Meta-analysis of Randomized Trials of Ivermectin to Treat SARS-CoV-2 Infection.

Ivermectin is an antiparasitic drug being investigated for repurposing against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Ivermectin showed in vitro activity against SARS-COV-2, but only at high concentrations. This meta-analysis investigated ivermectin in 23 randomized clinical trials (3349 patients) identified through systematic searches of PUBMED, EMBASE, MedRxiv, and trial registries. The primary meta-analysis was carried out by excluding studies at a high risk of bias. Ivermectin did not show a statistically significant effect on survival (risk ratio [RR], 0.90; 95% CI, 0.57 to 1.42; P = .66) or hospitalizations (RR, 0.63; 95% CI, 0.36 to 1.11; P = .11). Ivermectin displayed a borderline significant effect on duration of hospitalization in comparison with standard of care (mean difference, -1.14 days; 95% CI, -2.27 to -0.00; P = .05). There was no significant effect of ivermectin on time to clinical recovery (mean difference, -0.57 days; 95% CI, -1.31 to 0.17; P = .13) or binary clinical recovery (RR, 1.19; 95% CI, 0.94 to 1.50; P = .15). Currently, the World Health Organization recommends the use of ivermectin only inside clinical trials. A network of large clinical trials is in progress to validate the results seen to date.


Introduction
The SARS-CoV-2 pandemic continues to grow, with over 350,000 new infections and over 7,000 deaths recorded worldwide daily in May 2021 [1]. Protective vaccines have been developed, but current supplies are too low to cover worldwide demand in the coming months [2]. Researchers worldwide are urgently looking for interventions to prevent new infections, or prevent disease progression, and lessen disease severity for those already infected.
While research on new therapeutic agents for COVID-19 is key, there is also great interest in evaluating the potential of already existing medicines against COVID-19, and many clinical trials are in progress to ‗re-purpose' drugs normally indicated for other diseases. The known safety profiles, shortened development timelines, and well-established markets (with low price points and higher capacity to deliver at scale) for most of the already existing compounds proposed for COVID-19 are particularly advantageous compared to new drug discovery in a pandemic situation. Three re-purposed anti-inflammatory drugs have shown significant survival benefits to date: the corticosteroid dexamethasone in the UK RECOVERY trial [3], and the Interleukin-6 (IL-6) receptor antagonist drugs, tocilizumab and sarilumab, in the REMAP-CAP trial and RECOVERY trial [4,5]. Other re-purposed antimicrobials such as, hydroxychloroquine, lopinavir/ritonavir, remdesivir and interferonbeta, have shown no significant survival benefit in two large, randomized trials [3,6] despite initial reports of efficacy, underscoring the need for caution when interpreting early clinical trial data.
Dexamethasone is recommended for use by the WHO and has proven survival benefits for oxygen-dependent patients with COVID-19, while tocilizumab and sarilumab improve survival for patients in intensive care [3,4]. Preliminary data suggest that nitazoxanide and budesonide may have a role in mild infection [7,8]. However, there are no approved treatments for patients with mild SARS-CoV-2 infection, either to prevent disease progression or reduce viral transmission. Treatments increasing viral clearance rate may reduce the risk of onward transmission but this requires empirical demonstration.
Downloaded from https://academic.oup.com/ofid/advance-article/doi/10.1093/ofid/ofab358/6316214 by guest on 08 July 2021 A c c e p t e d M a n u s c r i p t 3 Ivermectin is a well-established anti-parasitic drug used worldwide for a broad number of parasites and also for topical use against rosacea. Antiviral activity of ivermectin has been demonstrated recently for SARS-CoV-2 in Vero/hSLAM cells [9]. However, concentrations required to inhibit viral replication in-vitro (EC 50 =2.2 -2.8μM; EC 90 =4.4μM) are not achieved systemically after oral administration of the drug to humans [9,10].
The drug is estimated to accumulate in lung tissues (2.67 times that of plasma) [11], but this is also unlikely to be sufficient to maintain target concentrations for pulmonary antiviral activity [10,12]. Notwithstanding, ivermectin is usually present as a mixture of two agents and although mainly excreted unchanged in humans, has two major metabolites [13].
Current data are insufficient to determine whether the minor form or a circulating metabolite has higher direct potency against SARS-CoV-2, but it seems likely that it would need to be profoundly more potent than the reported values.
Ivermectin has also demonstrated immunomodulatory and anti-inflammatory mechanisms of action in preclinical models of several other indications. In-vitro studies have demonstrated that ivermectin suppresses production of the inflammatory mediators nitric oxide and prostaglandin E2 [14]. Furthermore, avermectin (from which ivermectin is derived) significantly impairs pro-inflammatory cytokine secretion (IL-1β and TNF-α) and increases secretion of the immunoregulatory cytokine IL-10 [15]. Ivermectin also reduced TNF-α, IL-1, and IL-6, and improved survival in mice given a lethal dose of lipopolysaccharide [16].
Preclinical evidence to support these immunomodulatory and anti-inflammatory mechanisms of action have also been generated in murine models [17,18]. Finally, in Syrian golden hamsters infected with SARS-CoV-2, subcutaneous ivermectin demonstrated a reduction in the IL-6/IL-10 ratio in lung tissues. In this study, ivermectin also prevented pathological deterioration [19]. Ultimately, various potential mechanisms of action for ivermectin against COVID19 exist and are undergoing further investigation, as recently summarised in a review article [20].
At standard doses, of 0.2-0.4mg/kg for 1-2 days, ivermectin has a good safety profile and has been distributed to billions of patients worldwide in mass drug administration programs.
A recent meta-analysis found no significant difference in adverse events in those given Downloaded from https://academic.oup.com/ofid/advance-article/doi/10.1093/ofid/ofab358/6316214 by guest on 08 July 2021 A c c e p t e d M a n u s c r i p t 4 higher doses of ivermectin, of up to 2mg/kg, and those receiving longer courses, of up to 4 days, compared to those receiving standard doses [21]. Ivermectin is not licensed for pregnant or breast-feeding women, or children <15kg. The WHO Guidelines Group found that in 16 RCTs with 2407 participants ivermectin improved mortality outcomes compared with control but rated the quality of available evidence as low or very low [22]. Currently, the WHO does not recommend the use of ivermectin outside clinical trials.
The objective of this systematic review and meta-analysis was to combine available results from new published or unpublished randomized trials of ivermectin in SARS-CoV-2 infection to inform current guidelines.

