'Risk of bias' figure (Higgins 2011). We classified each 'Risk of bias' domain as either at high, low, or unclear risk of bias (Higgins 2011). We assessed the risk of bias associated with blinding for each outcome separately and used these judgements in the GRADE assessment, but made an overall judgement in the 'Risk of bias' assessment for each study based on the primary outcome as stated by the study authors. For other domains we assessed the risk of bias for the trial as a whole. We planned to attempt to contact the trial authors if information was not specified or was unclear. Any disagreements were resolved by discussion between the review authors.

Measures of treatment effect

We presented dichotomous outcomes as risk ratios (RR) with 95% confidence intervals (CIs). We reported continuous outcomes as mean differences (MD) with 95% CIs if the outcomes were measured in the same way across all included trials. If included trials measured continuous outcomes in different ways, we would use the standardized mean difference (SMD) and 95% CI as the effect measure. If using the SMD, we would re‐express the SMD in the units of one or more of the specific measurement instruments used in the original studies, to aid interpretation. We presented time‐to‐event outcomes as hazard ratios (HRs) and 95% CIs.

Unit of analysis issues

We did not anticipate that any cluster‐randomized studies would meet our inclusion criteria. In the case that cluster‐randomized studies did meet our inclusion criteria, we would ensure appropriate analysis adjusting for the effect of cluster randomization was carried out before including effects estimates in a meta‐analysis. If available, we planned to extract adjusted measures of effect from the trial reports. If only unadjusted data were available, we would adjust these data ourselves using the intracluster correlation coefficient (ICC). If the ICC was not reported, we would contact the study authors to obtain it, or borrow an ICC value from a similar study, or estimate the ICC. If the ICC was estimated, we would perform sensitivity analyses to investigate the robustness of our analyses.

If we identified multi‐arm trials, we would either select relevant arms for inclusion in our analyses, or if more than two arms were relevant to this review, we would either combine intervention arms so that there was one comparison, or split the control group between multiple comparisons so that participants are not double‐counted in meta‐analysis.

We did not anticipate that any cross‐over trials for treatment of COVID‐19 would meet our inclusion criteria, as cross‐over trials are used to evaluate interventions that have a temporary effect in the treatment of stable, chronic conditions.

We also thought it unlikely that cross‐over trials would have been conducted for the prevention of COVID‐19, due to the long half‐life of CQ/HCQ, meaning that a long wash‐out period would be required. It is also unknown whether the effects of receiving CQ or HCQ in the first period of the trial may have an irreversible effect that would subsequently impact outcomes in the second period of the trial. If we identified cross‐over trials for the prevention of COVID‐19, we would include data from the first period of the trial only. We would carefully consider whether studies that reported data only for the first period of a cross‐over trial were at risk of bias, and whether the omission of studies that did not report data from the first period of the trial (i.e. only a paired analysis was reported) would lead to bias at the meta‐analysis level.

Dealing with missing data

The primary analysis for efficacy outcomes was an available‐case analysis where the denominator is the number of patients completing follow‐up to the point of outcome assessment, where possible. Where this was not possible, we performed an intention‐to‐treat analysis, with investigation of the effects of missing data. For safety outcomes, we planned to include all participants receiving at least one dose of the intervention drug or placebo.

We planned to carry out sensitivity analyses to explore the impact of missing data on the primary outcomes. For dichotomous outcomes, we planned to vary the event rate within the missing patients from intervention and control groups within plausible limits. For continuous data, we planned to also perform sensitivity analyses using the methods described by Ebrahim 2013 and Ebrahim 2014.

Assessment of heterogeneity

We assessed heterogeneity by visually inspecting the forest plots to determine closeness of point estimates with each other and overlap of CIs. We used the Chi² test with a P value of 0.10 to indicate statistical significance, and the I² statistic to measure heterogeneity. We used the following ranges outlined in the Cochrane Handbook for Systematic Reviews of Interventions to interpret the I² statistic (Higgins 2019):

0% to 40%: might not be important;
30% to 60%: may represent moderate heterogeneity;
50% to 90%: may represent substantial heterogeneity;
75% to 100%: considerable heterogeneity.

We also considered the magnitude and direction of effects, and the strength of evidence for heterogeneity (e.g. P value from the Chi² test), when determining the importance of the observed I² value.

Assessment of reporting biases

We planned to construct a funnel plot to investigate any potential reporting bias if 10 or more studies were included for a given outcome.

Data synthesis

We analysed the data using Review Manager Web (RevMan Web 2019). We performed all meta‐analyses using random‐effects models. Where a meta‐analysis was not appropriate due to important clinical or methodological heterogeneity, or if study results differed to the extent that combining them in a pooled analysis would not make sense, we summarized data in tables.

Subgroup analysis and investigation of heterogeneity

We planned to investigate heterogeneity by performing the following subgroup analyses for people with COVID‐19.

Disease severity at presentation
Time in the illness when treatment started (< 7 days, and ≥ 7 days after symptoms started)
Comorbidity, such as cardiovascular disease, diabetes, and immunosuppression
Age
Sex
Admitted to hospital versus receiving ambulatory/outpatient treatment
CQ or HCQ dosing regimen

We planned to investigate heterogeneity by performing the following subgroup analyses for people exposed to SARS‐CoV‐2 or at risk of exposure to SARS‐CoV‐2.

Healthcare workers
Household contacts
Laboratory staff
Age
Comorbidity, such as cardiovascular disease, diabetes, and immunosuppression

Sensitivity analysis

To explore the possible effect of losses to follow‐up on the effect estimates for the primary outcomes, we planned to perform sensitivity analyses. For dichotomous outcomes, we planned to vary the event rate within the missing patients from intervention and control groups within plausible limits. For continuous data, we planned to perform sensitivity analyses using the methods described by Ebrahim 2013 and Ebrahim 2014.

Summary of findings and assessment of the certainty of the evidence

We summarized the results of the analysis in 'Summary of findings' tables, and presented the summary effects estimates for the primary outcomes and other important outcomes with illustrative comparative risks. We used the GRADE framework to evaluate the certainty of evidence for each outcome, as developed by the GRADE Working Group and described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019).

Results

Description of studies

Results of the search

Our searches identified 953 records, 93 of which were excluded as duplicate records. Of the remaining 860 records, we excluded 603 based on the assessment of titles and abstracts. We retrieved 257 full‐text publications to assess for inclusion. The screening process is illustrated in a flow diagram in Figure 1.

Ongoing studies

From our search on 15 September 2020 and reviewing the COVID‐NMA website, we identified 122 ongoing trials registered for treatment or prevention of COVID‐19. Due to the pressures of the pandemic and fluctuating interest in CQ and HCQ, many trials have been suspended or terminated, or had significant changes in protocol. We have therefore presented a summary of those ongoing trials that are reported to be recruiting actively, or that have completed recruitment but are yet to publish, and have a target recruitment of 500 or more participants, in tables (Table 1 for 22 ongoing treatment trials; Table 2 for 15 ongoing prevention trials). Up‐to‐date lists of ongoing trials can be found at www.covid-nma.com, updated daily.

Table 1. Ongoing trials for treatment: actively recruiting or completed; not yet published

Trial registration number; trial registry	Location(s)	Interventions; abbreviated name	Recruitment status	Estimated completion	Target enrolment
NCT02735707 ClinicalTrials.gov	13 countries; registered in the Netherlands	Adaptive platform trial including HCQ, or HCQ + lopinavir/ritonavir, vs no HCQ REMAP‐CAP	Recruiting	December 2021	7100
NCT04351724 ClinicalTrials.gov	Austria	Platform trial including CQ/HCQ vs placebo ACOVACT	Recruiting	December 2020	500
NCT04328012 ClinicalTrials.gov	USA	Pragmatic adaptive HCQ vs lopinavir/ritonavir vs losartan vs placebo COVID MED	Recruiting	January 2021	4000
NCT04334382 ClinicalTrials.gov	USA	HCQ vs azithromycin HyAzOUT	Recruiting	December 2020	1550
NCT04332991 ClinicalTrials.gov	USA	HCQ vs placebo for hospitalized patients with COVID‐19 ORCHID	Completed	April 2021	510
NCT04363827 ClinicalTrials.gov	Italy	HCQ vs observation PROTECT	Recruiting	September 2020	2300
NCT04359953 ClinicalTrials.gov	France	HCQ vs telmisartan vs azithromycin	Recruiting	June 2021	1600
NCT04356495 ClinicalTrials.gov	France	HCQ vs favipiravir vs imatinib vs telmisartan vs placebo COVERAGE	Recruiting	July 2020	1057
PACTR202004801273802 Pan African Clinical Trials Registry	Nigeria	CQ vs HCQ vs placebo	Recruiting	October 2020	600
ISRCTN86534580 ISRCTN registry	UK	HCQ vs standard care for treatment	Recruiting	March 2021	3000
NCT04324463 ClinicalTrials.gov	Canada	Azithromycin plus hydroxychloroquine or chloroquine (AZCT) vs AZCT plus interferon beta vs interferon beta vs usual care	Recruiting	September 2020	1500
NCT04345289 ClinicalTrials.gov	Denmark	Convalescent plasma vs sarilumab vs HCQ vs baricitinib vs intravenous and subcutaneous placebo vs oral placebo	Recruiting	June 2021	1500
NCT04358068 ClinicalTrials.gov	USA and Puerto Rico	HCQ vs azithromycin	Completed	October 2020	2000
NCT04340544 ClinicalTrials.gov	Germany	HCQ vs placebo	Recruiting	November 2021	2700
NCT04338698 ClinicalTrials.gov	Pakistan	HCQ vs oseltamivir vs azithromycin	Recruiting	September 2020	500
NCT04353037 ClinicalTrials.gov	USA	HCQ vs placebo	Recruiting	April 2021	850
NCT04321616 ClinicalTrials.gov	Norway	HCQ vs remdesivir vs standard care	Recruiting	August 2020	700
ACTRN12620000445976 ANZCTR	Australia and New Zealand	HCQ vs lopinavir/ritonavir vs HCQ plus lopinavir/ritonavir vs standard care	Recruiting	Not reported	2500
NCT04315948 ClinicalTrials.gov	France and Luxembourg	HCQ vs remdesivir vs lopinavir/ritonavir vs interferon beta‐1A vs standard care	Recruiting	March 2023	3100
UTN‐A27736297878 Ensaiosclinicos.gov.br	Brazil	HCQ vs placebo	Recruiting	July 2020	1300
NCT04410562 ClinicalTrials.gov	Spain	HCQ vs placebo (pregnant women)	Recruiting	May 2021	714
NCT04392973 ClinicalTrials.gov	Saudi Arabia	HCQ with favipiravir vs standard care	Recruiting	November 2021	520

CQ, chloroquine; HCQ, hydroxychloroquine

Table 2. Ongoing trials for prevention: actively recruiting or completed; not yet published

Trial registration number; trial registry	Location(s)	Interventions; population; abbreviated name	Recruitment status	Estimated completion	Target enrolment
NCT04333732 ClinicalTrials.gov	USA	Low‐/medium‐/high‐dose chloroquine vs placebo Healthcare workers	Recruiting	February 2021	55,000
NCT04303507 ClinicalTrials.gov	Europe, Asia, Africa	HQC vs CQ vs placebo Healthcare workers COPCOV	Recruiting	April 2021	40,000
NCT04334928 ClinicalTrials.gov	Spain	Emtricitabine/tenofovir (Truvada) vs HCQ vs Truvada + HCQ vs placebo Healthcare workers EPICOS	Recruiting	June 2020	4000
NCT04334148 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	July 2020	15,000
NCT04363450 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers (pre‐exposure) HCQPreP	Recruiting	July 2020	1700
NCT04318444 ClinicalTrials.gov	USA	HCQ vs placebo Household contacts (postexposure)	Recruiting	March 2021	1600
NCT04341441 ClinicalTrials.gov	USA	Daily HCQ vs weekly HCQ vs placebo Healthcare workers and first responders	Recruiting	June 2020	3000
IRCT20190122042450N4 Iranian Clinical Trials Registry	Iran	HCQ vs no HCQ All contacts (postexposure)	Completed	Not reported	1000
ISRCTN14326006 ISRCTN registry	Canada	HCQ vs placebo Healthcare workers	Recruiting	January 2022	988
NCT04363827 ClinicalTrials.gov	Italy	HCQ vs no HCQ All contacts	Recruiting	September 2020	2300
NCT04352933 ClinicalTrials.gov	UK	HCQ weekly vs HCQ daily vs placebo Healthcare workers	Recruiting	October 2020	1000
NCT04353037 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	April 2021	850
ACTRN12620000501943 ANZCTR	Australia	HCQ vs placebo Healthcare workers	Recruiting	December 2020	2250
NCT04374942 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	January 2022	988
EudraCT 2020‐001987‐28 EudraCT	Italy	HCQ vs no HCQ Healthcare workers	Recruiting	Not reported	1000

CQ, chloroquine; HCQ, hydroxychloroquine

Included studies

We included 14 RCTs with a total of 11,915 participants. Further details of the trials are provided in subsections for each of the review's objectives. A summary description is provided in Table 3, with more details in the Characteristics of included studies section.

Table 3. Summary of characteristics of included studies

Study

Objective; comparisons

Study design

Countries; recruitment dates

Age

Number of participants in primary comparison

Types of participant at enrolment (type of contact; place of care; disease severity)

Abd‐Elsalam 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Egypt

March to June 2020

HCQ: mean 40.4 y (SD 18.7 y)

Standard care: mean 41.1 y (SD 20.1 y)

194 total: 97 HCQ; 97 standard care

All hospitalized.

“The patients were randomized equally between the two groups regarding the disease severity.” (Numbers not reported.)

Boulware 2020

3: Postexposure prophylaxis

5: HCQ vs placebo (individually randomized)

RCT, double‐blind

USA and Canada

17 March to 6 May 2020

HCQ: median 41 y (IQR 33 to 51)

Placebo: median 40 y (IQR 32 to 50)

821 total: 414 HCQ; 407 placebo

HCQ: 275 healthcare contacts; 125 household contacts; 14 NR

Placebo: 270 healthcare contacts; 120 household contacts; 17 NR

Cavalcanti 2020

1: Treatment

1: HCQ vs standard care

3: HCQ + azithromycin vs standard care

RCT, open‐label

Brazil

29 March to 17 May 2020

HCQ + azithromycin: mean 49.6 y (SD 14.2 y)

HCQ: mean 51.3 y (SD 14.5 y)

Standard care: mean 49.9 y (SD 15.1 y)

665 total: 217 HCQ + azithromycin; 221 HCQ; 227 standard care

All hospitalized.

HCQ + azithromycin: 125/217 mild; 92/217 moderate disease

HCQ: 132/221 mild; 89/221 moderate disease

Standard care: 130/227 mild; 97/227 moderate disease

Chen 2020a

1: Treatment

1: HCQ vs standard care

RCT, open‐label

China

6 February to 25 February 2020

HCQ: mean 50.5 y (SD 3.8 y)

Standard care: mean 46.7 y (SD 3.6 y)

30 total: 15 HCQ; 15 standard care

All hospitalized.

All 30 participants had moderate disease.

