Is hydroxychloroquine safety for COVID-19? a systematic review and meta-analysis of randomized trials

,


INTRODUCTION
Although the war against COVID-19 in China has ushered in the dawn, the worldwide epidemic became more widely over the world.Chloroquine, which is widely used against malaria and autoimmunity diseases, has been reported as a potential broad-spectrum antiviral drug in previous articles.A recent study demonstrated that chloroquine was effective against COVID-19 in vitro 1 , and then it was incorporated into antiviral treatment options in the sixth 2 and seventh trial editions of COVID-19 protocol of China against COVID-19 3 , with the emergence of a series of clinical trials on chloroquine or HCQ treatment of COVID-19 4,5 .Despite the positive efficacy of chloroquine in the treatment of COVID-19, it also raised concerns about the safety of chloroquine.As a derivative of chloroquine, HCQ has a similar antiviral mechanism with chloroquine, but it is well tolerated [6][7][8][9] .The adverse reactions (ADRs) of HCQ involve various systemic organs, among which irreversible retinopathy is the most concerned.Previous studies indicated that the overall prevalence rate of patients receiving HCQ for more than 5 years is 7.5%, which can rise to almost 20% after 20 years 10 .Other frequently reported SAEs were mainly focused on cardiotoxicities such as cardiomyopathy [11][12][13][14] , and cutaneous toxicity such as acute generalized exanthematous pustulosis [15][16][17] , pigmentation [18][19][20] and toxic epidermal necrolysis 21,22 .
Preliminary results from a small, single-center clinical trial conducted by the People's Hospital of Wuhan University revealed good short-term efficacy for 20 patients treated with HCQ 23 .The experts consensus of the comprehensive treatment of coronavirus diseases in Shanghai ("Shanghai protocol") agreed that HCQ would also be one of the main antiviral treatment 4 .According to the Shanghai epidemic prevention and control press conference on March 19, 2020, HCQ ranks as the first therapeutic drug in the "Shanghai protocol".The recent clinical trial performed in Shanghai Public Health Clinical Center enrolling 184 patients suggested that HCQ is effective and safe in the treatment of COVID-19 24,25 .The purpose of this study was to investigate the incidence of AEs of HCQ in randomized controlled trials, and then to provide evidence for the safety of COVID-19 therapy.

METHODS
We conducted the study following the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines 26 .The study protocol was registered in PROSPERO (CRD42020176407).

Search strategies
Randomized controlled studies published in English were searched systematically in PubMed, Embase and Cochrane Library up to February 27, 2020.The Medical Subject Headings "HCQ," "Drug-Related Side Effects and Adverse Reactions," "Adverse event," "Adverse drug reaction," and free text word such as "HCQ," "Adverse drug effect," "Adverse event," "Adverse drug reaction," "random," "randomization," "randomized," "randomly" were combined with the Boolean operator "AND" and "OR".See Tables S1-3 for detailed retrieval strategies.Additionally, references cited in the articles were checked for and found to be available.

Study selections
We had access to all publications which evaluated the safety of HCQ, included randomized controlled studies that were enrolled adult patients and published in English.We excluded studies whose full text was unavailable, as well as studies that not published as randomized controlled studies, studied in children, not reported the safety outcomes, and focused on other irrelevant topics.The primary outcome was the incidence of AEs, which was defined as any undesirable experience associated with the use of a medical product in a patient 27 .The secondary outcome was the incidence of SAEs that defined as death, life-threatening, hospitalization (initial or prolonged), disability or permanent damage, or congenital anomaly/birth defect, or required intervention to prevent permanent impairment or damage (devices) 27 .The literature selection was performed independently by two researchers (C.C and KM.P), and any disagreements were resolved through negotiation or seeking the help of another reviewer.

Data abstraction
Two researchers (C.C and KM.P) extracted the data of the eligible studies independently, including the author, year of publication, country, region, study type, study population, age, HCQ dosage, follow-up time, the occurrence of AEs and SAEs The extracted AEs were classified according to the Medical dictionary for regulatory activities (MedDRA) 28 , and cross-checked by the two researchers (C.C and KM.P).If inconsistencies could not reach a consensus, another reviewer would participate in making a decision.For studies that covered two intervention groups, we will integrate the two intervention groups into one to analyzed 29 .