Methods
The systematic review and meta-analysis was conducted according to PRISMA guidelines.
A systematic search of PUBMED and EMBASE was conducted to identify randomized control trials (RCTs) evaluating treatment with ivermectin for SARS-CoV-2 infected patients. Clinical trials with no control arm, or those evaluating prevention of infection were excluded alongside non-randomized trials and case-control studies. Key data extracted included baseline characteristics (age, sex, weight, oxygen saturation, stage of infection), changes in inflammatory markers, viral suppression after treatment, clinical recovery, hospitalization and survival. Data were extracted and cross-checked by two independent reviewers (HW and LE).  [25] to identify additional trials listed on other national, and international registries. Literature searches via PubMed, Embase, and the preprint servers MedRxiv and Researchsquare were conducted to identify published studies. Duplicate registrations, non-randomised studies and prevention studies were excluded following discussion between the authors.

Search strategy and selection criteria
Additionally, the research teams conducting unpublished clinical trials were contacted and requested to join regular international team meetings from December 2020 to May 2021. All results available from eligible unpublished studies were also included in this systematic review.
All of the clinical trials included in this meta-analysis were approved by local ethics committees and all patients gave informed consent.
The primary outcome was all-cause mortality from randomization to the end of follow-up.

Data analysis
Statistical analyses for all-cause mortality, time to viral clearance and clinical recovery were conducted using published data summaries. For the mortality outcome, clinical trials with at least one death reported were included in this analysis. Furthermore, any hospitalization within 12 hours of randomization was excluded. Treatment effects were expressed as risk ratios (RR) for binary outcomes and mean difference (MD) for continuous outcomes. For each outcome, we pooled the individual trial statistics using the random-effects inversevariance model; a continuity correction of 0.5 was applied to treatment arms with no deaths.
Heterogeneity was evaluated by I 2 . The significance threshold was set at 5% (two-sided) and all analyses were conducted using Revman 5.3. A funnel plot for the mortality outcome was Downloaded from https://academic.oup.com/ofid/advance-article/doi/10.1093/ofid/ofab358/6316214 by guest on 08 July 2021 A c c e p t e d M a n u s c r i p t 6 created to assess publication bias and small study effects; the p-value was estimated from the regression-based Harbord test for small study effects.
All studies included in this analysis were assessed for risk of bias using the Cochrane Collaboration risk of bias standardized assessment tool [26]. The outcome of this assessment is given in Supplementary Table 3. Each study was assessed for risk of bias for the primary endpoint, viral load, and survival outcomes. The primary endpoint in the trials tended to be clinical recovery which is more subjective and likely to be influenced by knowledge of treatment arms. An assessment was also carried out on more objective endpoints including survival and viral load which are less likely to be influenced by this bias.
Where information was not available in published papers, clinical trial investigators were proactively contacted to inform the risk bias analysis.

24
RCTs involving a total of 3328 participants were included in this meta-analysis. The sample sizes of each trial ranged from 24 to 400 participants. Of the 24 included studies, eight were published papers, nine were available as pre-prints, six were unpublished results shared for this analysis, and one reported results via a trial registry website.

