Top
Introduction
Preclinical
Results
RCTs
Exclusions
Heterogeneity
Discussion
Merck, FDA, NIH, ..
Perspective
Conclusion
 
Responses
Study Notes
Revisions
Methods and Data
Supplementary
References
 
All studies
Mortality
Ventilation
ICU admission
Hospitalization
Recovery
COVID‑19 cases
Viral clearance
Peer reviewed
Exclusions
All RCTs
RCT mortality
RCT cases
RCT viral
 
Feedback
Home
c19ivm.org COVID-19 treatment researchIvermectinIvermectin (more..)
Melatonin Meta
Metformin Meta
Azvudine Meta
Bromhexine Meta Molnupiravir Meta
Budesonide Meta
Colchicine Meta
Conv. Plasma Meta Nigella Sativa Meta
Curcumin Meta Nitazoxanide Meta
Famotidine Meta Paxlovid Meta
Favipiravir Meta Quercetin Meta
Fluvoxamine Meta Remdesivir Meta
Hydroxychlor.. Meta Thermotherapy Meta
Ivermectin Meta

Loading...
More

Ivermectin for COVID-19: real-time meta analysis of 101 studies

 
0 0.5 1 1.5+ All studies 61% 101 142,247 Improvement, Studies, Patients Relative Risk Primary outcome 51% 101 142,290 Mortality 49% 51 122,827 Ventilation 29% 19 37,107 ICU admission 35% 14 28,039 Hospitalization 34% 29 44,784 Recovery 39% 37 12,829 Cases 81% 16 13,696 Viral clearance 42% 30 4,157 RCTs 53% 48 16,847 Peer-reviewed 61% 84 126,969 Prophylaxis 85% 17 19,764 Early 62% 38 58,744 Late 39% 46 63,739 Ivermectin for COVID-19 c19ivm.org March 2024 after exclusions Favorsivermectin Favorscontrol
Abstract
Statistically significant lower risk is seen for mortality, ventilation, ICU admission, hospitalization, recovery, cases, and viral clearance. All remain significant for higher quality studies. 61 studies from 55 independent teams in 25 different countries show statistically significant improvements.
Meta analysis using the most serious outcome shows 62% [51‑70%] and 85% [77‑90%] lower risk for early treatment and prophylaxis, with similar results for higher quality studies, primary outcomes, peer-reviewed studies, and for RCTs.
0 0.5 1 1.5+ All studies 61% 101 142,247 Improvement, Studies, Patients Relative Risk Primary outcome 51% 101 142,290 Mortality 49% 51 122,827 Ventilation 29% 19 37,107 ICU admission 35% 14 28,039 Hospitalization 34% 29 44,784 Recovery 39% 37 12,829 Cases 81% 16 13,696 Viral clearance 42% 30 4,157 RCTs 53% 48 16,847 Peer-reviewed 61% 84 126,969 Prophylaxis 85% 17 19,764 Early 62% 38 58,744 Late 39% 46 63,739 Ivermectin for COVID-19 c19ivm.org March 2024 after exclusions Favorsivermectin Favorscontrol
Results are very robust — in worst case exclusion sensitivity analysis 61 of 101 studies must be excluded to avoid finding statistically significant efficacy.
No treatment or intervention is 100% effective. All practical, effective, and safe means should be used based on risk/benefit analysis. Multiple treatments are typically used in combination, which may be significantly more effective. Pharmacokinetics show significant inter-individual variability Guzzo. Efficacy may vary depending on the manufacturer Williams.
Over 20 countries adopted ivermectin for COVID-19. The evidence base is much larger and has much lower conflict of interest than typically used to approve drugs.
All data to reproduce this paper and sources are in the appendix. Multiple other meta analyses show efficacy Bryant, Hariyanto, Kory, Lawrie, Nardelli.