Chen 2020b

1: Treatment

1: HCQ vs standard care

RCT, double‐blind (no placebo)

China

4 February to 28 February 2020

HCQ: mean 44.1 y (SD 16.1 y)

Standard care: mean 45.2 y (SD 14.7 y)

62 total: 31 HCQ; 31 standard care

All hospitalized.

All 62 participants had mild disease.

Chen 2020c

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Taiwan

1 April to 31 May 2020

HCQ: mean 33 y (SD 12 y)

Standard care: mean 32.8 y (SD 8.3 y)

33 total: 21 HCQ; 12 standard care

All hospitalized.

HCQ: 19/21 mild; 2/21 moderate

Standard care: 10/12 mild; 2/12 moderate

Davoodi 2020

1: Treatment

4: HCQ vs febuxostat

RCT, open‐label

Iran

16 March to 10 April 2020

HCQ: mean 57.3 y (standard error 2.2 y)

Febuxostat: mean 58 y (standard error 1.47 y)

54 total: 25 HCQ; 29 febuxostat

All ambulatory patients, symptomatic, with abnormalities on CT scan of the chest, but no features of severe acute illness or severe underlying chronic disease.

Horby 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

25 March to 5 June 2020

HCQ: mean 65.2 y (SD 15.2 y)

Standard care: mean 65.4 y (SD 15.4 y)

4716 total: 1561 HCQ; 3155 standard care

All hospitalized.

Inferred from level of oxygen/respiratory support need:

HCQ: 362/1561 asymptomatic/mild (no oxygen received); 938/1561 moderate/severe (received oxygen); 261/1561 critical disease (invasive ventilation)

Standard care: 750/3155 asymptomatic/mild (no oxygen received); 1873/3155 moderate/severe (received oxygen); 532/3155 critical disease (invasive ventilation)

Huang 2020

1: Treatment

2: CQ vs lopinavir/ritonavir (LPV/r)

RCT, open‐label

China

27 January to 15 February 2020

CQ: median 41.5 y (IQR 33.8 to 50 y)

LPV/r: median 53 y (IQR 41.8 to 63.5 y)

22 total: 10 CQ; 12 LPV/r

All hospitalized.

CQ: 7/10 moderate; 3/10 severe disease

LPV/r: 7/12 moderate; 5/12 severe disease

Mitjà 2020a

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Spain

17 March to 26 May 2020

HCQ: mean 41.6 y (SD 12.4 y)

Standard care: mean 41.7 y (SD 12.6 y)

293 total: 136 HCQ; 157 standard care

All ambulatory patients with mild disease, except for 1 patient with severe disease included in the HCQ arm, despite this being an exclusion criterion (included in ITT analysis).

Mitjà 2020b

3: Postexposure prophylaxis

6: HCQ vs standard care (cluster randomized)

Cluster‐RCT, open‐label

Spain

17 March to 28 April 2020

HCQ: mean 48.6 y (SD 18.7 y)

Standard care: mean 48.7 y (SD 19.3 y)

2525 total: 1225 HCQ; 1300 standard care

HCQ: 131 (12%) healthcare workers; 302 (27%) household contacts; 550 (49%) nursing home workers; 133 (12%) nursing home residents

Standard care: 130 (11%) healthcare workers; 338 (28%) household contacts; 584 (49%) nursing home workers; 160 (13%) nursing home residents

Pan 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

30, across all WHO regions

22 March to 18 June 2020

HCQ: 335 (< 50 years), 410 (50 to 69 years), 202 (≥ 70 years)

Standard care: 317 (< 50 years), 396 (50 to 69 years), 193 (≥ 70 years)

1853 total: 947 HCQ; 906 standard care

All hospitalized.

HCQ: 862/947 moderate or severe (of whom 517 receiving oxygen), 85 critical

Standard care: 824/906 moderate or severe (of whom 483 receiving oxygen), 82 critical

Skipper 2020

1: Treatment

1: HCQ vs placebo

RCT, double‐blind

USA and Canada

22 March to 6 May 2020

HCQ: median 41 y (IQR 33 to 49 y)

Placebo: median 39 y (IQR 31 to 50 y)

491 total: 244 HCQ; 247 placebo

All ambulatory patients, so presumed to have mild disease if symptomatic.

HCQ: 48/244 asymptomatic

Placebo: 52/247 asymptomatic

Tang 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label (no placebo)

China

11 February to 29 February 2020

HCQ: mean 48 y (SD 14.1 y)

Standard care: mean 44.1 y (SD 15 y)

150 total: 75 HCQ; 75 standard care

All hospitalized.

HCQ: 15/75 mild; 59/75 moderate; 1/75 severe disease

Standard care: 7/75 mild; 67/75 moderate; 1/75 severe disease

CQ: chloroquine; CT: computed tomography; HCQ: hydroxychloroquine; IQR: interquartile range; ITT: intention‐to‐treat; NR: not reported; RCT: randomized controlled trial; SD: standard deviation; WHO: World Health Organization; y: years.

Objective 1. For treatment of COVID‐19 disease

We included 12 RCTs (8569 participants) assessing treatment of patients diagnosed with COVID‐19.

Trial size

Trial size varied widely, from 22 participants in Huang 2020 to 4716 participants in Horby 2020. Five trials recruited fewer than 100 participants each (Chen 2020a; Chen 2020b; Chen 2020c; Davoodi 2020; Huang 2020).

Geographical location and time period

Four trials were conducted in China, early in the pandemic; all completed recruitment in February 2020 (Chen 2020a; Chen 2020b; Huang 2020; Tang 2020). The other trials recruited from March until May or June 2020: in Brazil (Cavalcanti 2020); Egypt (Abd‐Elsalam 2020); Iran (Davoodi 2020); Spain (Mitjà 2020a); Taiwan (Chen 2020c); the UK (Horby 2020); the USA and Canada (Skipper 2020; around 90% of participants were in the USA); and one trial recruited participants in 30 countries globally (Pan 2020).

Participants

None of the trials recruited children. The protocol of one trial was modified on 9 May 2020 to allow recruitment of children, but none of the participants in the study arms included in this review (i.e. HCQ and standard care) were children (Horby 2020). The average age in most trials was between 40 and 50 years old, except for Horby 2020, in which the mean age of participants was around 65 years in both arms, and Pan 2020, with a median somewhere between 50 and 69 years old.

Nine trials recruited hospitalized patients (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Horby 2020; Huang 2020; Pan 2020; Tang 2020), whilst the other three trials were focused on ambulatory care and only included outpatients (Davoodi 2020; Mitjà 2020a; Skipper 2020).

Overall, 7347/8569 (85.7%) participants had COVID‐19 confirmed by SARS‐CoV‐2 PCR on clinical samples. Six trials recruited participants only if they had a positive PCR (Chen 2020a; Chen 2020b; Chen 2020c; Huang 2020; Mitjà 2020a; Tang 2020). In three of the remaining six trials, the majority of participants had a positive PCR: 504/665 (75.8%; Cavalcanti 2020), 4234/4716 (89.8%; Horby 2020), and 1850/1853 (> 99%; Pan 2020). Skipper 2020 reported 169/491 (34.4%) to have positive PCR testing, though the test result was pending for 48/491 (9.8%), and not available or not done for 204/491 (41.5%) (Skipper 2020). Abd‐Elsalam 2020 and Davoodi 2020 did not report number of participants with positive PCR test results.

Where severity of COVID‐19 disease at enrolment was not reported using author label or defined criteria equivalent to asymptomatic, mild, moderate, severe or critical, this was inferred using classification as described by WHO guidance (WHO 2020c). Of the 1800 participants (9 trials) amenable to classification, 100 (6%) were asymptomatic, 1183 (66%) had mild disease, 506 (28%) moderate disease, and 11 (0.6%) severe disease. Participants in Horby 2020 were classified according to receipt of oxygen or other respiratory support: 1112/4716 (24%) were not receiving oxygen or ventilation at enrolment (who would be labelled as asymptomatic or mild); 2811/4716 (60%) received oxygen (who could have moderate, severe or critical disease, depending on oxygen needs); and 793/4716 (17%) received invasive ventilation (who would be classified as having critical disease). Participant disease severity was reported similarly by Pan 2020: 686/1853 (37%) were not receiving oxygen at enrolment; 1000/1853 (54%) were receiving oxygen or other respiratory support but not invasive ventilation; 167/1853 (9%) were receiving invasive ventilation.

Where reported, hypertension was usually the most common comorbidity, though its prevalence varied widely: from 6% of participants in Tang 2020 and 11% in Skipper 2020, to 27% in Chen 2020a and 39% in Cavalcanti 2020. The next most common comorbidity was usually diabetes mellitus, though its prevalence varied from < 10% (Chen 2020a; Huang 2020; Skipper 2020), to 19% in Cavalcanti 2020, 21% in Pan 2020, and 27% in Davoodi 2020 and Horby 2020. In three of the five trials reporting chronic heart and lung disease (including asthma), prevalence for each was < 15% of participants (Cavalcanti 2020; Mitjà 2020a; Skipper 2020); Horby 2020 reported 26% of participants to have heart disease and 22% chronic lung disease; Pan 2020 reported 21% of participants to have cardiac disease and 12% chronic lung disease or asthma. Other reported comorbidities were present in < 5% of participants, such as cancer and chronic renal or liver disease. Two of the three outpatient‐treatment trials reported proportions of participants with no known comorbidities: 47% for Mitjà 2020a and 31% for Skipper 2020. The third outpatient‐treatment trial reported 28% of participants to have diabetes mellitus, and 1 of 54 participants had underlying lung disease (Davoodi 2020). Two trials did not report comorbid conditions for their participants (Chen 2020b; Chen 2020c).

Special patient populations were not commonly recruited. Most trials excluded pregnant women (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Huang 2020; Mitjà 2020a; Tang 2020). Whilst not excluding pregnant women from their trials, Horby 2020 and Pan 2020 did not report how many pregnant women were included, and Skipper 2020 recruited none. Only Skipper 2020 reported recruitment of people with immunosuppression other than due to HIV (3 of 491 total participants); across all trials, 26 participants were reported to have HIV.

Two trials provided a breakdown of contact history: 238/293 (81%) had healthcare exposure history and 2% were household contacts in Mitjà 2020a; 51% of participants in Skipper 2020 were healthcare workers, whilst 29% had household exposure to someone with COVID‐19.

Time from onset of symptoms to enrolment varied widely between trials. The outpatient trials reporting this information enrolled very soon after symptom onset, with medians of between one and two days in Skipper 2020 and three days in Mitjà 2020a. Three of the hospital‐based trials recruited on average between six and nine days from onset (Cavalcanti 2020; Chen 2020a; Horby 2020). Tang 2020 enrolled at a mean of 16 to 17 days from onset, which contributed to the change in timing of their primary outcome, from negative SARS‐CoV‐2 PCR at 28 days to 10 days from enrolment. Huang 2020 recruited relatively early from onset, but this appeared to be earlier for the CQ arm (median 2.5 days) than for the lopinavir/ritonavir arm (6.5 days). Abd‐Elsalam 2020, Chen 2020b, Chen 2020c, Davoodi 2020, and Pan 2020 did not report time from symptom onset to enrolment.

Interventions and comparators

Four comparisons are reported for Objective 1 (see Effects of interventions), as follows.

1. HCQ versus standard care without HCQ, or placebo

Ten trials were included in this comparison (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020). Nine trials compared HCQ to standard of care, and one trial, Skipper 2020, compared HCQ to placebo (folic acid). Two trials were multi‐arm trials: Horby 2020 allocated to five arms in a 2:1:1:1:1 ratio (the control arm (standard care) was twice the size of each intervention arm), and Pan 2020 randomized to one of five arms in a 1:1:1:1:1 ratio, of which HCQ was one arm. Horby 2020 and Pan 2020 are ongoing adaptive trials that have each dropped the HCQ arm.

2. CQ versus lopinavir/ritonavir

One trial was included in this comparison (Huang 2020).

3. HCQ + azithromycin versus standard care

One trial was included in this comparison, in which participants were randomized 1:1:1 to receive HCQ, HCQ and azithromycin, or standard of care without HCQ or azithromycin (Cavalcanti 2020).

4. HCQ versus febuxostat

One trial was included in this comparison (Davoodi 2020). In this trial, febuxostat was the experimental drug of interest, and HCQ was the comparator.

Dosing regimens for HCQ varied widely, and are summarized in Table 4. To highlight the heterogeneity of regimens between the trials, the loading daily dose on day 1 for participants in Horby 2020 and Pan 2020 (2000 mg) was equivalent to the total cumulative dose given to participants in Chen 2020a, Chen 2020b, and Davoodi 2020.

Table 4. Dosing regimens in hydroxychloroquine treatment trials1

Study	Hydroxychloroquine (HCQ) dose regimen	Control group	Total hydroxychloroquine dose
Abd‐Elsalam 2020	800 mg on day 1, followed by 400 mg daily for further 14 days (total duration of treatment 15 days)	Standard care	6400 mg
Cavalcanti 2020²	400 mg orally twice daily for 7 days	Standard care	5600 mg
Chen 2020a³	400 mg once daily for 5 days	Standard care	2000 mg
Chen 2020b	200 mg orally twice daily for 5 days	Standard care	2000 mg
Chen 2020c	800 mg on day 1, followed by 400 mg daily for further 6 days (total duration of treatment 7 days)	Standard care	3200 mg
Davoodi 2020	200 mg orally twice daily for 5 days	Standard care	2000 mg
Horby 2020	800 mg at 0 and 6 hours, then 400 mg at 12 hours from first dose and every 12 hourly for 10 days	Standard care	10,000 mg
Mitjà 2020a	800 mg on day 1, followed by 400 mg daily for further 6 days (total duration of treatment 7 days)	Standard care	3200 mg
Pan 2020	2000 mg on day 1, followed by 800 mg daily for further 9 days (total duration of treatment 10 days)	Standard care	9200 mg
Skipper 2020	800 mg (4 tablets) once, then 600 mg (3 tablets) 6 to 8 hours later, then 600 mg (3 tablets) once daily for 4 more days (5 days in total)	Placebo: folic acid in USA and lactose in Canada	3800 mg
Tang 2020	400 mg orally 3 times a day for 3 days, then twice daily from day 4, for a total of 14 days for those with mild/moderate disease and 21 days for those with severe disease	Standard care	12,400 mg mild/moderate disease; 18,000 mg severe disease

¹See Table 5 for co‐interventions given in each trial.
²Cavalcanti 2020 ‐ hydroxychloroquine plus azithromycin group received HCQ 400 mg orally twice daily and azithromycin 500 mg orally once daily for seven days.
³Chen 2020a ‐ additionally, all participants in the HCQ arm had nebulized interferon alpha; 12/15 had umifenovir (Arbidol). Standard care arm: no HCQ; all had nebulized interferon alpha; 10/15 had umifenovir (Arbidol).

Huang 2020 administered 500 mg of CQ twice daily for 10 days to participants in the CQ arm, without a loading dose on day one, for a cumulative total dose of 10,000 mg.

Co‐interventions

The pharmacological co‐interventions reported per arm in the treatment trials for comparison 1 (HCQ versus standard care without HCQ or placebo) are shown in Table 5. Considerable variability in reporting was observed. The following are of particular note regarding co‐interventions.