Risk of bias assessment
The risks of bias of the eligible studies were assessed using the Cochrane Collaboration's tool for assessing risk of bias in randomized trials 30 , which including the following 7 domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and anything else.

Statistical analysis
We categorized AEs by frequencies according to Council for International Organization of Medical Science (CIOMS) as follows: common AEs (risk per 100 participants [?]1/100 and <1/10), uncommon AEs ( risk per 100 participants [?]1/1000 and <1/100) 31 .The pooled incidence of AEs or SAEs was calculated in HCQ and control group, respectively, and the Chi-square test or Fisher's exact test in SPSS software (version 23.0) was used to evaluate the difference in the pooled incidence of AEs or SAEs between the HCQ and the control group.
For AEs reported in two or more studies, we performed meta-analysis using the Review Manager 5.3.5 software 32 , and the Risk Difference (RD) and 95% Confidence Interval (CI) in the pooled incidence of AEs between HCQ and control group was calculated.We used the χ2 test for the exploration of heterogeneity, and the P-value <0.05 was considered significant.Besides, I 2 statistics was used to quantify heterogeneity, and the random-effect model was used for quantitative synthesis, otherwise, the fixed-effect model was used.If there was significant clinical heterogeneity across studies, descriptive analysis was used to present the results.Furthermore, Funnel plots were used to assess the risk of publication bias for outcomes reported by nine or more studies.A P value of less than 0.05 was considered significant.Subgroup analysis was performed for common AEs based on whether the daily doses of HCQ were greater than 400mg.We considered the daily doses greater than 400mg as high dosage group, and less than 400mg as standard dosage group.

RESULTS
A total of 885 articles were identified, of which 58 studies were potentially eligible, and 34 randomized studies  published in English between 1976 and 2020 were finally included after title/abstract and fulltext screening process, the PRISMA 2009 flow diagram in literature screening referred to Fig 1 . Studis excluded in the full-text screening process were listed in Table S4 in the Supplementary Material.
Risks of bias assessment of all eligible studies are presented inFig 2, 3 , most of the studies are considered to be at low or moderate risk of bias in the important domains as measured by the Cochrane Collaboration's tool for assessing risk of bias in randomized trials.The reason for the increased risk of bias is mainly that the researchers or those who evaluated the outcomes were not blinded.

The incidence of AEs
Among 3,639 patients enrolled in the 34 studies, 1,878 patients received HCQ and 1,761 received placebo or not.933 AEs were reported in HCQ group, while 651 in the control group.The cumulative number of AEs in the two groups was significantly different (P<0.0001).When all the AEs were categorized by the system organ class in MedDRA, the gastrointestinal disorders were reported most frequently, of which 290 AEs reported in HCQ group and 199 in control group, and there was a significant difference between the two groups (P<0.0001).In addition, the incidences of skin and subcutaneous tissues disorders (P = 0.011), ear and ear labyrinthine disorders (P = 0.045), renal and urinary disorders (P=0.011), as well as surgical and medical procedures (P = 0.020) in the HCQ group were significantly higher than that in the control group.We presented the pooled incidences of AEs of each system organ in Table 2.Moreover, the cumulative number of common AEs in the HCQ group (381/1878) was significantly increased compared with that in the control group (268/1761) (P<0.0001),among which the pooled incidence of rash was increased obviously (P=0.037).The pooled incidences of common AEs and all AEs referred to Table 3 and Table S5, respectively.The cumulative number of the treatment discontinuation caused by AEs in HCQ group was greatly raised compared with the control group (P= 0.016).