Evolution of COVID-19 clinical evidence Ivermectin p<0.0000000001 Acetaminophen p=0.00000029 2020 2021 2022 2023 2024 Effective Harmful c19ivm.org March 2024 meta analysis results (pooled effects) 100% 50% 0% -50%
Highlights
Ivermectin reduces risk for COVID-19 with very high confidence for mortality, ventilation, hospitalization, progression, recovery, cases, viral clearance, and in pooled analysis, and high confidence for ICU admission.
Ivermectin was the 4th treatment shown effective with ≥3 clinical studies in August 2020, now with p < 0.00000000001 from 101 studies, and recognized in 22 countries.
We show traditional outcome specific analyses and combined evidence from all studies, incorporating treatment delay, a primary confounding factor in COVID-19 studies.
Real-time updates and corrections, transparent analysis with all results in the same format, consistent protocol for 66 treatments.
A
Loading..
Loading..
B
Loading..
Figure 1. A. Random effects meta-analysis excluding late treatment. This plot shows pooled effects, analysis for individual outcomes is below, and more details on pooled effects can be found in the heterogeneity section. Effect extraction is pre-specified, using the most serious outcome reported. Simplified dosages are shown for comparison, these are the total dose in the first four days for treatment, or the monthly dose for prophylaxis, for a 70kg person. For details of effect extraction and full dosage information see the appendix. B. Timeline of results in ivermectin studies. The marked dates indicate the time when efficacy was known with a statistically significant improvement of ≥10% from ≥3 studies for pooled outcomes, one or more specific outcome, pooled outcomes in RCTs, and one or more specific outcome in RCTs. Efficacy based on RCTs only was delayed by 3.6 months, compared to using all studies. Efficacy based on specific outcomes was delayed by 1.3 months, compared to using pooled outcomes.
Introduction
SARS-CoV-2 infection primarily begins in the upper respiratory tract and may progress to the lower respiratory tract, other tissues, and the nervous and cardiovascular systems, which may lead to cytokine storm, pneumonia, ARDS, neurological issues Duloquin, Hampshire, Scardua-Silva, Yang, cardiovascular complications Eberhardt, organ failure, and death. Minimizing replication as early as possible is recommended.
SARS-CoV-2 infection and replication involves the complex interplay of 50+ host and viral proteins and other factors Note A, Malone, Murigneux, Lv, Lui, Niarakis, providing many therapeutic targets for which many existing compounds have known activity. Scientists have predicted that over 7,000 compounds may reduce COVID-19 risk c19early.org, either by directly minimizing infection or replication, by supporting immune system function, or by minimizing secondary complications.
Ivermectin, better known for antiparasitic activity, is a broad spectrum antiviral with activity against many viruses including H7N7 Götz, Dengue Tay, Wagstaff, HIV-1 Wagstaff, Simian virus 40 Wagstaff (B), Zika Barrows, Yang (B), West Nile Yang (B), Yellow Fever Mastrangelo, Varghese, Japanese encephalitis Mastrangelo, Chikungunya Varghese, Semliki Forest virus Varghese, Human papillomavirus Li, Epstein-Barr Li, BK Polyomavirus Bennett, and Sindbis virus Varghese.
Ivermectin inhibits importin-α/β-dependent nuclear import of viral proteins Götz, Kosyna, Wagstaff, Wagstaff (B), inhibits SARS-CoV-2 3CLpro Mody, shows spike-ACE2 disruption at 1nM with microfluidic diffusional sizing Fauquet, binds to glycan sites on the SARS-CoV-2 spike protein preventing interaction with blood and epithelial cells and inhibiting hemagglutination Boschi, Scheim, exhibits dose-dependent inhibition of lung injury Abd-Elmawla, Ma, may inhibit SARS-CoV-2 induced formation of fibrin clots resistant to degradation Vottero, may inhibit SARS-CoV-2 RdRp activity Parvez, may be beneficial for COVID-19 ARDS by blocking GSDMD and NET formation Liu, shows protection against inflammation, cytokine storm, and mortality in an LPS mouse model sharing key pathological features of severe COVID-19 DiNicolantonio, Zhang, may be beneficial in severe COVID-19 by binding IGF1 to inhibit the promotion of inflammation, fibrosis, and cell proliferation that leads to lung damage Zhao, may minimize SARS-CoV-2 induced cardiac damage Liu (B), Liu (C), increases Bifidobacterium which plays a key role in the immune system Hazan, has immunomodulatory Munson and anti-inflammatory DiNicolantonio (B), Yan properties, and has an extensive and very positive safety profile Descotes.