See: Summary of findings 1 Hydroxychloroquine (HCQ) compared to standard care or placebo for the treatment of people with COVID‐19; Summary of findings 2 Hydroxychloroquine (HCQ) compared to placebo for the prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2

Table 5. Pharmacological co‐interventions given in treatment trials for comparison 1 (HCQ versus standard care or placebo)

Study	Co‐interventions in HCQ arm	Co‐interventions in comparator arm
Abd‐Elsalam 2020	Authors report: "The Egyptian Ministry of Health (MOH) adopted a standard of care treatment protocol for COVID‐19 patients. It included paracetamol, oxygen, fluids (according to assessment), empiric antibiotic (cephalosporins), oseltamivir if needed (75 mg/12 hours for 5 days), and invasive mechanical ventilation with hydrocortisone for severe cases if PaO2 < 60 mmHg, O2 saturation < 90% despite oxygen or noninvasive ventilation, progressive hypercapnia, respiratory acidosis (pH < 7.3), and progressive or refractory septic shock".
Cavalcanti 2020¹	Corticosteroids 5/221 Oseltamivir 38/221 Aciclovir 1/221 Lopinavir/ritonavir 0/221 Ceftriaxone 86/221 Ceftaroline 11/221 Piperacillin/tazobactam 8/221 Oxacillin 0/221 Vancomycin 1/221 Carbapenem 6/221 Quinolone 22/221 No other antiviral, antibiotic, or corticosteroids 21/221	Corticosteroids 8/227 Oseltamivir 51/227 Aciclovir 0/227 Lopinavir/ritonavir 0/227 Ceftriaxone 99/227 Ceftaroline 17/227 Piperacillin/tazobactam 15/227 Oxacillin 1/227 Vancomycin 4/227 Carbapenem 3/227 Quinolone 28/227 No other antiviral, antibiotic, or corticosteroids 18/227
Chen 2020a²	Nebulized interferon alpha 15/15 Umifenovir 12/15	Nebulized interferon alpha 15/15 Umifenovir 10/15
Chen 2020b	Authors report “all received the standard treatment (oxygen therapy, antiviral agents, antibacterial agents, and immunoglobulin, with or without corticosteroids)”.
Chen 2020c³	Azithromycin 1/21	Azithromycin 2/12
Horby 2020⁴	Dexamethasone 8% Azithromycin 17%	Dexamethasone 9% Azithromycin 19%
Mitjà 2020a⁵	Cobicistat‐boosted darunavir 49/136	Cobicistat‐boosted darunavir 0/157
Pan 2020	The authors report that co‐medications will appear in supplementary tables, but these are not provided with the currently available preprint publication.
Skipper 2020⁶	Zinc 63/212 Vitamin C 101/212	Zinc 53/211 Vitamin C 101/211
Tang 2020	Umifenovir 37/75 Ribavirin 13/75 Lopinavir/ritonavir 13/75 Oseltamivir 8/75 Entecavir 1/75 Antibiotics 32/75 Corticosteroids 6/75	Umifenovir 33/75 Ribavirin 15/75 Lopinavir/ritonavir 12/75 Oseltamivir 9/75 Entecavir 1/75 Antibiotics 27/75 Corticosteroids 4/75

HCQ, hydroxychloroquine; PaO2, partial pressure of oxygen

¹Cavalcanti 2020 ‐ this was a three‐arm trial, of which the third arm received HCQ + azithromycin.
²Chen 2020a ‐ authors report that two participants received lopinavir/ritonavir, but it is unclear which study arms these participants were in. Whether or not any participants received corticosteroids or antibiotics is not reported.
³Chen 2020c ‐ in addition to the above, authors report: "Both study group and comparison group received standard of care comprising supportive treatment for subjects with mild clinical COVID‐19 symptoms and antimicrobial therapy for subjects presenting with moderate clinical COVID‐19 symptoms. The treatment consisted of: (1) ceftriaxone 2 g daily for 7 days +/‐ azithromycin 500 mg on day 1 and 250 mg on days 2–5; or (2) levofloxacin 750 mg daily for 5 d; or (3) levofloxacin 500 mg daily; or (4) moxifloxacin 400 mg daily for 7–14 days for subjects allergic to ceftriaxone or azithromycin or according to physician discretion. Oseltamivir 75 mg b.i.d. will be administered for 5 days to subjects presenting with concomitant influenza A or B infection".
⁴Horby 2020 ‐ authors presented the percentage of participants in each arm receiving dexamethasone or azithromycin. Data on antibiotics and other antivirals not reported. This trial was a platform trial with other arms testing tocilizumab, azithromycin, and dexamethasone, as well as convalescent plasma.
⁵Mitjà 2020a ‐ the trial was originally designed to test HCQ with cobicistat‐boosted darunavir, but this was modified during the trial as further information became available that cobicistat‐boosted darunavir had no in vitro activity against SARS‐CoV‐2.
⁶Skipper 2020 ‐ whether or not participants received antimicrobials or corticosteroids is not reported.

Cavalcanti 2020 reported that fewer than 10% of participants did not receive concurrent treatment with an antiviral, antibiotic, or corticosteroid. However, corticosteroids were rarely given (13 of 448 participants).
All participants in Chen 2020a received nebulized interferon‐alpha, and the majority (22/30) received umifenovir (Arbidol). Both are postulated anti‐SARS‐CoV‐2 drugs.
Horby 2020 reported that a minority of participants received concurrent corticosteroids (dexamethasone) (< 10%) and azithromycin (< 20%).
Participants initially enrolled into the HCQ arm of Mitjà 2020a received cobicistat‐boosted darunavir with HCQ as a planned combination, which was stopped when its activity against SARS‐CoV‐2 was called into question.
Skipper 2020 reported subgroup analyses for self‐reported use of zinc and vitamin C; this was common, with ~25% and ~50% of participants reporting their use, respectively.

There did not appear to be a difference in receipt of pharmacological co‐interventions between trial arms, where this information was reported. No trials reported concurrent use of remdesivir.

Follow‐up

One trial measured all outcomes up to six days (Chen 2020b); six trials followed participants up until 14 to 15 days (Cavalcanti 2020; Chen 2020a; Chen 2020c; Davoodi 2020; Huang 2020; Skipper 2020); and four trials completed data collection at 28 days from enrolment (Abd‐Elsalam 2020; Horby 2020; Mitjà 2020a; Tang 2020). Pan 2020 followed participants up to discharge from hospital. Two trials used telephone follow‐up in place of or in addition to in‐person outcome assessment (Cavalcanti 2020; Mitjà 2020a); one trial employed online surveys for enrolment and all follow‐up (Skipper 2020).

Outcome measures

Our predefined primary outcomes were death and time to negative PCR for SARS‐CoV‐2 on respiratory samples. Ten trials reported death (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020c; Davoodi 2020; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020).

PCR‐based outcomes varied amongst the included trials. Three trials reported time to negative PCR (Abd‐Elsalam 2020; Chen 2020a; Huang 2020); four trials reported negative PCR at specified time points: 7 days (Chen 2020a; Tang 2020); 10 days (Huang 2020; Tang 2020); and 14 days from enrolment (Chen 2020c; Huang 2020; Tang 2020); and the primary outcome in one trial was reduction in 'viral load' (amount of virus per swab sample) at day 3 and day 7 after enrolment (Mitjà 2020a).

Regarding our secondary outcomes, the following information was reported.

Number of participants admitted to hospital (if receiving ambulatory treatment): this was reported by the three outpatient‐based trials (Davoodi 2020; Mitjà 2020a; Skipper 2020).
Number of participants requiring mechanical ventilation: three trials reported this outcome (Cavalcanti 2020; Horby 2020; Tang 2020).
Length of hospital admission: this was reported as a mean by Abd‐Elsalam 2020 and Cavalcanti 2020; the authors of Tang 2020 provided this upon request. Horby 2020 reported a median, but without interquartile range, and no mean. Huang 2020 provided a Kaplan‐Meier chart, but no mean; however, proportion discharged by day 14 from enrolment was reported.
Time to clinical improvement was reported as survival data only by Tang 2020. For the remaining trials, either a mean (Abd‐Elsalam 2020) or median (Chen 2020a; Mitjà 2020a) was reported, and/or the definitions of time to clinical improvement were not comparable (Chen 2020a; Chen 2020b).
Duration of mechanical ventilation postenrolment in survivors of COVID‐19 was not reported by any trials.

Five of the 12 included trials did not report the number of participants experiencing any adverse events (Abd‐Elsalam 2020; Chen 2020c; Davoodi 2020; Horby 2020; Pan 2020). Five, with some overlap (Abd‐Elsalam 2020; Davoodi 2020; Horby 2020; Pan 2020; Skipper 2020), did not report the number of participants experiencing serious adverse events, with Skipper 2020 stating: “No serious adverse events attributable to the study drug occurred”. The remaining trials reported events without attribution to a particular drug.

Additionally, Skipper 2020 used the change in symptoms over 14 days from enrolment as their primary outcome. This differed significantly from our predefined outcomes, and was not comparable with the outcomes of other trials.

Objective 2. For prevention of COVID‐19 disease in people at risk of exposure to SARS‐CoV‐2

No eligible trials were identified for this objective.

Objective 3. For prevention of COVID‐19 disease in people who have been exposed to SARS‐CoV‐2

We included two trials for this objective: one with double‐blind individual randomization to HCQ or placebo that enrolled 821 participants (Boulware 2020), and one open‐label cluster‐RCT comparing HCQ with standard care that enrolled 2525 participants (Mitjà 2020b).

Geographical location and time period

Boulware 2020 was based in the USA and Canada, and recruited from 17 March to 6 May 2020. Mitjà 2020b recruited in Spain between 17 March and 28 April 2020.

Participants

Both trials only recruited asymptomatic people with a history of exposure to people with laboratory‐confirmed COVID‐19 (Boulware 2020; Mitjà 2020b).

In Boulware 2020, exposure history was most commonly in a healthcare setting (545/821, 66%), followed by household contact (245/821, 30%). The corresponding figures for Mitjà 2020b were 12% for healthcare workers and 28% household exposure; additionally, 49% worked and 13% lived in a nursing home. Exposure was deemed to be high risk (neither eye protection nor a surgical mask/respirator was worn) in 88% of participants, with 60% in Boulware 2020 wearing no personal protective equipment. Participants were enrolled at a median of three days after exposure in Boulware 2020 and four days after exposure in Mitjà 2020b.

Children were excluded. Median age was 41 years in the HCQ arm and 40 years in the placebo arm in Boulware 2020; mean age was 49 years in both the HCQ and standard care arms in Mitjà 2020b.

Most participants did not have comorbidities associated with increased risk of severe acute COVID‐19. In Boulware 2020, 12% had hypertension, 8% chronic respiratory disease (mostly asthma), 3% diabetes, and < 1% reported each of heart disease, kidney disease, and cancer; 73% reported no pre‐existing conditions. Mitjà 2020b reported underlying cardiovascular disease in 13% of participants, respiratory disease in 4%, metabolic disease in 8%, and some nervous system disease in 15%. HIV and non‐HIV immunosuppression were reported in 1/821 and 4/821 participants, respectively (Boulware 2020). Whilst pregnant women were not excluded, their representation in the participants was not reported (Boulware 2020). Mitjà 2020b did not report on participants with HIV or other immunosuppression, nor whether pregnant women were included.

Interventions and comparators

The HCQ dosing regimen in Boulware 2020 was the same as in Skipper 2020: 1400 mg (800 mg, then 600 mg 6 to 8 hours later) on day 1, followed by 600 mg once daily for a further four days, translating to a cumulative total of 3800 mg over five days. Mitjà 2020b used the same HCQ dosing as in the paired treatment trial Mitjà 2020a: 800 mg on day 1, followed by 400 mg once daily for a further six days, for a total of 3200 mg over seven days.

The comparator in Boulware 2020 was placebo in the form of unmarked folic acid tablets, which closely resembled HCQ tablets, to be taken on the same schedule as HCQ. Mitjà 2020b used neither placebo nor an active comparator.

Follow‐up

In Boulware 2020, follow‐up was conducted using online surveys exclusively, with the final survey to be completed four to six weeks after enrolment. Mitjà 2020b used a combined approach of in‐person visits to the participant's home on days 1 and 14, and telephone interviews on days 3, 7, and 28.

Outcome measures

Our primary outcome of development of COVID‐19 was assessed at 14 days in both trials. In Boulware 2020, the definition of COVID‐19 was expanded beyond confirmed (i.e. by PCR for SARS‐CoV‐2) to include probable COVID‐19 due to difficulty accessing PCR testing, whereas in Mitjà 2020b development of COVID‐19 required both symptoms and a positive PCR test. Our second primary outcome, production of antibodies to SARS‐CoV‐2, was assessed by Mitjà 2020b at 14 days.

A variety of secondary outcomes were measured, including hospitalization due to COVID‐19, which partly addressed our outcome of disease severity in participants developing COVID‐19 (Boulware 2020; Mitjà 2020b). Onward transmission to household contacts from index participants was not assessed.

Adverse events were assessed through self‐reporting by participants using an online survey in Boulware 2020, and through telephone and in‐person visits in Mitjà 2020b. QT prolongation was not assessed due to lack of in‐person assessment (which would be necessary for electrocardiography to be performed) in Boulware 2020; there was one in‐person assessment in Mitjà 2020b, but at the participant's home, where electrocardiography may not have been practical.

Excluded studies

We excluded 791 articles (see Figure 1), 88 of which were at the full‐text stage (see the Characteristics of excluded studies section), for the following reasons: 35 were not RCTs; 16 lacked a control group without CQ or HCQ; one did not include mention of CQ or HCQ; 32 duplicates were found, which had not been apparent at first screening; and four were excluded for other reasons.

Risk of bias in included studies

See Characteristics of included studies, which includes a 'Risk of bias' table for each included trial. The results of the 'Risk of bias' assessments across all included trials are summarized in Figure 2.

Figure 2

Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included trial.

Allocation

We judged that 10 out of the 14 included trials were at low risk of bias (Abd‐Elsalam 2020; Boulware 2020; Cavalcanti 2020; Chen 2020b; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020), two were at unclear risk of bias (Davoodi 2020; Mitjà 2020b), and two were at high risk of bias for random sequence generation (Chen 2020a; Huang 2020). The description of the method of randomisation was inadequate in Davoodi 2020 and Mitjà 2020b. Chen 2020a had 15 participants in each arm, and Huang 2020 had a notable imbalance between treatment arms raising concerns about the integrity of the randomisation process; neither trial explicitly described the method of randomisation.

We assessed six trials as at low risk of bias for allocation concealment (Boulware 2020; Cavalcanti 2020; Davoodi 2020; Horby 2020; Skipper 2020; Tang 2020), and seven trials as at unclear risk of bias due to lack of clear reporting of the method of allocation concealment (Chen 2020a; Chen 2020b; Chen 2020c; Huang 2020; Mitjà 2020a; Mitjà 2020b; Pan 2020). We judged Abd‐Elsalam 2020 to be at high risk of bias for allocation concealment, as the method used was not reported, and there were more participants with comorbidity (obesity and smoking history) in the intervention arm, although there was not a statistically significant difference in these characteristics between the treatment arms.