The incidence of SAEs
Twenty-four of the 34 included studies 34,35,37,39,41,42,44,[46][47][48][50][51][52][53][54] , which contributing to 2,760 patients, reported the incidence of SAEs as an outcome. 1,428patients in the HCQ group reported 162 SAEs, while 1,332 patients in the control group reported 132 SAEs.The cumulative number of SAEs in the HCQ group was not significantly greater than that in the control group (P=0.222).When all the SAEs were categorized by the system organ class in MedDRA, the pooled incidence of SAEs in all system organs (P>0.05) between the two groups except for gastrointestinal tract (P=0.005).The pooled incidences of SAEs in each system organ were presented inTable 4 .Of the 2,760 patients, 138 specified SAEs in the HCQ group and 84 in the control group were reported.In HCQ group, the SAE occurred most frequently was neutropenia, but it was not significantly different from the control groups (P = 0.577).Moreover, the patients in HCQ group had low risk of anemia (P = 0.026) and high risk of fatigue (P = 0.045).We listed the pooled incidence of all specified SAEs inTable 5 .

Meta-analysis results
The results of the meta-analysis indicated that the pooled incidence of treatment discontinuation caused by AEs (RD 0.02, 95% CI: 0.00 to 0.06, 23 studies) were significantly different between HCQ and control group using the fixed-effect model.However, the pooled incidences of other AEs in HCQ group were similar to that in control group.The meta-analysis results of each AE were listed in Table 6 .
We performed a descriptive analysis for the 5 studies 42,44,45,53,55 in high dosage groups.Two studies 44,55 conducted in pancreatic cancer receiving gemcitabine and nab-paclitaxel plus HCQ or not, and the dosage of HCQ was 600mg twice a day.One 44 of two studies concluded that the chemotherapy regimen plus HCQ significantly increased neutropenia (P = 0.03) compared with the chemotherapy regimen without HCQ, however, the other one 55 did not find any difference in the incidence of AEs or SAEs between the two groups.In the remaining three studies 42,45,53 , the eligible patients were given a daily dose of 800mg of HCQ, and the incidences of AEs were not significantly increased in two 45,53 of the studies, while the incidence of diarrhea was significantly higher in one study 42 than in the control group.
The funnel plot of the AEs reported by nine or more studies had no obvious asymmetry, indicating that there was no significant publication bias (see figure S1-4).