We analyze all significant controlled studies of ivermectin for COVID-19. Search methods, inclusion criteria, effect extraction criteria (more serious outcomes have priority), all individual study data, PRISMA answers, and statistical methods are detailed in Appendix 1. We present random effects meta-analysis results for all studies, studies within each treatment stage, mortality, ventilation, ICU admission, hospitalization, recovery, cases, viral clearance, peer-reviewed studies, Randomized Controlled Trials (RCTs), and after exclusion of lower quality studies.
Figure 2 shows stages of possible treatment for COVID-19. Prophylaxis refers to regularly taking medication before becoming sick, in order to prevent or minimize infection. Early Treatment refers to treatment immediately or soon after symptoms appear, while Late Treatment refers to more delayed treatment.
Figure 2. Treatment stages.
Preclinical Research
Ivermectin, better known for antiparasitic activity, is a broad spectrum antiviral with activity against many viruses including H7N7 Götz, Dengue Tay, Wagstaff, HIV-1 Wagstaff, Simian virus 40 Wagstaff (B), Zika Barrows, Yang (B), West Nile Yang (B), Yellow Fever Mastrangelo, Varghese, Japanese encephalitis Mastrangelo, Chikungunya Varghese, Semliki Forest virus Varghese, Human papillomavirus Li, Epstein-Barr Li, BK Polyomavirus Bennett, and Sindbis virus Varghese.
Ivermectin inhibits importin-α/β-dependent nuclear import of viral proteins Götz, Kosyna, Wagstaff, Wagstaff (B), inhibits SARS-CoV-2 3CLpro Mody, shows spike-ACE2 disruption at 1nM with microfluidic diffusional sizing Fauquet, binds to glycan sites on the SARS-CoV-2 spike protein preventing interaction with blood and epithelial cells and inhibiting hemagglutination Boschi, Scheim, exhibits dose-dependent inhibition of lung injury Abd-Elmawla, Ma, may inhibit SARS-CoV-2 induced formation of fibrin clots resistant to degradation Vottero, may inhibit SARS-CoV-2 RdRp activity Parvez, may be beneficial for COVID-19 ARDS by blocking GSDMD and NET formation Liu, shows protection against inflammation, cytokine storm, and mortality in an LPS mouse model sharing key pathological features of severe COVID-19 DiNicolantonio, Zhang, may be beneficial in severe COVID-19 by binding IGF1 to inhibit the promotion of inflammation, fibrosis, and cell proliferation that leads to lung damage Zhao, and may minimize SARS-CoV-2 induced cardiac damage Liu (B), Liu (C).
13 In Vivo animal studies support the efficacy of ivermectin Abd-Elmawla, Albariqi, Arévalo, Chaccour, de Melo, Errecalde, Ma, Madrid, Mountain Valley MD, Uematsu, Yan, Zhang, Zheng.
7 studies investigate novel formulations of ivermectin that may be more effective for COVID-19 Albariqi, Albariqi (B), Chaccour, Errecalde, Mansour, Mohammed, Saha (B).
Preclinical research is an important part of the development of treatments, however results may be very different in clinical trials. Preclinical results are not used in this paper.