Evaluation of Studies.
An evaluation of the quality of the studies included in this meta-analysis was conducted according to the Cochrane Collaboration tool to assess the risk of bias across the following outcomes: primary endpoints, viral load, and survival. For the primary outcome assessment, 6

Effects on Inflammatory Markers
Five trials provided results of the effect of ivermectin on inflammatory markers including Creactive protein (CRP), ferritin and d-dimer ( Table 2). Four of these trials demonstrated significant reductions in CRP compared to control. Furthermore, in the Elgazzar trial [36], ivermectin significantly reduced ferritin levels compared to control in the severe patient population while no significant difference was demonstrated in the mild/moderate population.
The Okumus trial [47] showed significantly greater reductions in ferritin on day 10 of followup for ivermectin versus control. The Chaccour [35] and Ahmed [46] trials showed no significant difference in ferritin count between ivermectin and control. Elgazzar [36] showed significant differences in d-dimer between ivermectin and control in both the mild/moderate and severe populations. Okumus [47] showed significant differences in d-dimer on day 5 whilst Chaccour [35] found no significant differences in d-dimer between ivermectin and control, but with a smaller sample size.

Effects on Viral Clearance
Three different endpoints were used to analyze viral clearance: the percentage of patients undetectable on a set day (Table 3A), the number of days from randomization to negativity (Table 3B), and other measures such as cycle time (Ct) values and dose-response correlations (Table 3C). The Kirti [43] and Okumus [47] trials included viral load analysis only in a subset of patients. The effects of ivermectin on viral clearance were generally smaller when dosed on only one day. Several studies showed no statistically significant effect of ivermectin on viral clearance [28,29,34].
The three studies randomizing patients to different doses or durations of ivermectin showed apparent dose-dependent effects on viral clearance. First, in the Babalola trial (n=60) [48], the 0.4mg/kg dose showed trends for faster viral clearance than the 0.2mg/kg dose. Second, in the Mohan trial (n=125) [28], the 0.4 mg/kg dose of ivermectin led to a numerically higher  Figure 1B].

Effects on Clinical Recovery and Duration of Hospitalization
Definitions of clinical recovery varied across trials, as shown in Table 4. In Table 4A, three of the six trials showed significantly faster time to clinical recovery on ivermectin compared to control. In four trials, ivermectin showed significantly shorter duration of hospitalization compared to control (Table 4B).
In a meta-analysis of clinical recovery with subgroups of dose duration, there were significant differences in time to clinical recovery in favour of ivermectin (Mean Difference -1.58 days [95%CI -2.80, -0.35]; p=0.01, Figure 1C]. Additionally, ivermectin showed a 29% improvement in clinical recovery in an analysis with subgroups of dose duration (RR 1.29 [95%CI 1.12-1.47]; p=0.0003, Figure 1D]. Ivermectin demonstrated a shorter duration of hospitalization compared to control (Mean Difference -4.27 days [95%CI -8.60-0.06]; p=0.05, Figure 1E). Ivermectin was not associated with a lower risk of hospitalization compared to control (RR 0.40 [95%CI 0.14-1.08]; p=0.07, Figure 1F). However, this analysis involved only four trials in 704 participants. In a sensitivity Additional subgroup analysis of the mortality outcome with trials separated by dose-duration, blinding and control group showed consistent survival benefit and no significant subgroup differences were found (Supplementary Figures 6, 7 and 8).
A leave-one-out sensitivity analysis was performed and no single study had a substantial effect on the overall effect size (Supplementary Table 4

Mechanism of action
At the time of writing, knowledge gaps prevent a robust conclusion about the mechanism of action of ivermectin. Ivermectin's broad-spectrum anti-viral effects have been proposed to be related to its impact on the NF-κB pathway and via binding to the host cell importin α/β1 heterodimer, nuclear transport proteins responsible for nuclear entry of cargoes, and these effects in turn also prevent viral replication.

Limitations
A key limitation to this meta-analysis is the comparability of the data, with studies differing in dosage, treatment duration, and inclusion criteria. Furthermore, the standard of care used in the control arm differed between trials. In this meta-analysis, trials that used active controls such as hydroxychloroquine or lopinavir/ritonavir were combined together with those that used placebo or standard care. However, lopinavir/ritonavir and hydroxychloroquine have shown no overall benefit or harm in large randomized trials and meta-analyses. [7,[59][60][61] Furthermore, additional analyses in this paper separating trials by subgroups of standard care/ placebo and active control showed no significant difference between groups.
Another limitation is that ivermectin was given in combination with doxycycline in three trials. For open label studies, there is a risk of bias in the evaluation of subjective endpoints such as clinical recovery and hospital discharge. However, the risk is lower for objective endpoints such as viral clearance and survival. We have attempted to control for publication bias by contacting each research team conducting the trials directly. This has generated more results than would be apparent from a survey of published clinical trials only but means that many of the included trials have not been peer-reviewed. Review and publication of RCTs generally takes three to six months. It has become common practice for clinical trials of key COVID-19 treatments to be evaluated from pre-prints, such as for the WHO SOLIDARITY, RECOVERY and REMAP-CAP trials [4,5,7].

Acknowledgements
We would like to thank all the clinical staff, the research teams and the patients who participated in these studies.               A c c e p t e d M a n u s c r i p t   A c c e p t e d M a n u s c r i p t 32