Blinding

We assessed the risk of bias associated with blinding for each outcome separately (details are provided in the 'Risk of bias' table for each trial), but made our overall judgement for each trial based on the primary outcome as stated by the trial authors.

We assessed 10 trials as at low risk of performance bias (blinding of participants and personnel) (Abd‐Elsalam 2020; Boulware 2020; Chen 2020a; Chen 2020c; Horby 2020; Huang 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020). We judged Cavalcanti 2020 to be at unclear risk of bias, as it was not blinded, and the primary outcome consisted of an ordinal scale ranking clinical improvement or deterioration. We judged Chen 2020b to be at high risk of bias because although the authors stated that the researchers and patients were unaware of treatment assignments, no placebo was used and the methods of blinding were not described, and the primary outcome was based on patient‐reported clinical recovery. We judged Davoodi 2020 to be at high risk of bias as it was an open‐label trial, and the primary outcome of hospitalization could have been influenced by clinicians knowing the treatment allocation. Similarly, we judged Mitjà 2020b to be at high risk of bias as it was an open‐label trial, and the primary outcome involved a subjective assessment of symptoms.

We assessed 11 trials as at low risk of detection bias (blinding of outcome assessment) (Abd‐Elsalam 2020; Boulware 2020; Cavalcanti 2020; Chen 2020a; Chen 2020c; Davoodi 2020; Horby 2020; Huang 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020). We judged Chen 2020b and Mitjà 2020b to be at high risk of detection bias, as the outcome assessors were not blinded to treatment allocation, and the primary outcomes of time to clinical improvement and development of symptoms are likely to have been subjectively assessed.

Incomplete outcome data

We assessed eight trials as at low risk of bias for incomplete outcome data (Boulware 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Horby 2020; Huang 2020; Mitjà 2020b; Pan 2020). Three trials were at unclear risk of bias for this domain: Abd‐Elsalam 2020 did not report on losses to follow‐up or missing data; Skipper 2020 had significant losses to follow‐up that were balanced between each group, but no explanations for losses were provided; and Tang 2020 had significant loss to follow‐up beyond 21 days of follow‐up. We assessed Mitjà 2020a as at high risk of bias for incomplete outcome data: 60 participants were excluded from the intention‐to‐treat (ITT) analysis due to negative baseline SARS‐CoV‐2 swab, missing reverse transcription polymerase chain reaction (RT‐PCR) at all follow‐up visits, or consent withdrawal, and a further 23 participants had protocol deviations including eight participants lost to follow‐up. We judged Chen 2020c as at high risk of attrition bias, as the authors reported loss to follow‐up of ~10% (3/33), and missing participants were imputed as having negative results, which could have impacted on the results as the sample size was small. We judged Davoodi 2020 as at high risk of bias as there were no outcome data for 10% (6/60), which could have impacted the results due to the small sample size.

Selective reporting

We assessed Horby 2020 as at low risk of bias for selective reporting. Three trials were at unclear risk of bias for this domain: Boulware 2020 changed the primary outcome from confirmed COVID‐19 cases to probable due to a problem with access to testing; Pan 2020 was accessed as a preprint at the time of writing of this review, which did not include all outcome data, and referencing one change between protocol and trial report that was not explained; Tang 2020 changed their primary outcome and gave justification for this, but did not report the secondary outcomes. Ten trials were at high risk of bias for selective reporting (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Davoodi 2020; Huang 2020; Mitjà 2020a; Mitjà 2020b; Skipper 2020), all of which reported outcomes that deviated from those stated in the protocol (described in Characteristics of included studies).

Other potential sources of bias

We identified other potential sources of bias in four trials. Skipper 2020 and Tang 2020 were at high risk of bias as they were terminated early, which could have introduced bias as both trials had a time‐updating variable as the primary outcome. Mitjà 2020a was also at high risk of other bias: a small number of participants were randomized who were in fact not eligible for the trial, but these participants were kept as part of the ITT population.

Mitjà 2020b was a cluster‐randomized trial, and so was assessed for risk of bias against five further domains specific to cluster‐randomized trials. We judged the trial to be at low risk for four out of five domains (see Characteristics of included studies), and at high risk of bias for comparability with individually randomized trials: contamination was possible due to the open‐label design, and the intervention would be expected to work best when given to all contacts of a case rather than some being randomized to the intervention and some randomized to no intervention, which would preclude comparability with an individually randomized trial.

Effects of interventions

See summary of findings Table 1 for Objective 1, Comparison 1, and summary of findings Table 2 for Objective 3.

Due to inability to extract data disaggregated for subgroups on outcomes predefined in the review protocol, we did not perform subgroup analyses. Furthermore, heterogeneity in most analyses was minimal.

Objective 1. For treatment of COVID‐19 disease

Comparison 1: HCQ versus standard care without HCQ or placebo

Ten trials of treatment of people with COVID‐19 compared HCQ to standard care or placebo (8270 participants; Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020). The arm randomizing participants to a combination of HCQ with azithromycin in Cavalcanti 2020 was not included in this comparison, but is included in Comparison 3 below.

Nine of the 10 trials reported death due to any cause (Abd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020). Meta‐analysis using ITT results for each trial found little or no difference between HCQ and standard care without HCQ or placebo in all‐cause death (risk ratio (RR) 1.09, 95% confidence interval (CI) 0.99 to 1.19; 8208 participants; 9 RCTs; Analysis 1.1). Sensitivity analysis performed using modified ITT results as reported by three trials revealed no difference in the pooled effect estimate (RR 1.09, 95% CI 0.99 to 1.19; 8043 participants; 9 RCTs; Analysis 1.2) (Cavalcanti 2020; Mitjà 2020a; Skipper 2020).

Our predefined outcome involving tests for SARS‐CoV‐2, time to negative PCR for SARS‐CoV‐2 on respiratory samples, was reported as a median by Chen 2020a and Chen 2020c, and as a mean by Abd‐Elsalam 2020 and Tang 2020; all trials reported no significant difference between the arm that received HCQ and the arm that did not. Two of the trials reported the related outcome of negative PCR for SARS‐CoV‐2 at day 7 after enrolment as dichotomous outcomes (Chen 2020a; Tang 2020), and three trials reported negative PCR at day 14 (Chen 2020a; Chen 2020c; Tang 2020). We deemed the latter (i.e. negative PCR at day 14) to be more important based on the current understanding of COVID‐19, so this is displayed in summary of findings Table 1 . No significant difference between HCQ and standard care without HCQ was revealed in meta‐analysis of either negative PCR at day 14 (RR 1.00, 95% CI 0.91 to 1.10; 213 participants; 3 RCTs; Analysis 1.3) or negative PCR at day 7 (RR 0.86, 95% CI 0.68 to 1.09; 180 participants; 2 RCTs; Analysis 1.4) after enrolment.

Of the two trials assessing ambulatory treatment of people with COVID‐19, only Skipper 2020 was included in the analysis of risk of admission to hospital; Mitjà 2020a did not report denominators disaggregated for allocation to HCQ versus standard care without HCQ. In Skipper 2020, though the risk ratio may suggest an important benefit from HCQ, the confidence intervals were wide, and included potential important harm (RR 0.41, 95% CI 0.13 to 1.27; 465 participants; 1 RCT; Analysis 1.5).

Three trials reported results for people hospitalized with COVID‐19 going on to require mechanical ventilation (Cavalcanti 2020; Horby 2020; Tang 2020). No significant difference was found in participants progressing to mechanical ventilation when comparing HCQ to no HCQ (RR 1.11, 95% CI 0.91 to 1.37; 4521 participants; 3 RCTs; Analysis 1.6).

Three trials reported mean length of hospital admission (Abd‐Elsalam 2020; Cavalcanti 2020; Tang 2020). We noted that early in the pandemic suitability for discharge was often driven by infection prevention and control considerations, and therefore might not have been a good reflection of the efficacy of HCQ. Accordingly, we decided not to include results from Tang 2020 in the analysis for this outcome, as participants remained in hospital until they were deemed no longer infectious. Pooled length of admission in Abd‐Elsalam 2020 and Cavalcanti 2020 did not differ between participants who received HCQ and those who did not (mean difference (MD) 0.15 days shorter with HCQ, 95% CI 0.75 shorter to 0.45 longer; 642 participants; 2 RCTs; Analysis 1.7).

Time to clinical improvement (for symptomatic patients) and time to negative PCR for SARS‐CoV‐2 on respiratory samples were reported as hazard ratios (HRs) and corresponding 95% CIs by Tang 2020. No significant difference was found for time to clinical improvement (HR 1.01, 95% CI 0.59 to 1.74; 119 participants; 1 RCT; Analysis 1.8) or time to negative PCR for SARS‐CoV‐2 on respiratory samples (HR 0.85, 95% CI 0.58 to 1.23; 150 participants; 1 RCT; Analysis 1.9).

Duration of mechanical ventilation postenrolment in survivors of COVID‐19 was not reported by any trials.

Six trials reported number of participants with any adverse events (Cavalcanti 2020; Chen 2020a; Chen 2020b; Mitjà 2020a; Skipper 2020; Tang 2020). Meta‐analysis revealed a higher risk of adverse events in participants receiving HCQ versus those receiving standard care or placebo (RR 2.90, 95% CI 1.49 to 5.64; 1394 participants; 6 RCTs; Analysis 1.10). Adverse events reported in the six trials are summarized in Table 6.

Table 6. Adverse events for HCQ versus standard care without HCQ, or placebo, for treatment

Study

HCQ

No HCQ

Cavalcanti 2020

QTc prolongation (13/89)

Arrhythmia (3/199)

Bradycardia (1/199)

Supraventricular tachycardia (2/199)

Pneumothorax (1/199)

Bloodstream infection (1/199)

Itching (1/199)

Nausea (9/199)

Anaemia (14/199)

Elevated ALT or AST (17/199)

Elevated bilirubin (5/199)

Hypoglycaemia (1/199)

Leucopenia (3/199)

QTc prolongation (1/58)

Arrhythmia (1/177)

Bradycardia (1/177)

Bronchospasm (1/177)

Nausea (2/177)

Vomiting (1/177)

Anaemia (11/177)

Elevated ALT or AST (6/177)

Elevated bilirubin (2/177)

Leucopenia (3/177)

Chen 2020a¹

Transient elevated AST with anaemia (1/15)

Diarrhoea (2/15)

Fatigue (1/15)

Elevated AST (1/15)

Elevated creatinine (1/15)

Chen 2020b

Headache (1/31)

Rash (1/31)

Mitjà 2020a

Gastrointestinal disorders (148/169)

General disorders (30/169)

Infections and infestations (9/169)

Injury, poisoning, and procedural complications (1/169)

Metabolic and nutrition disorders (2/169)

Musculoskeletal and connective tissue disorders (1/169)

Nervous system disorders (63/169)

Psychiatric disorders (2/169)

Renal and urinary disorders (1/169)

Reproductive system and breast disorders (1/169)

Respiratory, thoracic, and mediastinal disorders (2/169)

Skin and subcutaneous tissue disorders (11/169)

Vascular disorders (1/169)

Gastrointestinal disorders (7/184)

General disorders (1/184)

Infections and infestations (12/184)

Metabolic and nutrition disorders (1/184)

Nervous system disorders (3/184)

Skipper 2020²

Upset stomach/nausea (66/212)

Diarrhoea, other GI symptoms, vomiting (50/212)

Neurologic (nervousness, irritability, dizziness, vertigo) (20/212)

Skin reaction, rash (6/212)

Ringing in ears (8/212)

Allergic reaction, self‐reported (5/212)

Changes in vision (4/212)

Warmth, hot flashes, night sweats (2/212)

Headache (2/212)

Upset stomach/nausea (26/211)

Diarrhoea, other GI symptoms, vomiting (20/211)

Neurologic (nervousness, irritability, dizziness, vertigo) (13/211)

Skin reaction, rash (2/211)

Ringing in ears (5/211)

Changes in vision (5/211)

Taste, dry mouth (1/211)

Heart racing, anxiety, panic attack (1/211)

Tang 2020

Disease progression (1/70)

Upper respiratory tract infection (1/70)

Diarrhoea (7/70)

Vomiting (2/70)

Nausea (1/70)

Abdominal discomfort (1/70)

Blurred vision (1/70)

Thirst (1/70)

Sinus bradycardia (1/70)

Hypertension (1/70)

Orthostatic hypotension (1/70)

Hypertriglyceridaemia (1/70)

Decreased appetite (1/70)

Fatigue (1/70)

Dyspnoea (1/70)

Flush (1/70)

Kidney injury (1/70)

Coagulation dysfunction (1/70)

Decreased white blood cell (1/70)

Increased ALT (1/70)

Increased serum amylase (1/70)

Decreased neutrophil count (1/70)

Abdominal bloating (1/80)

Fever (1/80)

Liver abnormality (1/80)

Hepatic steatosis (1/80)

Otitis externa (1/80)

Increased serum amyloid A (1/80)

ALT: alanine aminotransferase; AST: aspartate transaminase; GI: gastrointestinal; HCQ: hydroxychloroquine

¹Authors of Chen 2020a comment that “among the test group the occurrence of adverse events in subjects with moderate to severe disease is not related to medication. All adverse reactions after drug withdrawal or symptomatic treatment disappeared”.
²Skipper 2020 ‐ authors describe these adverse events as side effects reported at day 5.

Meta‐analysis of six trials that reported the number of participants experiencing serious adverse events showed no significant difference between participants receiving HCQ and those receiving standard care (RR 0.82, 95% CI 0.37 to 1.79; 1004 participants; 6 RCTs; Analysis 1.11) (Cavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Mitjà 2020a; Tang 2020). It was not possible to disaggregate data for specific serious adverse events for each trial, nor was it possible to disaggregate data for serious adverse events attributed to the intervention drug for each trial.

Our predefined specific adverse event, prolongation of the QT‐interval on electrocardiogram (ECG), was reported by one trial (Cavalcanti 2020). Risk of QT‐interval prolongation increased in participants receiving HCQ (without azithromycin) versus those receiving standard care or azithromycin (RR 8.47, 95% CI 1.14 to 63.03; 147 participants; 1 RCT; Analysis 1.12). Fewer than half of participants had an ECG performed within seven days of enrolment; this appeared to be higher in those receiving HCQ (89/199, 44.7%) than in those receiving standard care (58/177, 32.8%).

Comparison 2: CQ versus lopinavir/ritonavir (LPV/r)

One trial (22 participants) reported this comparison (Huang 2020). Due to the comparison not having been predefined, and this being a single small trial with high risk of selection and reporting bias, reporting few of our predefined outcomes, a separate 'Summary of findings' table is not provided.

Death was not reported as an outcome (Huang 2020).

Time to negative PCR for SARS‐CoV‐2 on respiratory samples was not reported, but the proportion with negative PCR ranged from appreciable benefit to appreciable harm between arms at day 7 from enrolment (RR 1.20, 95% CI 0.64 to 2.25; 22 participants; 1 RCT; Analysis 2.1) and day 14 from enrolment (RR 1.08, 95% CI 0.85 to 1.36; 22 participants; 1 RCT; Analysis 2.2).

