A Meta-Analysis on the Effects of Hydroxychloroquine on COVID-19

Introduction Since December 2019, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has rapidly spread throughout the world with a large medical and economic impact. On March 12, 2020, the World Health Organization (WHO) classified SARS-CoV-2 as a pandemic. As a result of this worldwide public health crisis, politicians, elected officials, and healthcare professionals emergently began trialing hydroxychloroquine (HCQ) in efforts to treat and prevent the transmission of the virus. This meta-analysis was performed to assess the effects of HCQ on patients with COVID-19. Methods This meta-analysis adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMA) guidelines. Selected articles published between December 2019 and July 2020 were found utilizing the following search engines: PubMed, Google Scholar, Cochrane Library, DisasterLit, Clinicaltrials.gov, Medrxiv, and Embase. Two independent physician reviewers screened eligible articles that met the inclusion and exclusion criteria of the analysis. The outcome measures analyzed were mortality rate, rate of disease progression/improvement, rate of disease severity, and adverse effects of treatment. Six out of 14 studies that met the study’s eligibility criteria were selected and further analyzed, with a total of 381 participants (n= 381). Conclusion From the studies analyzed, it was found that groups treated with HCQ had an overall mortality rate that was 2.5 times greater than that of the control group. HCQ treated patients had higher rates of adverse clinical outcomes and side effects compared with the control populations. Lastly, there was a 1.2 times higher rate of improvement in the group of HCQ treated patients with mild to moderate symptoms as compared to the control group.

The SARS-CoV-2 virus has an affinity for the ciliated cells of the respiratory conducting airway, with increased viral replication as it progresses further along the respiratory tract and gastrointestinal mucosa [12]. The SARS CoV-2 infection occurs in three distinct stages: an asymptomatic stage, an upper airway stage, and, finally, the conducting airway response stage, which leads to the classically seen ground-glass infiltrates on chest X-ray and clinical hypoxia with progression to acute respiratory distress syndrome (ARDS) and multi-system dysfunction [13]. In stage 1, the virus binds to the angiotensin-converting enzyme 2 (ACE2) receptor, a transmembrane protease, serine 2 (TMPRSS2). TMPRSS2 is ubiquitous in the human body; it is found in the nasal cavity and lung and is also expressed throughout the intestine and prostate [14]. ACE2 receptors can also be found in the heart, esophagus, kidneys, stomach, bladder, and ileum [12,14]. As SARS-CoV-2 progresses down the respiratory tract, the virus begins to activate a more potent immune response and certain patients may manifest clinically with respiratory failure and ARDS.
Most patients will have a mild disease, with the disease restricted to the upper respiratory tract [13]. About one out of five SAR-CoV-2 infected patients will progress to more severe respiratory disease and further to ARDS [13]. The proposed mechanism is the destruction of type II pneumocytes once the virus reaches the alveoli [12]. The virus would then begin the replication process within these cells and the cell would undergo apoptosis, releasing viral particles. This cellular apoptosis results in diffuse alveolar damage with the formation of hyaline membranes, which decrease gas exchange and lead to clinical hypoxia. Furthermore, the healing of the affected areas may worsen the patient's condition through more severe parenchymal scarring and fibrosis. Because the cytokines mentioned above have binding sites within the lungs, they may serve as therapeutic targets.
4-Aminoquinolones such as hydroxychloroquine (HCQ) and chloroquine have gained a lot of steam in the medical field and media for their possible efficacy against COVID-19. HCQ has immunomodulatory properties and was originally developed as an antimalarial drug with further applications in patients with rheumatoid arthritis and systemic lupus erythematosus [15]. In vitro studies of HCQ have additionally shown antiviral properties; it supposedly prevents COVID-19 related ARDS [8,[15][16]. The treatment of COVID-19 positive patients with HCQ has been met with controversy, as there have been no large multicenter randomized control trials to support its use. Up to this point, there is a lack of statistically significant reduction in morbidity or mortality in COVID-19 patients who have undergone HCQ trials.
The treatment of COVID-19 with a combination of hydroxychloroquine and azithromycin was first proposed in a controversial, small non-randomized trial from the South of France that concluded that the drug combination was effective for the treatment of COVID-19 [17]. Criticism was brought on immediately when it was presented for peer review due to many methodological flaws, with the biggest being the lack of a randomized control group [17]. This led to various expert researchers criticizing the efficacy of hydroxychloroquine, with the majority concluding no statistically significant difference between treatment groups.
The emergent approval of HCQ at the height of the COVID-19 pandemic was considered controversial but necessary given the overwhelming lack of effective treatment options at that time. The controversy was limited not only to the unknown efficacy and side-effect profile of HCQ but also to the limited supply of the drug [18].

