A Systematic Review, Meta-analysis, and Trial Sequential Analysis to Inform Clinical Guidelines.
Abstract
Background:
Repurposed medicines may have a role against the SARS-CoV-2 virus. The antiparasitic ivermectin, with antiviral and anti-inflammatory properties, has now been tested in numerous clinical trials. Areas of uncertainty: We assessed the efficacy of ivermectin treatment in reducing mortality, in secondary outcomes, and in chemoprophylaxis, among people with, or at high risk of, COVID-19 infection.
Data sources:
We searched bibliographic databases up to April 25, 2021. Two review authors sifted for studies, extracted data, and assessed risk of bias. Meta-analyses were conducted and certainty of the evidence was assessed using the GRADE approach and additionally in trial sequential analyses for mortality. Twenty-four randomized controlled trials involving 3406 participants met review inclusion.
Therapeutic Advances:
Meta-analysis of 15 trials found that ivermectin reduced risk of death compared with no ivermectin (average risk ratio 0.38, 95% confidence interval 0.19–0.73; n = 2438; I2 = 49%; moderate-certainty evidence). This result was confirmed in a trial sequential analysis using the same DerSimonian–Laird method that underpinned the unadjusted analysis. This was also robust against a trial sequential analysis using the Biggerstaff–Tweedie method. Low-certainty evidence found that ivermectin prophylaxis reduced COVID-19 infection by an average 86% (95% confidence interval 79%–91%). Secondary outcomes provided less certain evidence. Low-certainty evidence suggested that there may be no benefit with ivermectin for “need for mechanical ventilation,” whereas effect estimates for “improvement” and “deterioration” clearly favored ivermectin use. Severe adverse events were rare among treatment trials and evidence of no difference was assessed as low certainty. Evidence on other secondary outcomes was very low certainty.
Conclusions:
Moderate-certainty evidence finds that large reductions in COVID-19 deaths are possible using ivermectin. Using ivermectin early in the clinical course may reduce numbers progressing to severe disease. The apparent safety and low cost suggest that ivermectin is likely to have a significant impact on the SARS-CoV-2 pandemic globally.
INTRODUCTION
To date, very few treatments have been demonstrated to reduce the burden of morbidity and mortality from COVID-19. Although corticosteroids have been proven to reduce mortality in severe disease,1 there has been little convincing evidence on interventions that may prevent disease, reduce hospitalizations, and reduce the numbers of people progressing to critical disease and death.
Ivermectin is a well-known medicine that is approved as an antiparasitic by the World Health Organization and the US Food and Drug Administration. It is widely used in low- and middle-income countries (LMICs) to treat worm infections.2,3 Also used for the treatment of scabies and lice, it is one of the World Health Organization’s Essential Medicines.4 With total doses of ivermectin distributed apparently equaling one-third of the present world population,5 ivermectin at the usual doses (0.2–0.4 mg/kg) is considered extremely safe for use in humans.6,7 In addition to its antiparasitic activity, it has been noted to have antiviral and anti-inflammatory properties, leading to an increasing list of therapeutic indications.8
Since the start of the SARS-CoV-2 pandemic, both observational and randomized studies have evaluated ivermectin as a treatment for, and as prophylaxis against, COVID-19 infection. A review by the Front Line COVID-19 Critical Care Alliance summarized findings from 27 studies on the effects of ivermectin for the prevention and treatment of COVID-19 infection, concluding that ivermectin “demonstrates a strong signal of therapeutic efficacy” against COVID-19.9 Another recent review found that ivermectin reduced deaths by 75%.10 Despite these findings, the National Institutes of Health in the United States recently stated that “there are insufficient data to recommend either for or against the use of ivermectin for the treatment of COVID-19,”11 and the World Health Organization recommends against its use outside of clinical trials.12
Ivermectin has exhibited antiviral activity against a wide range of RNA and some DNA viruses, for example, Zika, dengue, yellow fever, and others.13 Caly et al14 demonstrated specific action against SARS-CoV-2 in vitro with a suggested host-directed mechanism of action being the blocking of the nuclear import of viral proteins14,15 that suppress normal immune responses. However, the necessary cell culture EC50 may not be achievable in vivo.16 Other conjectured mechanisms include inhibition of SARS-CoV-2 3CLPro activity17,18 (a protease essential for viral replication), a variety of anti-inflammatory effects,19 and competitive binding of ivermectin with the viral S protein as shown in multiple in silico studies.20 The latter would inhibit viral binding to ACE-2 receptors suppressing infection. Hemagglutination via viral binding to sialic acid receptors on erythrocytes is a recently proposed pathologic mechanism21 that would be similarly disrupted. Both host-directed and virus-directed mechanisms have thus been proposed, the clinical mechanism may be multimodal, possibly dependent on disease stage, and a comprehensive review of mechanisms of action is warranted.