DISCUSSION
This review found that the cumulative number of all AEs in HCQ group was markedly greater than that of the control group.According to the system organ class, the difference in the pooled incidence of gastrointestinal AEs that occurred most frequently in HCQ group, was significant between the two groups.In addition, the pooled incidence of skin and subcutaneous tissue AEs, ear and labyrinthine AEs, and renal and urinary AEs of HCQ group are also significantly higher than that of control group.Furthermore, the cumulative number of the common AEs of HCQ group was also significantly increased compared with that of the control group, among which the pooled incidence of rash was raised up obviously.Compared with the control group, the cumulative number of SAEs in the HCQ group did not reach a significant increase, but the pooled incidence of SAEs in the gastrointestinal tract was significantly different between the two groups.Moreover, the patients in HCQ group had a high risk of fatigue.The meta-analysis suggested that the pooled incidence of treatment discontinuation caused by AEs in the HCQ group was significantly higher than that in the control group.HCQ or chloroquine, which are currently widely medicated for rheumatoid arthritis, systemic lupus erythematosus and other rheumatic diseases, has been proved to be a potential broad-spectrum antiviral agent 69,70 .Previous studies revealed that HCQ or chloroquine could inhibit retroviruses 51,53,71 , dengue virus 72 , handfoot-mouth virus 73 , avian influenza A (H5N1) virus 74 and coronavirus [75][76][77] .
The antiviral mechanisms of HCQ or chloroquine include inhibiting multiple processes such as virus invasion, transport, and replication by increasing the pH value of vesicle organelles such as intracellular bodies, as well as inhibiting the production and release of tumor necrosis factor and interleukin-6 that mediate the inflammatory complications of various viral diseases 70 .For SARS, it seems to interfere with the glycosylated terminal of the cell receptor angiotensin-converting enzyme-2, thereby negatively affecting the virus-receptor binding, leading to the failure of infection, and finally influencing the infection and transmission of SARS coronavirus within a certain concentration range 78 .
Although a series of studies and case reports [79][80][81][82][83][84][85][86] found that HCQ could increase the risk of retinopathy, this review demonstrated that the pooled incidence of eye AEs in HCQ group was not significantly increased in comparison with the control group and only one specified proliferative retinopathy was reported in eligible studies.The reason might be explained by the follow-up times of randomized studies that included in this review were not abundant enough.one retrospective study found that the potential risk factors for HCQ retinopathy were high dose and long-term (>5 years) treatment by multivariate regression 87 .However, most follow-up times of the included studies were within 12 months, and the longest one was about 40 months.Fin bloom et al. 9 assessed the frequency of retinal toxicity in patients receiving either chloroquine or HCQ, and the overall frequency of retinopathy was 6% (7/110).Compared with the chloroquine group, the risk of the retinopathy in HCQ group was lower (6/31 vs. 0/66).
This study found that the most common AEs and SAEs significantly increased in the HCQ group were gastrointestinal events such as diarrhea, nausea, and vomiting, and a retrospective cohort study showed the incidence of skin and subcutaneous tissue disorders in patients with cutaneous lupus and dermatomyositis was the highest 88 .In this review, the patients receiving HCQ also had a high incidence of skin and subcutaneous tissues AEs, which was significantly different from the patients in the control group.The common skin and subcutaneous tissues AE with a relatively increased risk was rash, and other AEs such pigmentation and itching were also reported frequently.There were two SAEs reported in HCQ group, one was erythema multiforme and the other one was acute generalized erythematous pustulosis, while none reported in control group.The difference in pooled incidence was not significant between the two groups.Previous studies revealed that skin ADRs generally occurred 5˜14 days after the beginning of HCQ, and the rash is characterized by lichen-like, urticaria or rash.Additionally, the symptoms are generally mild, which could be relieved after the withdrawal 89,90 .Skin and subcutaneous tissue SAEs of HCQ that were frequently reported in the literature were acute generalized exanthematous pustulosis (AGEP) [15][16][17][91][92][93][94] , pigmentation 18,20,[95][96][97][98][99][100][101] , Stevens-Johnson syndrome 22,102 , and toxic epidermal necrolysis 21,103,104 . One mltinational case-control study suggested that HCQ or chloroquine was highly associated with AGEP 105 . Besides, the risk factors idenified in previous studies of HCQ-induced pigmentation were previous treatment with oral anticoagulants and/or antiplatelet agents and with higher blood HCQ concentration 100,101 .In addition, we found that the patients in the HCQ group suffered from more ear and labyrinth AEs than in the control group, and the ADR reported more than once was tinnitus.There was no ear and labyrinth SAE reported in all patients enrolled.A case of hearing loss caused by HCQ in HIV-infected patients was reported in a previous publication, however, the hearing was restored two months after the withdrawal 106 .Moreover, patients who received HCQ had increased incidences of renal and urinary AEs compared with the patients without HCQ, and the frequently reported AE was proteinuria. Meawhile, the meta-analysis results suggested that the difference in the pooled incidence of proteinuria between the two groups was not significant.One SAE that reported in renal and urinary system was acute kidney injury.