Results
Table 1 summarizes the results of random-effects meta analysis for all stages combined, for Randomized Controlled Trials, for peer-reviewed studies, with different exclusions, and for specific outcomes. Table 2 shows results by treatment stage. Figure 3 plots individual results by treatment stage. Figure 4, 5, 6, 7, 8, 9, 10, and 11 show forest plots for random effects meta-analysis of all studies with pooled effects, and for specific outcomes: mortality, ICU admission, mechanical ventilation, hospitalization, recovery, cases, and viral clearance. Figure 12 shows results for peer reviewed trials only, and the supplementary data contains peer reviewed and individual outcome results after exclusions.
Table 1. Random effects meta-analysis for all stages combined, for Randomized Controlled Trials, for peer-reviewed studies, with different exclusions, and for specific outcomes. Results show the percentage improvement with treatment and the 95% confidence interval. * p<0.05  ** p<0.01  *** p<0.001  **** p<0.0001.
Improvement Studies Patients Authors
All studies61% [53‑67%] p < 0.0001
****
101 142,247 1,129
After exclusions65% [58‑72%] p < 0.0001
****
68 124,855 766
Peer-reviewed studiesPeer-reviewed61% [51‑68%] p < 0.0001
****
84 126,969 977
Randomized Controlled TrialsRCTs53% [37‑65%] p < 0.0001
****
48 16,847 691
RCTs after exclusionsRCTs w/exc.60% [45‑71%] p < 0.0001
****
37 12,392 468
Mortality49% [35‑60%] p < 0.0001
****
51 122,827 585
VentilationVent.29% [12‑42%] p = 0.0014
**
19 37,107 281
ICU admissionICU35% [7‑54%] p = 0.019
*
14 28,039 207
HospitalizationHosp.34% [20‑45%] p < 0.0001
****
29 44,784 405
Recovery39% [29‑48%] p < 0.0001
****
37 12,829 439
Cases81% [71‑87%] p < 0.0001
****
16 13,696 144
Viral42% [29‑53%] p < 0.0001
****
30 4,157 443
RCT mortality29% [5‑48%] p = 0.022
*
18 7,239 301
RCT casesRCT cases87% [56‑96%] p = 0.00092
***
4 2,173 39
RCT viral18% [4‑30%] p = 0.014
*
22 3,058 385
Table 2. Random effects meta-analysis results by treatment stage. Results show the percentage improvement with treatment, the 95% confidence interval, and the number of studies for the stage.treatment and the 95% confidence interval. * p<0.05  ** p<0.01  *** p<0.001  **** p<0.0001.
Early treatment Late treatment Prophylaxis
All studies62% [51‑70%]
****
39% [25‑51%]
****
85% [77‑90%]
****
After exclusions69% [61‑76%]
****
46% [30‑59%]
****
84% [73‑91%]
****
Peer-reviewed studiesPeer-reviewed61% [49‑70%]
****
43% [26‑57%]
****
83% [73‑90%]
****
Randomized Controlled TrialsRCTs57% [40‑69%]
****
27% [10‑40%]
**
89% [57‑97%]
**
RCTs after exclusionsRCTs w/exc.66% [54‑75%]
****
31% [11‑46%]
**
89% [57‑97%]
**
Mortality40% [12‑59%]
**
44% [27‑57%]
****
90% [50‑98%]
**
VentilationVent.19% [-16‑44%]45% [24‑60%]
***
ICU admissionICU52% [3‑76%]
*
26% [-3‑46%]
HospitalizationHosp.53% [27‑69%]
***
16% [3‑28%]
*
67% [54‑77%]
****
Recovery55% [34‑70%]
****
27% [16‑37%]
****
Cases81% [71‑87%]
****
Viral44% [29‑56%]
****
37% [7‑58%]
*
RCT mortality16% [-37‑49%]36% [7‑56%]
*
RCT casesRCT cases87% [56‑96%]
***
RCT viral20% [5‑33%]
*
10% [-43‑44%]
Loading..
Figure 3. Results by treatment stage.