Developing new medications can take years; therefore, identifying existing drugs that can be repurposed against COVID-19 that already have an established safety profile through decades of use could play a critical role in suppressing or even ending the SARS-CoV-2 pandemic. Using repurposed medications may be especially important because it could take months, possibly years, for much of the world's population to get vaccinated, particularly among LMIC populations.
Currently, ivermectin is commercially available and affordable in many countries globally.6 A 2018 application for ivermectin use for scabies gives a direct cost of $2.90 for 100 12-mg tablets.22 A recent estimate from Bangladesh23 reports a cost of US$0.60—US$1.80 for a 5-day course of ivermectin. For these reasons, the exploration of ivermectin's potential effectiveness against SARS-CoV-2 may be of particular importance for settings with limited resources.24 If demonstrated to be effective as a treatment for COVID-19, the cost-effectiveness of ivermectin should be considered against existing treatments and prophylaxes.
The aim of this review was to assess the efficacy of ivermectin treatment among people with COVID-19 infection and as a prophylaxis among people at higher risk of COVID-19 infection. In addition, we aimed to prepare a brief economic commentary (BEC) of ivermectin as treatment and as prophylaxis for COVID-19.25
METHODS
The conduct of this review was guided by a protocol that was initially written using Cochrane's rapid review template and subsequently expanded to a full protocol for a comprehensive review.26
Search strategy and selection criteria
Two reviewers independently searched the electronic databases of Medline, Embase, CENTRAL, Cochrane COVID-19 Study Register, and Chinese databases for randomized controlled trials (RCTs) up to April 25, 2021 (see Appendix 1–3, Supplemental digital content 1, http://links.lww.com/AJT/A95); current guidance25 for the BEC was followed for a supplementary search of economic evaluations. There were no language restrictions, and translations were planned to be performed when necessary.
We searched the reference list of included studies, and of two other 2021 literature reviews on ivermectin,9 as well as the recent WHO report, which included analyses of ivermectin.12 We contacted experts in the field (Drs. Andrew Hill, Pierre Kory, and Paul Marik) for information on new and emerging trial data. In addition, all trials registered on clinical trial registries were checked, and trialists of 39 ongoing trials or unclassified studies were contacted to request information on trial status and data where available. Many preprint publications and unpublished articles were identified from the preprint servers MedRχiv and Research Square, and the International Clinical Trials Registry Platform. This is a rapidly expanding evidence base, so the number of trials are increasing quickly. Reasons for exclusion were recorded for all studies excluded after full-text review.
Data analysis
We extracted information or data on study design (including methods, location, sites, funding, study author declaration of interests, and inclusion/exclusion criteria), setting, participant characteristics (disease severity, age, gender, comorbidities, smoking, and occupational risk), and intervention and comparator characteristics (dose and frequency of ivermectin/comparator). The primary outcome for the intervention component of the review included death from any cause and presence of COVID-19 infection (as defined by investigators) for ivermectin prophylaxis. Secondary outcomes included time to polymerase chain reaction (PCR) negativity, clinical recovery, length of hospital stay, admission to hospital (for outpatient treatment), admission to ICU or requiring mechanical ventilation, duration of mechanical ventilation, and severe or serious adverse events, as well as post hoc assessments of improvement and deterioration. All of these data were extracted as measured and reported by investigators. Numerical data for outcomes of interest were extracted according to intention to treat.
If there was a conflict between data reported across multiple sources for a single study (eg, between a published article and a trial registry record), we contacted the authors for clarification. Assessments were conducted by 2 reviewers (T.L., T.D., A.B., or G.G.) using the Cochrane RCT risk-of-bias tool.27 Discrepancies were resolved by discussion.
Continuous outcomes were measured as the mean difference and 95% confidence intervalss (CI), and dichotomous outcomes as risk ratio (RR) and 95% CI.
We did not impute missing data for any of the outcomes. Authors were contacted for missing outcome data and for clarification on study methods, where possible, and for trial status for ongoing trials.
We assessed heterogeneity between studies by visual inspection of forest plots, by estimation of the I2 statistic (I2 ≥60% was considered substantial heterogeneity), by a formal statistical test to indicate statistically significant heterogeneity, and, where possible, by subgroup analyses (see below). If there was evidence of substantial heterogeneity, the possible reasons for this were investigated and reported. We assessed reporting biases using funnel plots if more than 10 studies contributed to a meta-analysis.