Materials And Methods
This systematic review and meta-analysis adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines ( Figure 1) [2]. The search terms used were hydroxychloroquine, chloroquine, azithromycin, COVID-19, coronavirus, and SARS-CoV-2. Using these terms, the systematic search strategies used were boolean and fuzzy logic, truncated terms, and wild cards. Selected articles published between December 2019 and July 2020 were found utilizing the following search engines: PubMed, Google Scholar, Cochrane Library, DisasterLit, Clinicaltrials.gov, Medrxiv, and Embase. Two independent physician reviewers screened eligible articles that met the analysis' inclusion and exclusion criteria. The inclusion criteria were (1) age range 12-65, (2) prospective control trial, and (3) use of hydroxychloroquine, chloroquine, lopinavir-ritonavir, or azithromycin. The exclusion criteria were (1) presence of a co-morbid medical condition, i.e., advanced heart, liver, or renal disease or diabetes mellitus, (2) treatment with remdesivir, convalescent plasma, corticosteroids, vaccines, IL-6 inhibitors, T-cell therapy, α-ketoamide inhibitors, resiniferatoxin, teicoplanin, favipiravir, extracorporeal therapy, or HCQ prophylaxis.

PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses
The outcome measures analyzed were HCQ's effect on COVID-19 mortality rate, rate of disease progression/rate of improvement, rate of disease severity, and adverse effects of treatment. Six out of 14 studies that met the study's eligibility criteria were selected and further analyzed, with a total of 381 participants (n= 381).
Data collected from six different studies looked at the effects of HCQ on patients with clinically proven COVID-19 infection. These six studies have been labeled S1, S2, S3, S4, S5, and S6 for better data visualization. Each study varies in sample sizes and the distribution of treatment and control groups. The defined outcomes studied in this meta-analysis are: 1) The mortality rate of patients after applying HCQ on patients with COVID-19 2) The rate of progression/improvement of COVID-19 disease 3) The rate of COVID-19 disease severity, for example, after applying the HCQ treatment, the rate of which patients went on to develop severe conditions such as acute hypoxic respiratory failure and adult respiratory distress syndrome.
The random-effects model was used on the assumption that the study effect estimates show more variance when drawn from a single population [19]. Therefore, this follows the so-called assumption of exchangeability [19]. This means that in a random-effects model fit, not only do assumptions of the effects of individual studies deviate from the true intervention effect of all studies due to sampling error but that there is another source of variance introduced by the fact that the studies do not stem from one single population [19]. The studies are sampled from a "universe" of populations [19]. In this study, the random effect model is a suitable choice because it is a risky assumption to state all the studies along with their respective effect sizes stem from a single homogeneous population.

Section A: meta-analysis on mortality rates
Among the six studies considered for meta-analysis, information on mortality rates was available in two of them, details of which are provided below (Tables 1-3).    These studies are perfectly homogeneous as tau^2 is 0 (equivalently, H^2 is 1). See Figures 2-3 for more information. Study 1 has more uncertainty in its results as evident due to the wide spread of the horizontal line. Studies 1 and 2 both do not cross the effect line at 0, indicating that they are not in agreement with the mortality rate of HCQ treated COVID-19 positive patients.

Interpretation
There is a marginal asymmetry between the studies, however, as the number of studies is small, this result can be attributed purely to chance rather than any actual publication bias.

Section B: meta-analysis on progression/improvement rates
Among the six studies considered for meta-analysis, information on disease progression rates are available in four of them, details of which are provided below (Tables 4-6). 2020     These studies are a bit heterogeneous, though by a very small amount. See Figures 4-5 for more information.