Number of participants admitted to hospital (if receiving ambulatory treatment) was not relevant for this hospital inpatient‐based trial.

Number of participants requiring mechanical ventilation after enrolment was not reported (Huang 2020).

We were unable to extract length of hospital admission as a mean, but visual inspection of the Kaplan‐Meier chart appeared to show a median time to discharge of around 11 days for the CQ arm, and around 14 days for the LPV/r arm (Huang 2020). The number of participants discharged by day 14 from enrolment was reported to be 10/10 in the CQ arm versus 6/12 in the LPV/r arm (RR 1.91, 95% CI 1.09 to 3.34; 22 participants; 1 RCT; Analysis 2.3).

Time to clinical improvement was not reported as a mean or median (Huang 2020). However, clinical recovery at day 10 was reported as showing no significant difference between study arms (RR 1.37, 95% CI 0.78 to 2.42; 22 participants; 1 RCT; Analysis 2.4).

There was no difference in the number of participants experiencing adverse events between study arms (RR 1.08, 95% CI 0.78 to 1.50; 22 participants; 1 RCT; Analysis 2.5); QT‐interval prolongation was not specifically reported. No serious adverse events were reported in either arm (Huang 2020).

Comparison 3: HCQ + azithromycin versus standard care

One trial (444 participants) reported this comparison (Cavalcanti 2020). Due to the comparison not having been predefined, and this trial having a high risk of reporting bias and unclear risk of performance and detection bias, a separate 'Summary of findings' table is not provided.

Death was reported as showing no difference between study arms (RR 0.52, 95% CI 0.13 to 2.07; 444 participants; 1 RCT; Analysis 3.1).

Time to negative PCR for SARS‐CoV‐2 was not reported, and as this was a trial of hospitalized patients, neither was number of participants admitted to hospital (Cavalcanti 2020).

The number of participants requiring mechanical ventilation did not differ between study arms (RR 1.61, 95% CI 0.82 to 3.15; 444 participants; 1 RCT; Analysis 3.2). Duration of mechanical ventilation was not reported (Cavalcanti 2020).

Length of hospital admission was similar between groups (MD 0.50 days longer with HCQ + azithromycin, 95% CI 0.81 days shorter to 1.81 days longer; 444 participants; 1 RCT; Analysis 3.3).

Time to clinical improvement was not reported.

Adverse events were experienced by a higher proportion of participants who received at least one dose of HCQ + azithromycin versus participants receiving neither HCQ nor azithromycin (RR 1.74, 95% CI 1.27 to 2.38; 416 participants; 1 RCT; Analysis 3.4). Serious adverse events did not differ significantly between study arms (RR 1.85, 95% CI 0.36 to 9.43; 416 participants; 1 RCT; Analysis 3.5).

When assessed, QT‐interval prolongation on ECG was more common amongst participants receiving HCQ + azithromycin (17/116) versus those receiving neither drug (1/58) (RR 8.50, 95% CI 1.16 to 62.31; 174 participants; 1 RCT; Analysis 3.6). Performance of ECG within seven days of enrolment appeared to be more frequent in the HCQ + azithromycin arm (116/239, 48.5%) than in the standard care arm (58/177, 32.8%).

Comparison 4: HCQ versus febuxostat

One trial (60 participants) reported this comparison (Davoodi 2020). A separate 'Summary of findings' table is not provided.

No deaths were reported in either study arm (Davoodi 2020).

Three participants in each arm (of 25 in the HCQ arm and 29 in the febuxostat arm) required hospitalization during the 14 days of follow‐up (RR 1.16, 95% CI 0.26 to 5.24; 54 participants; 1 RCT; Analysis 4.2).

Number of participants requiring mechanical ventilation was not reported explicitly, but the authors reported: “All
hospitalised patients … did not require ICU care" (Davoodi 2020).

Length of hospital admission was not reported precisely, but authors reported: “All
hospitalised patients were released from hospitals between 1 and 7 days of hospitalization” (Davoodi 2020).

Time to clinical improvement was not reported in a way that fit with our planned data extraction or analysis.

Duration of mechanical ventilation was not reported.

Reduction in involvement of the lungs on CT scan between days 1 and 14 was reported to be no different between the HCQ and febuxostat arms.

Adverse events were not reported.

Objective 2. Preventing COVID‐19 disease in people at risk of exposure to SARS‐CoV‐2

No eligible trials provided outcome results for this objective.

Objective 3. Preventing COVID‐19 disease in people who have been exposed to SARS‐CoV‐2

We deemed the effect of HCQ on the prevention of COVID‐19 to be susceptible to differences in administration to an individual, versus a cluster of individuals all in contact with one index person. We therefore did not pool results from the individually‐randomized RCT, Boulware 2020, with those from the cluster‐RCT (Mitjà 2020b).

Comparison 5: HCQ versus placebo by individual randomization

One trial (821 participants) reported this comparison (Boulware 2020). See summary of findings Table 2.

Development of confirmed COVID‐19 at 14 days from enrolment was not found to differ significantly between the two arms (RR 1.20, 95% CI 0.50 to 2.87; 821 participants; 1 RCT; Analysis 5.1).

Production of antibodies to SARS‐CoV‐2 and development of COVID‐19 in household contacts of the recipient of the prophylaxis were not reported (Boulware 2020).

For our predefined outcome of disease severity of participants who develop COVID‐19, we extracted data for participants hospitalized due to COVID‐19; this did not differ significantly between those receiving HCQ and those receiving placebo (RR 0.98, 95% CI 0.06 to 15.66; 821 participants; 1 RCT; Analysis 5.2).

Participants receiving at least one dose of HCQ had an increased risk of adverse events compared to those not receiving HCQ (RR 2.39, 95% CI 1.83 to 3.11; 700 participants; 1 RCT; Analysis 5.3). No serious adverse events were reported in either arm. QT‐interval prolongation on ECG was not reported, but follow‐up was performed remotely using an online survey, so ECG was not performed as part of the trial (Boulware 2020).

Comparison 6: HCQ versus standard care by cluster randomization

One trial (2525 participants) reported this comparison (Mitjà 2020b). Due to the cluster‐RCT design and appropriate analysis by the trial authors, adjusted risk ratios have been taken from the report.

Development of symptomatic confirmed COVID‐19 at 14 days from enrolment was not found to differ significantly between participants randomized to HCQ (64/1116; 5.7%) and those allocated to standard care (74/1198; 6.2%) (adjusted RR 0.89, 95% CI 0.54 to 1.46; 2314 participants; 1 RCT; Mitjà 2020b).

Production of antibodies to SARS‐CoV‐2 at 14 days was reported in 137/958 (14.3%) of the participants in HCQ clusters and 91/1042 (8.7%) in clusters not receiving HCQ (adjusted RR 1.6, 95% CI 0.96 to 2.69; 2000 participants; 1 RCT; Mitjà 2020b).

Development of COVID‐19 in household contacts of the recipient of the prophylaxis was not reported by either trial.

Disease severity of participants who developed COVID‐19 was not reported. Five participants in the HCQ clusters (with a denominator of 1197, which is unexplained in its deviation from the randomized total of 1225) and 8/1300 in the standard care clusters died (Mitjà 2020b). Causes of death were not reported.

Adverse events were reported in 671/1197 (56%) participants in the HCQ clusters versus 77/1300 (6%) participants in the clusters not receiving HCQ; a relative effect estimate was not reported (Mitjà 2020b). Serious adverse events were reported, but it was not clear whether they were reported as number of events or number of participants, and did not match the intensity grading reported by the pharmacovigilance consultants employed by the trial (Mitjà 2020b). QT‐interval prolongation was not measured in this trial.

Discussion

Summary of main results

Treating COVID‐19 disease

Ten trials compared HCQ versus standard care without HCQ, or placebo (see summary of findings Table 1). HCQ makes little or no difference to death due to any cause, compared with no HCQ (high‐certainty evidence). HCQ may make little or no difference to the likelihood of a negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment (low‐certainty evidence). HCQ probably results in little to no difference in progression to mechanical ventilation (moderate‐certainty evidence). We are very uncertain about the effect of HCQ on time to clinical improvement when compared to standard care without HCQ or placebo (very low‐certainty evidence). HCQ probably results in an increased risk of developing adverse events (moderate‐certainty evidence), but may make little or no difference to the risk of serious adverse events (low‐certainty evidence). We are very uncertain about the effect of HCQ on prolongation of QT‐interval on ECG when compared with standard care without HCQ, or placebo (very low‐certainty evidence).

We have drawn no conclusions from small single‐trial comparisons of CQ versus lopinavir/ritonavir; HCQ and azithromycin versus standard care; and HCQ versus febuxostat.

Objective 2. For prevention of COVID‐19 disease in people at risk of exposure to SARS‐CoV‐2

No eligible studies were identified for this objective.

Objective 3. For prevention of COVID‐19 disease in people who have been exposed to SARS‐CoV‐2

One individually randomized trial compared HCQ with placebo (see summary of findings Table 2). We are very uncertain about the effect of HCQ on the development of confirmed COVID‐19 at 14 days from enrolment and the risk of hospitalization due to COVID‐19, compared with placebo (very low‐certainty evidence). HCQ probably increases the risk of adverse events, compared with placebo (moderate‐certainty evidence). HCQ may result in little or no difference in serious adverse events, compared with placebo, though no participants in the trial experienced any events (low‐certainty evidence).

A cluster‐randomized trial compared HCQ with no intervention for postexposure prevention of COVID‐19. The results of this trial could not be combined with those of the individually randomized RCT. There was no difference in the risk of symptomatic confirmed COVID‐19 or production of antibodies to SARS‐CoV‐2 between study arms.

Overall completeness and applicability of evidence

Objective 1. For treatment of COVID‐19 disease

The 12 included trials were conducted in Brazil, Canada, China, Egypt, Iran, Spain, Taiwan, the UK, and the USA. The largest trial, contributing the majority of participants (4716/8569, 55%), was based in the UK. It is as yet uncertain whether geographical differences may impact on the efficacy or safety of interventions for the treatment of COVID‐19.

None of the trials included children or pregnant women, so the evidence cannot be applied to these populations. Most participants (86%) had COVID‐19 confirmed by positive RT‐PCR for SARS‐CoV‐2. Nine of the 12 trials recruited hospitalized patients, with the three ambulatory treatment trials contributing only 844/8569 (10%) of participants in the review, potentially affecting applicability of the findings to outpatient settings.

Severity of disease varied between trials. Whilst not all participants could be classified according to WHO severity grading, 3139/8569 (37%) of participants did not require oxygen or other respiratory support at enrolment; 5230/8569 (63%) were receiving oxygen or higher respiratory support. The two largest trials (Horby 2020; Pan 2020), which mostly included participants requiring oxygen or higher respiratory support, contributed the majority of participants to the meta‐analysis of the outcome death due to any cause for the comparison of HCQ versus standard care or placebo. Data for participants with any or serious adverse events could not be extracted from these trials. This means that evidence for the outcome of death was based on a population with more severe disease. For adverse events outcomes, the meta‐analysed population was less severely unwell, and so this effects estimate should be interpreted with this in mind as the baseline risk of adverse events in more severely unwell patients is likely to be higher. These trials were designed to assess the efficacy of HCQ, and may not be of sufficient power to detect any but the most common adverse events.

HCQ and CQ have similar pharmacological actions, but only one study used CQ, to which 10 participants were allocated, and so we could not draw conclusions about the efficacy and safety of CQ for the treatment of COVID‐19. This is likely due to the increased rate of adverse effects seen with CQ compared with HCQ in other conditions.

Only one trial included an arm with a combination of HCQ and azithromycin (217 participants), and so few conclusions can be drawn about the efficacy or safety of this combination treatment.

Dosing of HCQ varied considerably between trials (see Table 4). The two largest trials used relatively high total cumulative doses, and so it is unlikely that a lack of efficacy for the primary outcome of death is due to underdosing. As the data for adverse events were drawn from the trials using lower doses, it is possible that this meta‐analysis underestimates dose‐dependent adverse events.

Pharmacological co‐interventions also varied considerably between studies (see Table 5), and reporting was at times incomplete. Co‐interventions were mostly balanced between intervention arms across the studies, and are unlikely to have impacted on the summary effects estimates for the primary outcome.

Objective 2. For preventing COVID‐19 disease in people at risk of exposure to SARS‐CoV‐2

No eligible studies were identified for this objective.

Objective 3. For preventing COVID‐19 disease in people who have been exposed to SARS‐CoV‐2

One of the two trials included in this objective was conducted in the USA and Canada; the other in Spain. Most participants were healthcare or nursing home workers and had no comorbidities; average age was between 40 and 50 years. Consequently, the findings may have limited applicability of the evidence to older people with multi‐morbidity, household contacts, and possibly to lower‐income settings. Additionally, as the assessment for the development of COVID‐19 was based on the presence of symptoms, and no outcomes assessed infection or disease in household or other contacts of the person with exposure to SARS‐CoV‐2, no evidence was available for the effect of HCQ on the risk of asymptomatic infection or onward transmission.

Certainty of the evidence

We used the GRADE approach to assess the certainty of the evidence, employing GRADEpro GDT software (GRADEpro GDT). The GRADE assessment with explanatory footnotes is outlined in summary of findings Table 1 and summary of findings Table 2.

For Objective 1 ‐ treatment of COVID‐19, we included nine RCTs and assessed seven outcomes. We graded the effect estimate for death as high certainty, implying that treatment with HCQ results in no difference to death from any cause in people with COVID‐19. We graded the effect estimate for negative SARS‐CoV‐2 PCR at 14 days as low certainty, that is HCQ may make no difference to the proportion of people who have a negative SARS‐CoV‐2 swab at 14 days; the certainty of the evidence was downgraded by one level for serious risk of bias, as both trials in this analysis were at high risk of bias across several domains; and one level for serious indirectness, as almost all participants had mild or moderate COVID‐19, all were hospitalized, and all were from one country (Chen 2020a; Tang 2020). We graded the effect estimate for progression to mechanical ventilation as moderate certainty, implying that HCQ probably has no effect on progression to mechanical ventilation in people with COVID‐19; the certainty of the evidence was downgraded by one level for serious imprecision, as the lower bound of the confidence interval around the estimate represents no benefit nor harm from HCQ, whereas the upper bound suggests appreciable harm. For time to clinical improvement, we graded the estimate of effect as very low certainty, that is we do not know what effect HCQ has on this outcome. Data for this outcome came from a single trial (Tang 2020); we downgraded the certainty of the evidence for serious risk of bias, serious indirectness (all hospitalized patients with mild‐moderate COVID‐19 in one centre in China), and serious imprecision (confidence interval extends from appreciable benefit to appreciable harm).

For adverse effects in people with COVID‐19 treated with HCQ, we graded the effects estimate for participants with any adverse events as moderate certainty, meaning that HCQ probably increases the risk of developing adverse events. We downgraded the certainty of the evidence by one level for serious risk of bias, as all trials contributing to this analysis had high or unclear risk of bias across various domains, and all but one trial were open‐label. We graded the effects estimate for participants with serious adverse events as low certainty, downgrading by two levels for very serious imprecision, as the confidence intervals ranged from appreciable benefit to appreciable harm; overall the rate of serious adverse events was low. We graded the effects estimate for participants who developed prolonged QT interval on ECG as very low certainty; data for this outcome came from one trial, and the certainty of the evidence was downgraded for risk of bias, as the trial was open‐label; indirectness, as severe COVID patients were excluded; and imprecision, as the low event rate and small sample size led to a broad confidence interval.