We meta-analyzed data using the random effects model (DerSimonian and Laird method)30 using RevMan 5.4.1 software.27,31 The results used the inverse variance method for weighting.27 Some sensitivity analyses used other methods that are outlined below and some calculations were performed in R32 through an interface33 to the netmeta package. Where possible, we performed subgroup analyses grouping trials by disease severity, inpatients versus outpatients, and single dose versus multiple doses. We performed sensitivity analyses by excluding studies at high risk of bias. We conducted further post hoc sensitivity analyses using alternative methods to test the robustness of results in the presence of zero events in both arms in a number of trials35 and estimated odds ratios [and additionally RR for the Mantel–Haenszel (MH) method] using a fixed effects model. The models incorporate evidence from single-zero studies without having to resort to continuity corrections. However, double-zero studies are excluded from the analysis; so, the risk difference was also assessed using the MH method as this approach can adequately incorporate trials with double-zero events. This method can also use a random-effects component. A “treatment-arm” continuity correction was used, where the values 0.01, 0.1, and 0.25 were added where trials reported zero events in both arms. It has been shown that a nonfixed continuity correction is preferable to the usual 0.5.35 Other methods are available but were not considered due to difficulty in interpretation, sensitivity of assumptions, or the fact they are rarely used in practice.36–40
Trial sequential analysis
When a meta-analysis is subjected to repeated statistical evaluation, there is an exaggerated risk that “naive” point estimates and confidence intervals will yield spurious inferences. In a meta-analysis, it is important to minimize the risk of making a false-positive or false-negative conclusion. There is a trade-off between the risk of observing a false-positive result (type I error) and the risk of observing a false-negative result (type II error). Conventional meta-analysis methods (eg, in RevMan) also do not take into account the amount of available evidence. Therefore, we examined the reliability and conclusiveness of the available evidence using trial sequential analyses (TSA).The DerSimonian–Laird (DL) method was used because this is most often used in meta-analytic practice and was also used in the primary meta-analysis.30
The TSA was used to calculate the required information size (IS) to demonstrate or reject a relative reduction in the risk (RRR) of death in the ivermectin group, as found in the primary meta-analysis. We assumed the estimated event proportion in the control group from the meta-analysis because this is the best and most representative available estimate. Recommended type I and II error rates of 5% and 10% were used, respectively (power of 90%), powering the result on the effect observed in the primary meta-analyses. We did not identify any large COVID-19 trials powered on all-cause mortality, so powering on some external meaningful difference was not possible. Any small RRR is meaningful in this context, given the scale of the pandemic, but the required IS would be unfeasibly high for this analysis if powered on a small difference. The only reliable data on ivermectin in its repurposed role for treatment against COVID-19 will be from the primary meta-analysis. Therefore, assuming it does not widely deviate from other published systematic reviews, a pragmatic decision was therefore made to power on the pooled meta-analysis effect estimate for all-cause mortality a priori. This is more reflective of a true meaningful difference. We used a model variance-based estimate to correct for heterogeneity. A continuity correction of 0.01 was used in trials that reported zero events in one or both arms. The required IS is the sample size required for a reliable and conclusive meta-analysis and is at least as large as that needed in a single powered RCT. The heterogeneity corrected required IS was used to construct sequential monitoring boundaries based on the O'Brien–Fleming type alpha-spending function for the cumulative z-scores (corresponding to the cumulative meta-analysis), analogous to interim monitoring in an RCT, to determine when sufficient evidence had been accrued. These monitoring boundaries are relatively insensitive to the number of repeated significance tests. They can be used to further contextualize the original meta-analysis and enhance our certainty around its conclusions. We used a two-sided test, so also considered futility boundaries (to test for no statistically significant difference) and the possibility that ivermectin could harm. Sensitivity analyses were performed excluding the trial of Fonseca, which was a cause of substantial heterogeneity (but retained in the core analysis because it was at low risk of bias). Its removal dramatically reduced I2 and D2 (diversity) estimates, thus reducing the model variance-based estimate to correct for heterogeneity. Two further sensitivity analyses were performed using 2 alternative random effect models, namely the Biggerstaff–Tweedie (BT) and Sidik–Jonkman (SJ) methods.
All outcomes have been assessed independently by 2 review authors (T.D. and A.B.) using the GRADE approach, which ranks the quality and certainty of the evidence. The results of the TSAs will also form part of the judgment for the primary all-cause mortality outcome. The results are presented in a summary of findings table. Any differences in judgments were resolved by discussion with the wider group. We used Cochrane Effective Practice and Organisation of Care guidance to interpret the evidence.