FIGURE 4: Disease progression forest plot
Interpretation Study 6 has more uncertainty in its results as evident by the width of the horizontal line [19]. All studies, except Study 6, are in agreement with the results of a disease progression rate of HCQ treatment in patients with COVID [19]. 2020   The studies are symmetric except Study 6, which falls outside the triangle. This is in line with the conclusion drawn from the funnel plot. However, as evidenced by the funnel plot, Study 6 has very low power and thus its effect can be ignored.

Section C: meta-analysis on severity rates
Among the six studies considered, information on disease severity rates are available in four, the details of which are provided below (Tables 7-9).    Here, the estimated average log relative risk is equal to ˆμ=-0.3644 (95% CI: -2.3236 to 1.5949) [19]. For easier interpretation, these values are transformed back to the relative risk scale through exponentiation (i.e., exp(ˆμ) = 0.6946 with 95% CI: 0.10 to 4.92). The interpretation of these results suggests that the disease severity rate in HCQ treated individuals is on average 0.69 that of the non-HCQ individuals. The null hypothesis H0: μ= 0 can be rejected (p < 0.05).
These studies exhibit heterogeneity by a moderate amount. See Figures 6-7 for more

Discussion
The side effects of 4-Aminoquinolones are known to be dose-dependent increased risks for retinopathy, methemoglobinemia, and gastrointestinal (GI), renal, and cardiac toxicity [26]. HCQ co-administered with medications such as AZT further increases the risk of toxicity, particularly prolongation of the QT interval on electrocardiogram. The Borba et al. study revealed that males aged 50 with severe COVID symptoms and heart disease are at high risk for developing HCQ-related cardiac complications such as QT prolongation at higher doses of HCQ [21]. This toxicity is especially noted when combined with AZT, which is known to prolong the QT interval in populations with cardiac disease [21]. The studies by Tang et al. [24] and Chen J et al. [20] showed greater HCQ-related GI side effects as well.
In a post-marketing study by the Food and Drug Administration (FDA), it was also shown that the use of 4-Aminoquinolones increased rates of cardiac arrhythmias, ventricular tachycardia, fibrillation, and torsades de pointes. Their analysis also noted adverse cardiac events in combination with the use of other QT-prolonging medications such as azithromycin [27]. As a result, the FDA has cautioned the use of HCQ in COVID-19 patients, especially outside of the inpatient hospital setting [27]. Similarly, this meta-analysis supports that HCQ treated patients are more likely to have adverse side effects. It also appears that treatment with HCQ has a fatality rate of approximately 2.5 higher than with the control group.
The non-randomized study performed by Gautret et al. in the South of France included a total of 36 young patients with positive PCR test results and milder COVID-19 disease with no advanced comorbid medical conditions. A 50% reduction in viral load was noted at one week with a low dose of HCQ with AZT [17]. This study was not powered to detect mortality outcomes. Similarly, Yang et al. [19], Mingxing et al. [23], and Chen J et al. [20] studied females with a median age of 45 and mild COVID-19 related upper respiratory/pneumonia symptoms, without co-existing comorbid medical disease. Patients were stated to have improved time to clinical resolution in the HCQ treatment arm [20,[22][23]. These results seem to be in line with the meta-analysis' of a slight disease improvement in COVID-19 patients treated with HCQ as compared with the controls.
Furthermore, recent studies show a gender disparity, in that females show better outcomes as compared to similar male cohorts [14]. This gender disparity is seen in a recent study that noted that male patients with advanced age or multiple comorbid medical conditions are at higher risk for mortality [11,14]. The studies in this meta-analysis did not include these high-risk patients with underlying complex co-morbid medical conditions, severe cases of COVID-19, ARDS, or critical care patient populations.

Limitations
Of note, the studies included in this meta-analysis have various definitions of control groups, which might affect the conclusion. However, with respect to the disease progression and severity meta-analysis, it appears that most of the studies are in agreement with the results, with slight exceptions, which might be attributed to chance. To get a more robust conclusion, the metaanalysis can be performed on more studies rather than six prospective control trials. Currently, there are 107 HCQ clinical trials in the active recruitment phase [28]; as the pandemic continues to unfold, these future large multicenter randomized controlled clinical trials may be included in the meta-analysis to conclude the size effect of HCQ on COVID-19.