We found no studies addressing Objective 2 ‐ prevention of COVID‐19 disease in people at risk of exposure to SARS‐CoV‐2.

For Objective 3 ‐ prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2, we included one RCT and graded four outcomes (Boulware 2020). We graded the effects estimate for development of COVID‐19 at 14 days from enrolment as very low certainty, implying that we do not know whether HCQ prevents COVID‐19 in people exposed to SARS‐CoV‐2. We downgraded the certainty of the evidence by one level for serious indirectness, as data for this outcome came from a single trial in North America with few older or comorbid participants; and by two levels for very serious imprecision, as the confidence interval around the effects estimate included appreciable benefit and appreciable harm. We graded the effect estimate for participants hospitalized due to COVID‐19 as very low certainty, again downgrading by one level for serious indirectness and by two levels for very serious imprecision. We graded the effects estimate for participants with any adverse events as moderate certainty, implying that HCQ probably increases the risk of adverse events when compared with placebo; the certainty of the evidence was downgraded by one level for serious indirectness, as described above. We graded the effects estimate for participants with serious adverse events as low certainty, meaning that HCQ may result in no difference to the risk of developing serious adverse events compared with placebo; the certainty of the evidence was downgraded by one level for serious indirectness and one level for serious imprecision.

Potential biases in the review process

We took measures to limit bias in the review process by following the procedures outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). The Cochrane Infectious Diseases Group (CIDG) Information Specialist conducted the literature search using a variety of general and COVID‐19 specific resources, and included preprints. In addition, we also checked the COVID‐NMA website at www.covid-nma.com/ for further studies at regular intervals. We did not make a funnel plot, as fewer than 10 studies were included per comparison. Two review authors independently examined the search results, assessed studies for eligibility, and extracted data, in order to minimize bias in study selection and data extraction.

Agreements and disagreements with other studies or reviews

Several systematic reviews have been published examining the treatment of COVID‐19 with HCQ/CQ, all of which have included RCTs and non‐randomized studies. For the most part their conclusions match ours regarding the finding of HCQ showing no benefit for mortality from COVID‐19, but with less precision. Fiolet 2020, published in August 2020, describes results from 29 studies including 3 RCTs, but studies with no mortality were excluded. In participants treated with HCQ versus comparator group for the outcome of death, the RR was 0.83 (95% CI 0.65 to 1.06); excluding non‐randomized studies, the RR was 1.09 (95% CI 0.97 to 1.24). The authors concluded that HCQ is not effective for COVID‐19, and that further research is not needed. Elavarasi 2020, published in September 2020, is a systematic review of RCTs, case series, and cohort studies with a comparator arm including 12 non‐randomized studies and 3 RCTs. Meta‐analysis of the included studies revealed no difference in mortality with HCQ use (RR 0.98 95% CI 0.66 to 1.46), leading the authors to conclude that the available evidence does not support the use of HCQ and that further RCTs are required. Hernandez 2020, published in August 2020, is a living systematic review which includes 3 RCTs, 8 cohort studies, and 3 case series. No meta‐analysis was conducted due to high heterogeneity between studies; the authors concluded that the evidence on the benefits and harms of HCQ for COVID‐19 is weak and conflicting. Zang 2020, published in September 2020, includes 3 RCTs, 2 prospective observational studies, and 2 retrospective observational studies. In participants treated with HCQ compared with standard therapy, meta‐analysis suggested increased mortality with HCQ (RR 1.92, 95% CI 1.26 to 2.93), although the authors identified significant unexplained heterogeneity and problems with study quality, and concluded that better RCTs are urgently needed. All these systematic reviews cite the three Chinese RCTs included in this review (Chen 2020a; Chen 2020b; Tang 2020). Few systematic reviews have used GRADE to assess the certainty of the evidence.

There are fewer studies and fewer reviews examining CQ and HCQ as prophylaxis for COVID‐19 (Objectives 2 and 3). Shah 2020 is a systematic review of the evidence for HCQ in preventing COVID‐19, which was published in March 2020. Due to the lack of studies at that time, the authors included only two pre‐clinical studies and three commentaries, concluding that although evidence from pre‐clinical studies is promising, there was no evidence to support the efficacy of CQ or HCQ in preventing COVID‐19.

National and international guideline recommendations for the use of CQ and HCQ have changed over the course of the pandemic. The US National Institutes of Health published updated guidance on 27 August 2020 recommending against the use of CQ or HCQ for the treatment of COVID‐19 in hospitalized patients, and against the use of CQ or HCQ in non‐hospitalized patients except in the context of a clinical trial (NIH 2020). In May 2020, WHO recommended that CQ and HCQ not be administered to COVID‐19 patients outside of the context of a clinical trial (WHO 2020c).

Figure 1

Study flow diagram.

Figure 2

Risk of bias summary: review authors' judgements about each 'Risk of bias' item for each included trial.

Analysis 1.1

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 1: Death due to any cause

Analysis 1.2

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 2: Death due to any cause (sensitivity analysis)

Analysis 1.3

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 3: Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment

Analysis 1.4

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 4: Negative PCR for SARS‐CoV‐2 on respiratory samples at day 7 from enrolment

Analysis 1.5

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 5: Proportion admitted to hospital (if receiving ambulatory treatment)

Analysis 1.6

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 6: Progression to mechanical ventilation

Analysis 1.7

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 7: Length of hospital admission (in days)

Analysis 1.8

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 8: Time to clinical improvement

Analysis 1.9

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 9: Time to negative PCR for SARS‐CoV‐2 on respiratory samples

Analysis 1.10

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 10: Participants with any adverse events

Analysis 1.11

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 11: Participants with serious adverse events

Analysis 1.12

Comparison 1: HCQ versus standard care without HCQ, or placebo, for treatment, Outcome 12: Participants with prolongation of QT‐interval on electrocardiogram

Analysis 2.1

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 1: Negative PCR for SARS‐CoV‐2 on respiratory samples at day 7 from enrolment

Analysis 2.2

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 2: Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment

Analysis 2.3

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 3: Discharge from hospital at day 14 from enrolment

Analysis 2.4

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 4: Clinical improvement at day 10 from enrolment

Analysis 2.5

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 5: Total adverse events

Analysis 2.6

Comparison 2: CQ versus lopinavir/ritonavir for treatment, Outcome 6: Serious adverse events

Analysis 3.1

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 1: Death due to any cause

Analysis 3.2

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 2: Progression to mechanical ventilation

Analysis 3.3

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 3: Length of hospital stay in days

Analysis 3.4

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 4: Participants with any adverse events

Analysis 3.5

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 5: Participants with serious adverse events

Analysis 3.6

Comparison 3: HCQ + azithromycin versus standard care for treatment, Outcome 6: Participants with prolongation of QT‐interval on electrocardiogram

Analysis 4.1

Comparison 4: HCQ versus febuxostat for treatment, Outcome 1: Death due to any cause

Analysis 4.2

Comparison 4: HCQ versus febuxostat for treatment, Outcome 2: Admission to hospital

Analysis 5.1

Comparison 5: HCQ versus placebo for postexposure prophylaxis, Outcome 1: Development of confirmed COVID‐19 at 14 days from enrolment

Analysis 5.2

Comparison 5: HCQ versus placebo for postexposure prophylaxis, Outcome 2: Patients hospitalized due to COVID‐19

Analysis 5.3

Comparison 5: HCQ versus placebo for postexposure prophylaxis, Outcome 3: Participants with any adverse events

Analysis 5.4

Comparison 5: HCQ versus placebo for postexposure prophylaxis, Outcome 4: Participants with serious adverse events

Summary of findings 1. Hydroxychloroquine (HCQ) compared to standard care or placebo for the treatment of people with COVID‐19

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Patients or population: adults with mild to severe COVID‐19 Settings: hospital inpatients and ambulatory care in the community Intervention: HCQ Comparison: standard care or placebo (no HCQ)
Outcomes	Risk with standard care or placebo	Risk with HCQ	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Death due to any cause	18 per 100	19 per 100 (18 to 21)	RR 1.09 (0.99 to 1.19)	8208 (9 RCTs)^a	⨁⨁⨁⨁ HIGH ^b,c	HCQ results in little or no difference to death due to any cause.
Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment^d	83 per 100	83 per 100 (76 to 91)	RR 1.00 (0.91 to 1.10)	213 (3 RCTs)^e	⨁⨁◯◯ LOW ^f,g	HCQ may make little or no difference to proportion of people having negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment.
Progression to mechanical ventilation	8 per 100	9 per 100 (7 to 11)	RR 1.11 (0.91 to 1.37)	4521 (3 RCTs)^h	⨁⨁⨁◯ MODERATE ^i,j	HCQ probably results in little to no difference in progression to mechanical ventilation.
Time to clinical improvement	28 per 100	28 per 100 (18 to 44)	HR 1.01 (0.59 to 1.74)	119 (1 RCT)^k	⨁◯◯◯ VERY LOW ^f,l,m	We are uncertain whether HCQ increases or decreases the proportion of people with clinical improvement at day 28 from enrolment.
Participants with any adverse events	16 per 100	46 per 100 (24 to 90)	RR 2.90 (1.49 to 5.64)	1394 (6 RCTs)ⁿ	⨁⨁⨁◯ MODERATE ^o,p	HCQ probably increases the risk of developing adverse events.
Participants with serious adverse events	36 per 1000	30 per 1000 (13 to 64)	RR 0.82 (0.37 to 1.79)	1004 (6 RCTs)^q	⨁⨁◯◯ LOW ^r	HCQ may result in little or no difference to risk of serious adverse events.
Participants with prolongation of QT‐interval on ECG	2 per 100	17 per 100 (2 to 100)	RR 8.47 (1.14 to 63.03)	147 (1 RCT)^s	⨁◯◯◯ VERY LOW ^t,u,v	The evidence is very uncertain about the effect of HCQ on prolongation of QT‐interval on ECG.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; ECG: electrocardiogram; HCQ: hydroxychloroquine; HR: hazard ratio; PCR: polymerase chain reaction RCT: randomized controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aAbd‐Elsalam 2020; Cavalcanti 2020; Chen 2020a; Chen 2020c; Horby 2020; Mitjà 2020a; Pan 2020; Skipper 2020; Tang 2020. Of these, no participants died in Chen 2020a; Chen 2020c; Mitjà 2020a; Tang 2020. ^bNot downgraded for risk of bias, as most of the evidence comes from Horby 2020 and Pan 2020, which have low risk of bias for this outcome. ^cNot downgraded for indirectness, but it is noted that the population in the largest trial, Horby 2020, was mostly severely/critically unwell. ^dThis was selected as the most relevant of three related outcomes reported by trials in this review. Analyses for the other outcomes (time to negative PCR for SARS‐CoV‐2 on respiratory samples; negative PCR for SARS‐CoV‐2 at day 7 from enrolment) did not demonstrate an important benefit/harm. ^eChen 2020a; Chen 2020c; Tang 2020. ^fDowngraded by one level for serious indirectness: almost all people had mild or moderate COVID‐19; all were hospitalized; and all were from one region. ^gNot downgraded for imprecision: narrow confidence interval, not including appreciable benefit nor harm. The sample size has approximately 80% power to detect an absolute difference of 13%, or 90% power to detect an absolute difference of 15%, in this outcome for the group receiving HCQ versus those receiving standard care. ^hCavalcanti 2020; Horby 2020; Tang 2020. ⁱNot downgraded for indirectness: the three trials all recruited participants admitted to hospital. ^jDowngraded by one level for serious imprecision: lower confidence interval bound represents no benefit nor harm from HCQ, whereas the upper bound suggests appreciable harm. ^kTang 2020. ^lDowngraded by one level for serious risk of bias: unclear risk of attrition and reporting bias, and high risk of other bias. ^mDowngraded by one level for serious imprecision: lower confidence interval bound represents appreciable harm from HCQ, whereas the upper bound suggests no appreciable benefit. ⁿCavalcanti 2020; Chen 2020a; Chen 2020b; Mitjà 2020a; Skipper 2020; Tang 2020. ^oDowngraded by one level for serious risk of bias: all trials except Skipper 2020 were open‐label. Chen 2020a had a high risk of selection and reporting bias; Chen 2020b a high risk of performance, detection, and reporting bias and unclear risk of selection bias; Mitjà 2020a a high risk of performance, detection, attrition, and reporting bias for this outcome, and unclear risk of selection bias; Skipper 2020 a high risk of reporting bias and unclear risk of attrition bias; and Tang 2020 an unclear risk of attrition and reporting bias. We deemed Skipper 2020, Mitjà 2020a, and Tang 2020 as at high risk of other bias. ^pNot downgraded for inconsistency: despite high statistical heterogeneity (I² = 87%), all of the effect estimates were above a risk ratio of 1, with only one trial having a confidence interval that crossed 1. ^qCavalcanti 2020; Chen 2020a; Chen 2020b; Chen 2020c; Skipper 2020; Tang 2020. ^rDowngraded by two levels for very serious imprecision: low number of events, and lower confidence interval bound represents appreciable harm from HCQ, whereas the upper bound includes appreciable benefit. ^sCavalcanti 2020. ^tDowngraded by one level for risk of bias: Cavalcanti 2020 was unblinded, which could have led to detection bias, meaning more participants with QT prolongation were identified in the HCQ group. ^uDowngraded by one level for indirectness: Cavalcanti 2020 included only hospitalized patients, and excluded participants with severe disease, in whom problems with drug interactions and cardiac arrhythmia are more likely. ^vDowngraded by one level for imprecision: Cavalcanti 2020 had a low event rate for this outcome, and a small sample size leading to a very broad confidence interval.