A number of recent studies have reported the related cardiotoxicity of HCQ [11][12][13][14][107][108][109][110][111][112][113][114] , especially the cardiomyopathy [12][13][14]107,110,111,113,114 .The cardiac SAEs of HCQ reported in eligible studies were arrhythmias and heart failure. A systematic review aboutchloroquine or HCQ cardiac toxicity indicated that the incidence of cardiac AEs is rare, but generally more severe and may be irreversible.The AEs induced by HCQ included cardiomyopathy, atrioventricular block, valve dysfunction, acute myocardial infarction, heart failure, and abnormal left ventricular ejection fraction, and among them, the incidence of cardiomyopathy was higher than chloroquine 115 .However, the study also pointed out that the evidence of cardiotoxicity caused by HCQ was mainly from retrospective studies, and further pharmacovigilance was needed.
For SAEs, although the pooled incidence of neutropenia and neutrophils count decreased were higher in HCQ group, the differences did not reach a statistically significant between the two groups.The publications of Karasic et al. 44 and Zeh et al. 55 contributed to the increased neutropenia and neutrophils count decreased, for all the patients with pancreatic cancer enrolled in were treated with gemcitabine and nab-paclitaxel.Bone marrow suppression is a common ADR to nab-paclitaxel and gemcitabine 116,117 .In this review, the incidence of discontinuation caused by AEs in HCQ group was about 6%, which was significantly higher than that of the control group.
The descriptive analysis suggested that the incidence of neutropenia and diarrhea significantly increased in patients that received HCQ with a daily dose of 800˜1200 mg.The finding of the PLUS study 118 conducted in France indicated that the AEs rates of patients receiving HCQ once daily for 200mg, 400mg, 600mg, and 800mg were 38.9%, 15.5%, 25%, and 27.4%, respectively, and no differences were identified between groups in terms of nausea, vomiting, diarrhea or blurred vision.Moreover, patients who received high HCQ doses for 7 months, including patients with a dose of 800 mg/day, had no significant AEs.Another dose-dependent study enrolled 212 rheumatoid arthritis in a 6-week, double-blind study 38 comparing treatment with HCQ at 400 mg/day, 800 mg/day, and 1,200 mg/day, and the results revealed that gastrointestinal AEs of HCQ were dose-related, while ocular AEs were dose-independent.Besides, another study 119 also proved that there is an association between gastrointestinal AEs and elevated blood HCQ concentrations.
In vitro, based on the PB/PK models, Yao et al. 120 found that a loading dose of 400 mg twice a day of HCQ sulfate given orally, followed by a maintenance dose of 200 mg given twice a day for 4 days reached three times the potency of chloroquine phosphate when given 500 mg twice a day 5 days in advance.Furthermore, in a non-randomized study conducted by Didier Raoulta et al., 121 36 patients who infected with COVID-19 were given HCQ 200mg three times daily for 10 days, and the results demonstrated that 70% of HCQtreated patients were virologically cured comparing with 12.5% in the control group at day 6 post-inclusion (P<0.001).Of all the studies included in this review, only one study 68 used HCQ for COVID-19 patients, and the founding revealed that only two mild AEs occurred in patients with a daily dose of 400mg.The dosage of HCQ in the "Shanghai protocol" for the treatment of COVID-19 infection was 400mg daily for 5 days, and the preliminary study on 30 patients with common COVID-19 infection indicated that the HCQ group had no significant increase in the incidence of AEs compared with the control group (P > 0.05), and all ADRs disappeared after the withdrawal 25 .
There several limitations to this study.Firstly, the quantitative subgroup analysis could not be conducted because of the limited number of AEs reported in a limited number of studies in the high-dose group, thus we could not demonstrate that whether the incidence of AEs or SAEs were significantly increased in high-dose group.Secondly, when we counted the number of AEs in each study, some studies reported the number of AEs, but not the number of patients with AEs.Since different types of AEs might occur simultaneously in a single patient, as a result, the pooled incidence may be overestimated.Thirdly, the unspecified AEs or SAEs in some studies were not accounted for the total number of certain AEs or SAEs, leading to an underestimation of the pooled incidence.Finally, we merely enrolled randomized studies representing the exclusion of patients at high risk of harm 122 , the lack of enough time to determine long-term harm and small sample size to detect unusual events 123 .In the forward work, we will supplement the data on the safety of HCQ by incorporating observational studies.

CONCLUSION
The short time use of HCQ was well tolerated in other diseases and COVID-19, though there is not much evidence in the treatment of COVID-19.We considered that HCQ might be safe for clinical application under the outbreak of COVID-19.We will then conduct a pooled analysis of the AEs reported in the randomized and observational studies, in the hope of finding more safety information of HCQ in high dosage.

FUNDING
This study was funded by Key Clinical Specialty of Shanghai (Clinical Pharmacy) (No.shslczdzk06504).A summary table of the judgements for each risk of bias item for each study.

Figure 1 Figure 2
Figure 1 The PRISMA 2009 Flow Diagram in literature screening Figure 2 Risk of bias graph of the included studies A plot of the distribution of the judgements across studies for each risk of bias item.

Figure 3
Figure 3 Risk of bias summary of the included studies.Green means "low risk," yellow means "unclear risk," and red means "high risk."

Table 1
Basic characteristics of the eligible studies

Table 3
Differences in the pooled incidence of the common AEs between the HCQ and control groups *P value < 0.05 and difference was statistically significant.

Table 6
Meta-analysis results of the incidence of AEs between HCQ and control group