Summary of findings 1. Hydroxychloroquine (HCQ) compared to standard care or placebo for the treatment of people with COVID‐19

Summary of findings 2. Hydroxychloroquine (HCQ) compared to placebo for the prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2

Outcomes	*Anticipated absolute effects (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Patients or population: people who have been exposed to SARS‐CoV‐2 Settings: community Intervention: HCQ Comparison: placebo
Outcomes	Risk with placebo	Risk with HCQ	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Development of confirmed COVID‐19 at 14 days from enrolment	2 per 100	2 per 100 (1 to 6)	RR 1.20 (0.50 to 2.87)	821 (1 RCT)	⨁◯◯◯ VERY LOW ^a,b	The evidence is very uncertain about the effect of HCQ on development of confirmed COVID‐19 at 14 days from enrolment.
Hospitalized due to COVID‐19^c	2 per 1000	2 per 1000 (0 to 31)	RR 0.98 (0.06 to 15.66)	821 (1 RCT)	⨁◯◯◯ VERY LOW ^a,b	The evidence is very uncertain about the effect of HCQ on risk of being hospitalized due to COVID‐19.
Participants with any adverse events	17 per 100	41 per 100 (31 to 53)	RR 2.39 (1.83 to 3.11)	700 (1 RCT)	⨁⨁⨁◯ MODERATE ^a	HCQ probably increases the risk of adverse events when compared with placebo.
Participants with serious adverse events	0 per 1000	0 per 1000 (0 to 0)	Not estimable	700 (1 RCT)	⨁⨁◯◯ LOW ^a,d	HCQ may result in little or no difference in serious adverse events when compared with placebo.
The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect* of the intervention (and its 95% CI). CI: confidence interval; HCQ: hydroxychloroquine; RCT: randomized controlled trial; RR: risk ratio.
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded by one level for serious indirectness: one trial, limited to North America; few older and comorbid participants, possibly due to social media‐based recruitment and internet‐based data collection (Boulware 2020). ^bDowngraded by two levels for very serious imprecision: confidence interval around effect estimate includes appreciable benefit and appreciable harm. ^cThis outcome, as reported by Boulware 2020, was closest to our predefined outcome of 'disease severity of participants who develop COVID‐19, as defined by study authors'. ^dDowngraded by one level for imprecision: no events in either group, therefore risk ratio is not estimable. The optimal information size to be confident that this is a true reflection of risk of serious adverse events would be larger than the total number of participants in this trial. Risk difference = 0% (95% CI −1% to 1%).

Summary of findings 2. Hydroxychloroquine (HCQ) compared to placebo for the prevention of COVID‐19 in people who have been exposed to SARS‐CoV‐2

Table 1. Ongoing trials for treatment: actively recruiting or completed; not yet published

Trial registration number; trial registry	Location(s)	Interventions; abbreviated name	Recruitment status	Estimated completion	Target enrolment
NCT02735707 ClinicalTrials.gov	13 countries; registered in the Netherlands	Adaptive platform trial including HCQ, or HCQ + lopinavir/ritonavir, vs no HCQ REMAP‐CAP	Recruiting	December 2021	7100
NCT04351724 ClinicalTrials.gov	Austria	Platform trial including CQ/HCQ vs placebo ACOVACT	Recruiting	December 2020	500
NCT04328012 ClinicalTrials.gov	USA	Pragmatic adaptive HCQ vs lopinavir/ritonavir vs losartan vs placebo COVID MED	Recruiting	January 2021	4000
NCT04334382 ClinicalTrials.gov	USA	HCQ vs azithromycin HyAzOUT	Recruiting	December 2020	1550
NCT04332991 ClinicalTrials.gov	USA	HCQ vs placebo for hospitalized patients with COVID‐19 ORCHID	Completed	April 2021	510
NCT04363827 ClinicalTrials.gov	Italy	HCQ vs observation PROTECT	Recruiting	September 2020	2300
NCT04359953 ClinicalTrials.gov	France	HCQ vs telmisartan vs azithromycin	Recruiting	June 2021	1600
NCT04356495 ClinicalTrials.gov	France	HCQ vs favipiravir vs imatinib vs telmisartan vs placebo COVERAGE	Recruiting	July 2020	1057
PACTR202004801273802 Pan African Clinical Trials Registry	Nigeria	CQ vs HCQ vs placebo	Recruiting	October 2020	600
ISRCTN86534580 ISRCTN registry	UK	HCQ vs standard care for treatment	Recruiting	March 2021	3000
NCT04324463 ClinicalTrials.gov	Canada	Azithromycin plus hydroxychloroquine or chloroquine (AZCT) vs AZCT plus interferon beta vs interferon beta vs usual care	Recruiting	September 2020	1500
NCT04345289 ClinicalTrials.gov	Denmark	Convalescent plasma vs sarilumab vs HCQ vs baricitinib vs intravenous and subcutaneous placebo vs oral placebo	Recruiting	June 2021	1500
NCT04358068 ClinicalTrials.gov	USA and Puerto Rico	HCQ vs azithromycin	Completed	October 2020	2000
NCT04340544 ClinicalTrials.gov	Germany	HCQ vs placebo	Recruiting	November 2021	2700
NCT04338698 ClinicalTrials.gov	Pakistan	HCQ vs oseltamivir vs azithromycin	Recruiting	September 2020	500
NCT04353037 ClinicalTrials.gov	USA	HCQ vs placebo	Recruiting	April 2021	850
NCT04321616 ClinicalTrials.gov	Norway	HCQ vs remdesivir vs standard care	Recruiting	August 2020	700
ACTRN12620000445976 ANZCTR	Australia and New Zealand	HCQ vs lopinavir/ritonavir vs HCQ plus lopinavir/ritonavir vs standard care	Recruiting	Not reported	2500
NCT04315948 ClinicalTrials.gov	France and Luxembourg	HCQ vs remdesivir vs lopinavir/ritonavir vs interferon beta‐1A vs standard care	Recruiting	March 2023	3100
UTN‐A27736297878 Ensaiosclinicos.gov.br	Brazil	HCQ vs placebo	Recruiting	July 2020	1300
NCT04410562 ClinicalTrials.gov	Spain	HCQ vs placebo (pregnant women)	Recruiting	May 2021	714
NCT04392973 ClinicalTrials.gov	Saudi Arabia	HCQ with favipiravir vs standard care	Recruiting	November 2021	520
CQ, chloroquine; HCQ, hydroxychloroquine

Table 1. Ongoing trials for treatment: actively recruiting or completed; not yet published

Table 2. Ongoing trials for prevention: actively recruiting or completed; not yet published

Trial registration number; trial registry	Location(s)	Interventions; population; abbreviated name	Recruitment status	Estimated completion	Target enrolment
NCT04333732 ClinicalTrials.gov	USA	Low‐/medium‐/high‐dose chloroquine vs placebo Healthcare workers	Recruiting	February 2021	55,000
NCT04303507 ClinicalTrials.gov	Europe, Asia, Africa	HQC vs CQ vs placebo Healthcare workers COPCOV	Recruiting	April 2021	40,000
NCT04334928 ClinicalTrials.gov	Spain	Emtricitabine/tenofovir (Truvada) vs HCQ vs Truvada + HCQ vs placebo Healthcare workers EPICOS	Recruiting	June 2020	4000
NCT04334148 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	July 2020	15,000
NCT04363450 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers (pre‐exposure) HCQPreP	Recruiting	July 2020	1700
NCT04318444 ClinicalTrials.gov	USA	HCQ vs placebo Household contacts (postexposure)	Recruiting	March 2021	1600
NCT04341441 ClinicalTrials.gov	USA	Daily HCQ vs weekly HCQ vs placebo Healthcare workers and first responders	Recruiting	June 2020	3000
IRCT20190122042450N4 Iranian Clinical Trials Registry	Iran	HCQ vs no HCQ All contacts (postexposure)	Completed	Not reported	1000
ISRCTN14326006 ISRCTN registry	Canada	HCQ vs placebo Healthcare workers	Recruiting	January 2022	988
NCT04363827 ClinicalTrials.gov	Italy	HCQ vs no HCQ All contacts	Recruiting	September 2020	2300
NCT04352933 ClinicalTrials.gov	UK	HCQ weekly vs HCQ daily vs placebo Healthcare workers	Recruiting	October 2020	1000
NCT04353037 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	April 2021	850
ACTRN12620000501943 ANZCTR	Australia	HCQ vs placebo Healthcare workers	Recruiting	December 2020	2250
NCT04374942 ClinicalTrials.gov	USA	HCQ vs placebo Healthcare workers	Recruiting	January 2022	988
EudraCT 2020‐001987‐28 EudraCT	Italy	HCQ vs no HCQ Healthcare workers	Recruiting	Not reported	1000
CQ, chloroquine; HCQ, hydroxychloroquine

Table 2. Ongoing trials for prevention: actively recruiting or completed; not yet published

Table 3. Summary of characteristics of included studies

Study

Objective; comparisons

Study design

Countries; recruitment dates

Age

Number of participants in primary comparison

Types of participant at enrolment (type of contact; place of care; disease severity)

Abd‐Elsalam 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Egypt

March to June 2020

HCQ: mean 40.4 y (SD 18.7 y)

Standard care: mean 41.1 y (SD 20.1 y)

194 total: 97 HCQ; 97 standard care

All hospitalized.

“The patients were randomized equally between the two groups regarding the disease severity.” (Numbers not reported.)

Boulware 2020

3: Postexposure prophylaxis

5: HCQ vs placebo (individually randomized)

RCT, double‐blind

USA and Canada

17 March to 6 May 2020

HCQ: median 41 y (IQR 33 to 51)

Placebo: median 40 y (IQR 32 to 50)

821 total: 414 HCQ; 407 placebo

HCQ: 275 healthcare contacts; 125 household contacts; 14 NR

Placebo: 270 healthcare contacts; 120 household contacts; 17 NR

Cavalcanti 2020

1: Treatment

1: HCQ vs standard care

3: HCQ + azithromycin vs standard care

RCT, open‐label

Brazil

29 March to 17 May 2020

HCQ + azithromycin: mean 49.6 y (SD 14.2 y)

HCQ: mean 51.3 y (SD 14.5 y)

Standard care: mean 49.9 y (SD 15.1 y)

665 total: 217 HCQ + azithromycin; 221 HCQ; 227 standard care

All hospitalized.

HCQ + azithromycin: 125/217 mild; 92/217 moderate disease

HCQ: 132/221 mild; 89/221 moderate disease

Standard care: 130/227 mild; 97/227 moderate disease

Chen 2020a

1: Treatment

1: HCQ vs standard care

RCT, open‐label

China

6 February to 25 February 2020

HCQ: mean 50.5 y (SD 3.8 y)

Standard care: mean 46.7 y (SD 3.6 y)

30 total: 15 HCQ; 15 standard care

All hospitalized.

All 30 participants had moderate disease.

Chen 2020b

1: Treatment

1: HCQ vs standard care

RCT, double‐blind (no placebo)

China

4 February to 28 February 2020

HCQ: mean 44.1 y (SD 16.1 y)

Standard care: mean 45.2 y (SD 14.7 y)

62 total: 31 HCQ; 31 standard care

All hospitalized.

All 62 participants had mild disease.

Chen 2020c

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Taiwan

1 April to 31 May 2020

HCQ: mean 33 y (SD 12 y)

Standard care: mean 32.8 y (SD 8.3 y)

33 total: 21 HCQ; 12 standard care

All hospitalized.

HCQ: 19/21 mild; 2/21 moderate

Standard care: 10/12 mild; 2/12 moderate

Davoodi 2020

1: Treatment

4: HCQ vs febuxostat

RCT, open‐label

Iran

16 March to 10 April 2020

HCQ: mean 57.3 y (standard error 2.2 y)

Febuxostat: mean 58 y (standard error 1.47 y)

54 total: 25 HCQ; 29 febuxostat

All ambulatory patients, symptomatic, with abnormalities on CT scan of the chest, but no features of severe acute illness or severe underlying chronic disease.

Horby 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

25 March to 5 June 2020

HCQ: mean 65.2 y (SD 15.2 y)

Standard care: mean 65.4 y (SD 15.4 y)

4716 total: 1561 HCQ; 3155 standard care

All hospitalized.

Inferred from level of oxygen/respiratory support need:

HCQ: 362/1561 asymptomatic/mild (no oxygen received); 938/1561 moderate/severe (received oxygen); 261/1561 critical disease (invasive ventilation)

Standard care: 750/3155 asymptomatic/mild (no oxygen received); 1873/3155 moderate/severe (received oxygen); 532/3155 critical disease (invasive ventilation)

Huang 2020

1: Treatment

2: CQ vs lopinavir/ritonavir (LPV/r)

RCT, open‐label

China

27 January to 15 February 2020

CQ: median 41.5 y (IQR 33.8 to 50 y)

LPV/r: median 53 y (IQR 41.8 to 63.5 y)

22 total: 10 CQ; 12 LPV/r

All hospitalized.

CQ: 7/10 moderate; 3/10 severe disease

LPV/r: 7/12 moderate; 5/12 severe disease

Mitjà 2020a

1: Treatment

1: HCQ vs standard care

RCT, open‐label

Spain

17 March to 26 May 2020

HCQ: mean 41.6 y (SD 12.4 y)

Standard care: mean 41.7 y (SD 12.6 y)

293 total: 136 HCQ; 157 standard care

All ambulatory patients with mild disease, except for 1 patient with severe disease included in the HCQ arm, despite this being an exclusion criterion (included in ITT analysis).

Mitjà 2020b

3: Postexposure prophylaxis

6: HCQ vs standard care (cluster randomized)

Cluster‐RCT, open‐label

Spain

17 March to 28 April 2020

HCQ: mean 48.6 y (SD 18.7 y)

Standard care: mean 48.7 y (SD 19.3 y)

2525 total: 1225 HCQ; 1300 standard care

HCQ: 131 (12%) healthcare workers; 302 (27%) household contacts; 550 (49%) nursing home workers; 133 (12%) nursing home residents

Standard care: 130 (11%) healthcare workers; 338 (28%) household contacts; 584 (49%) nursing home workers; 160 (13%) nursing home residents

Pan 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label

30, across all WHO regions

22 March to 18 June 2020

HCQ: 335 (< 50 years), 410 (50 to 69 years), 202 (≥ 70 years)

Standard care: 317 (< 50 years), 396 (50 to 69 years), 193 (≥ 70 years)

1853 total: 947 HCQ; 906 standard care

All hospitalized.

HCQ: 862/947 moderate or severe (of whom 517 receiving oxygen), 85 critical

Standard care: 824/906 moderate or severe (of whom 483 receiving oxygen), 82 critical

Skipper 2020

1: Treatment

1: HCQ vs placebo

RCT, double‐blind

USA and Canada

22 March to 6 May 2020

HCQ: median 41 y (IQR 33 to 49 y)

Placebo: median 39 y (IQR 31 to 50 y)

491 total: 244 HCQ; 247 placebo

All ambulatory patients, so presumed to have mild disease if symptomatic.

HCQ: 48/244 asymptomatic

Placebo: 52/247 asymptomatic

Tang 2020

1: Treatment

1: HCQ vs standard care

RCT, open‐label (no placebo)

China

11 February to 29 February 2020

HCQ: mean 48 y (SD 14.1 y)

Standard care: mean 44.1 y (SD 15 y)

150 total: 75 HCQ; 75 standard care

All hospitalized.

HCQ: 15/75 mild; 59/75 moderate; 1/75 severe disease

Standard care: 7/75 mild; 67/75 moderate; 1/75 severe disease

Table 3. Summary of characteristics of included studies

Table 4. Dosing regimens in hydroxychloroquine treatment trials1

Study	Hydroxychloroquine (HCQ) dose regimen	Control group	Total hydroxychloroquine dose
Abd‐Elsalam 2020	800 mg on day 1, followed by 400 mg daily for further 14 days (total duration of treatment 15 days)	Standard care	6400 mg
Cavalcanti 2020²	400 mg orally twice daily for 7 days	Standard care	5600 mg
Chen 2020a³	400 mg once daily for 5 days	Standard care	2000 mg
Chen 2020b	200 mg orally twice daily for 5 days	Standard care	2000 mg
Chen 2020c	800 mg on day 1, followed by 400 mg daily for further 6 days (total duration of treatment 7 days)	Standard care	3200 mg
Davoodi 2020	200 mg orally twice daily for 5 days	Standard care	2000 mg
Horby 2020	800 mg at 0 and 6 hours, then 400 mg at 12 hours from first dose and every 12 hourly for 10 days	Standard care	10,000 mg
Mitjà 2020a	800 mg on day 1, followed by 400 mg daily for further 6 days (total duration of treatment 7 days)	Standard care	3200 mg
Pan 2020	2000 mg on day 1, followed by 800 mg daily for further 9 days (total duration of treatment 10 days)	Standard care	9200 mg
Skipper 2020	800 mg (4 tablets) once, then 600 mg (3 tablets) 6 to 8 hours later, then 600 mg (3 tablets) once daily for 4 more days (5 days in total)	Placebo: folic acid in USA and lactose in Canada	3800 mg
Tang 2020	400 mg orally 3 times a day for 3 days, then twice daily from day 4, for a total of 14 days for those with mild/moderate disease and 21 days for those with severe disease	Standard care	12,400 mg mild/moderate disease; 18,000 mg severe disease
¹See Table 5 for co‐interventions given in each trial. ²Cavalcanti 2020 ‐ hydroxychloroquine plus azithromycin group received HCQ 400 mg orally twice daily and azithromycin 500 mg orally once daily for seven days. ³Chen 2020a ‐ additionally, all participants in the HCQ arm had nebulized interferon alpha; 12/15 had umifenovir (Arbidol). Standard care arm: no HCQ; all had nebulized interferon alpha; 10/15 had umifenovir (Arbidol).

Table 4. Dosing regimens in hydroxychloroquine treatment trials1

Table 5. Pharmacological co‐interventions given in treatment trials for comparison 1 (HCQ versus standard care or placebo)

Study	Co‐interventions in HCQ arm	Co‐interventions in comparator arm
Abd‐Elsalam 2020	Authors report: "The Egyptian Ministry of Health (MOH) adopted a standard of care treatment protocol for COVID‐19 patients. It included paracetamol, oxygen, fluids (according to assessment), empiric antibiotic (cephalosporins), oseltamivir if needed (75 mg/12 hours for 5 days), and invasive mechanical ventilation with hydrocortisone for severe cases if PaO2 < 60 mmHg, O2 saturation < 90% despite oxygen or noninvasive ventilation, progressive hypercapnia, respiratory acidosis (pH < 7.3), and progressive or refractory septic shock".
Cavalcanti 2020¹	Corticosteroids 5/221 Oseltamivir 38/221 Aciclovir 1/221 Lopinavir/ritonavir 0/221 Ceftriaxone 86/221 Ceftaroline 11/221 Piperacillin/tazobactam 8/221 Oxacillin 0/221 Vancomycin 1/221 Carbapenem 6/221 Quinolone 22/221 No other antiviral, antibiotic, or corticosteroids 21/221	Corticosteroids 8/227 Oseltamivir 51/227 Aciclovir 0/227 Lopinavir/ritonavir 0/227 Ceftriaxone 99/227 Ceftaroline 17/227 Piperacillin/tazobactam 15/227 Oxacillin 1/227 Vancomycin 4/227 Carbapenem 3/227 Quinolone 28/227 No other antiviral, antibiotic, or corticosteroids 18/227
Chen 2020a²	Nebulized interferon alpha 15/15 Umifenovir 12/15	Nebulized interferon alpha 15/15 Umifenovir 10/15
Chen 2020b	Authors report “all received the standard treatment (oxygen therapy, antiviral agents, antibacterial agents, and immunoglobulin, with or without corticosteroids)”.
Chen 2020c³	Azithromycin 1/21	Azithromycin 2/12
Horby 2020⁴	Dexamethasone 8% Azithromycin 17%	Dexamethasone 9% Azithromycin 19%
Mitjà 2020a⁵	Cobicistat‐boosted darunavir 49/136	Cobicistat‐boosted darunavir 0/157
Pan 2020	The authors report that co‐medications will appear in supplementary tables, but these are not provided with the currently available preprint publication.
Skipper 2020⁶	Zinc 63/212 Vitamin C 101/212	Zinc 53/211 Vitamin C 101/211
Tang 2020	Umifenovir 37/75 Ribavirin 13/75 Lopinavir/ritonavir 13/75 Oseltamivir 8/75 Entecavir 1/75 Antibiotics 32/75 Corticosteroids 6/75	Umifenovir 33/75 Ribavirin 15/75 Lopinavir/ritonavir 12/75 Oseltamivir 9/75 Entecavir 1/75 Antibiotics 27/75 Corticosteroids 4/75
HCQ, hydroxychloroquine; PaO2, partial pressure of oxygen ¹Cavalcanti 2020 ‐ this was a three‐arm trial, of which the third arm received HCQ + azithromycin. ²Chen 2020a ‐ authors report that two participants received lopinavir/ritonavir, but it is unclear which study arms these participants were in. Whether or not any participants received corticosteroids or antibiotics is not reported. ³Chen 2020c ‐ in addition to the above, authors report: "Both study group and comparison group received standard of care comprising supportive treatment for subjects with mild clinical COVID‐19 symptoms and antimicrobial therapy for subjects presenting with moderate clinical COVID‐19 symptoms. The treatment consisted of: (1) ceftriaxone 2 g daily for 7 days +/‐ azithromycin 500 mg on day 1 and 250 mg on days 2–5; or (2) levofloxacin 750 mg daily for 5 d; or (3) levofloxacin 500 mg daily; or (4) moxifloxacin 400 mg daily for 7–14 days for subjects allergic to ceftriaxone or azithromycin or according to physician discretion. Oseltamivir 75 mg b.i.d. will be administered for 5 days to subjects presenting with concomitant influenza A or B infection". ⁴Horby 2020 ‐ authors presented the percentage of participants in each arm receiving dexamethasone or azithromycin. Data on antibiotics and other antivirals not reported. This trial was a platform trial with other arms testing tocilizumab, azithromycin, and dexamethasone, as well as convalescent plasma. ⁵Mitjà 2020a ‐ the trial was originally designed to test HCQ with cobicistat‐boosted darunavir, but this was modified during the trial as further information became available that cobicistat‐boosted darunavir had no in vitro activity against SARS‐CoV‐2. ⁶Skipper 2020 ‐ whether or not participants received antimicrobials or corticosteroids is not reported.

Table 5. Pharmacological co‐interventions given in treatment trials for comparison 1 (HCQ versus standard care or placebo)

Table 6. Adverse events for HCQ versus standard care without HCQ, or placebo, for treatment

Study	HCQ	No HCQ
Cavalcanti 2020	QTc prolongation (13/89) Arrhythmia (3/199) Bradycardia (1/199) Supraventricular tachycardia (2/199) Pneumothorax (1/199) Bloodstream infection (1/199) Itching (1/199) Nausea (9/199) Anaemia (14/199) Elevated ALT or AST (17/199) Elevated bilirubin (5/199) Hypoglycaemia (1/199) Leucopenia (3/199)	QTc prolongation (1/58) Arrhythmia (1/177) Bradycardia (1/177) Bronchospasm (1/177) Nausea (2/177) Vomiting (1/177) Anaemia (11/177) Elevated ALT or AST (6/177) Elevated bilirubin (2/177) Leucopenia (3/177)
Chen 2020a¹	Transient elevated AST with anaemia (1/15) Diarrhoea (2/15) Fatigue (1/15)	Elevated AST (1/15) Elevated creatinine (1/15)
Chen 2020b	Headache (1/31) Rash (1/31)
Mitjà 2020a	Gastrointestinal disorders (148/169) General disorders (30/169) Infections and infestations (9/169) Injury, poisoning, and procedural complications (1/169) Metabolic and nutrition disorders (2/169) Musculoskeletal and connective tissue disorders (1/169) Nervous system disorders (63/169) Psychiatric disorders (2/169) Renal and urinary disorders (1/169) Reproductive system and breast disorders (1/169) Respiratory, thoracic, and mediastinal disorders (2/169) Skin and subcutaneous tissue disorders (11/169) Vascular disorders (1/169)	Gastrointestinal disorders (7/184) General disorders (1/184) Infections and infestations (12/184) Metabolic and nutrition disorders (1/184) Nervous system disorders (3/184)
Skipper 2020²	Upset stomach/nausea (66/212) Diarrhoea, other GI symptoms, vomiting (50/212) Neurologic (nervousness, irritability, dizziness, vertigo) (20/212) Skin reaction, rash (6/212) Ringing in ears (8/212) Allergic reaction, self‐reported (5/212) Changes in vision (4/212) Warmth, hot flashes, night sweats (2/212) Headache (2/212)	Upset stomach/nausea (26/211) Diarrhoea, other GI symptoms, vomiting (20/211) Neurologic (nervousness, irritability, dizziness, vertigo) (13/211) Skin reaction, rash (2/211) Ringing in ears (5/211) Changes in vision (5/211) Taste, dry mouth (1/211) Heart racing, anxiety, panic attack (1/211)
Tang 2020	Disease progression (1/70) Upper respiratory tract infection (1/70) Diarrhoea (7/70) Vomiting (2/70) Nausea (1/70) Abdominal discomfort (1/70) Blurred vision (1/70) Thirst (1/70) Sinus bradycardia (1/70) Hypertension (1/70) Orthostatic hypotension (1/70) Hypertriglyceridaemia (1/70) Decreased appetite (1/70) Fatigue (1/70) Dyspnoea (1/70) Flush (1/70) Kidney injury (1/70) Coagulation dysfunction (1/70) Decreased white blood cell (1/70) Increased ALT (1/70) Increased serum amylase (1/70) Decreased neutrophil count (1/70)	Abdominal bloating (1/80) Fever (1/80) Liver abnormality (1/80) Hepatic steatosis (1/80) Otitis externa (1/80) Increased serum amyloid A (1/80)
ALT: alanine aminotransferase; AST: aspartate transaminase; GI: gastrointestinal; HCQ: hydroxychloroquine ¹Authors of Chen 2020a comment that “among the test group the occurrence of adverse events in subjects with moderate to severe disease is not related to medication. All adverse reactions after drug withdrawal or symptomatic treatment disappeared”. ²Skipper 2020 ‐ authors describe these adverse events as side effects reported at day 5.

Table 6. Adverse events for HCQ versus standard care without HCQ, or placebo, for treatment

Comparison 1. HCQ versus standard care without HCQ, or placebo, for treatment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Death due to any cause Show forest plot	9	8208	Risk Ratio (M‐H, Random, 95% CI)	1.09 [0.99, 1.19]

1.2 Death due to any cause (sensitivity analysis) Show forest plot	9	8043	Risk Ratio (M‐H, Random, 95% CI)	1.09 [0.99, 1.19]

1.3 Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment Show forest plot	3	213	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.91, 1.10]

1.4 Negative PCR for SARS‐CoV‐2 on respiratory samples at day 7 from enrolment Show forest plot	2	180	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.68, 1.09]

1.5 Proportion admitted to hospital (if receiving ambulatory treatment) Show forest plot	1	465	Risk Ratio (M‐H, Random, 95% CI)	0.41 [0.13, 1.27]

1.6 Progression to mechanical ventilation Show forest plot	3	4521	Risk Ratio (M‐H, Random, 95% CI)	1.11 [0.91, 1.37]

1.7 Length of hospital admission (in days) Show forest plot	2	642	Mean Difference (IV, Random, 95% CI)	‐0.15 [‐0.75, 0.45]

1.8 Time to clinical improvement Show forest plot	1		Hazard Ratio (IV, Random, 95% CI)	1.01 [0.59, 1.74]

1.9 Time to negative PCR for SARS‐CoV‐2 on respiratory samples Show forest plot	1		Hazard Ratio (IV, Random, 95% CI)	0.85 [0.58, 1.23]

1.10 Participants with any adverse events Show forest plot	6	1394	Risk Ratio (M‐H, Random, 95% CI)	2.90 [1.49, 5.64]

1.11 Participants with serious adverse events Show forest plot	6	1004	Risk Ratio (M‐H, Random, 95% CI)	0.82 [0.37, 1.79]

1.12 Participants with prolongation of QT‐interval on electrocardiogram Show forest plot	1	147	Risk Ratio (M‐H, Random, 95% CI)	8.47 [1.14, 63.03]

Comparison 1. HCQ versus standard care without HCQ, or placebo, for treatment

Comparison 2. CQ versus lopinavir/ritonavir for treatment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Negative PCR for SARS‐CoV‐2 on respiratory samples at day 7 from enrolment Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	1.20 [0.64, 2.25]

2.2 Negative PCR for SARS‐CoV‐2 on respiratory samples at day 14 from enrolment Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	1.08 [0.85, 1.36]

2.3 Discharge from hospital at day 14 from enrolment Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	1.91 [1.09, 3.34]

2.4 Clinical improvement at day 10 from enrolment Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	1.37 [0.78, 2.42]

2.5 Total adverse events Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	1.08 [0.78, 1.50]

2.6 Serious adverse events Show forest plot	1	22	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

Comparison 2. CQ versus lopinavir/ritonavir for treatment

Comparison 3. HCQ + azithromycin versus standard care for treatment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Death due to any cause Show forest plot	1	444	Risk Ratio (M‐H, Random, 95% CI)	0.52 [0.13, 2.07]

3.2 Progression to mechanical ventilation Show forest plot	1	444	Risk Ratio (M‐H, Random, 95% CI)	1.61 [0.82, 3.15]

3.3 Length of hospital stay in days Show forest plot	1	444	Mean Difference (IV, Random, 95% CI)	0.50 [‐0.81, 1.81]

3.4 Participants with any adverse events Show forest plot	1	416	Risk Ratio (M‐H, Random, 95% CI)	1.74 [1.27, 2.38]

3.5 Participants with serious adverse events Show forest plot	1	416	Risk Ratio (M‐H, Random, 95% CI)	1.85 [0.36, 9.43]

3.6 Participants with prolongation of QT‐interval on electrocardiogram Show forest plot	1	174	Risk Ratio (M‐H, Random, 95% CI)	8.50 [1.16, 62.31]

Comparison 3. HCQ + azithromycin versus standard care for treatment

Comparison 4. HCQ versus febuxostat for treatment

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
4.1 Death due to any cause Show forest plot	1	54	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

4.2 Admission to hospital Show forest plot	1	54	Risk Ratio (M‐H, Random, 95% CI)	1.16 [0.26, 5.24]

Comparison 4. HCQ versus febuxostat for treatment

Comparison 5. HCQ versus placebo for postexposure prophylaxis

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
5.1 Development of confirmed COVID‐19 at 14 days from enrolment Show forest plot	1	821	Risk Ratio (M‐H, Random, 95% CI)	1.20 [0.50, 2.87]

5.2 Patients hospitalized due to COVID‐19 Show forest plot	1	821	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.06, 15.66]

5.3 Participants with any adverse events Show forest plot	1	700	Risk Ratio (M‐H, Random, 95% CI)	2.39 [1.83, 3.11]

5.4 Participants with serious adverse events Show forest plot	1	700	Risk Ratio (M‐H, Random, 95% CI)	Not estimable

Comparison 5. HCQ versus placebo for postexposure prophylaxis