Loading..
Loading..
Figure 4. Random effects meta-analysis for all studies. This plot shows pooled effects, analysis for individual outcomes is below, and more details on pooled effects can be found in the heterogeneity section. Effect extraction is pre-specified, using the most serious outcome reported. Simplified dosages are shown for comparison, these are the total dose in the first four days for treatment, or the monthly dose for prophylaxis, for a 70kg person. For details of effect extraction and full dosage information see the appendix.
Loading..
Loading..
Figure 5. Random effects meta-analysis for mortality.
Loading..
Figure 6. Random effects meta-analysis for mechanical ventilation.
Loading..
Figure 7. Random effects meta-analysis for ICU admission.
Loading..
Figure 8. Random effects meta-analysis for hospitalization.
Loading..
Figure 9. Random effects meta-analysis for recovery results only.
Loading..
Figure 10. Random effects meta-analysis for COVID-19 case results.
Loading..
Figure 11. Random effects meta-analysis for viral clearance.
Loading..
Figure 12. Random effects meta-analysis for peer-reviewed trials. Zeraatkar et al. analyze 356 COVID-19 trials, finding no significant evidence that preprint results are inconsistent with peer-reviewed studies. They also show extremely long peer-review delays, with a median of 6 months to journal publication. A six month delay was equivalent to around 1.5 million deaths during the first two years of the pandemic. Authors recommend using preprint evidence, with appropriate checks for potential falsified data, which provides higher certainty much earlier. Davidson et al. also showed no important difference between meta analysis results of preprints and peer-reviewed publications for COVID-19, based on 37 meta analyses including 114 trials. Effect extraction is pre-specified, using the most serious outcome reported, see the appendix for details.
Randomized Controlled Trials (RCTs)
Results restricted to Randomized Controlled Trials (RCTs) are shown in Figure 13, 14, 15, and 16, Table 1, and Table 2. The supplementary data contains RCT results after exclusions.
Loading..
Figure 13. Randomized Controlled Trials. The distribution of results for RCTs is similar to the distribution for all other studies.
Loading..
Figure 14. Random effects meta-analysis for Randomized Controlled Trials only. Effect extraction is pre-specified, using the most serious outcome reported, see the appendix for details.
Loading..
Figure 15. Random effects meta-analysis for RCT mortality results.
Loading..
Figure 17. Random effects meta-analysis for RCT case results.
Loading..
Figure 16. Random effects meta-analysis for RCT viral clearance results.
Bias in clinical research may be defined as something that tends to make conclusions differ systematically from the truth. RCTs help to make study groups more similar and can provide a higher level of evidence, however they are subject to many biases Jadad, and analysis of double-blind RCTs has identified extreme levels of bias Gøtzsche. For COVID-19, the overhead may delay treatment, dramatically compromising efficacy; they may encourage monotherapy for simplicity at the cost of efficacy which may rely on combined or synergistic effects; the participants that sign up may not reflect real world usage or the population that benefits most in terms of age, comorbidities, severity of illness, or other factors; standard of care may be compromised and unable to evolve quickly based on emerging research for new diseases; errors may be made in randomization and medication delivery; and investigators may have hidden agendas or vested interests influencing design, operation, analysis, reporting, and the potential for fraud. All of these biases have been observed with COVID-19 RCTs. There is no guarantee that a specific RCT provides a higher level of evidence.
RCTs are expensive and many RCTs are funded by pharmaceutical companies or interests closely aligned with pharmaceutical companies. For COVID-19, this creates an incentive to show efficacy for patented commercial products, and an incentive to show a lack of efficacy for inexpensive treatments. The bias is expected to be significant, for example Als-Nielsen et al. analyzed 370 RCTs from Cochrane reviews, showing that trials funded by for-profit organizations were 5 times more likely to recommend the experimental drug compared with those funded by nonprofit organizations. For COVID-19, some major philanthropic organizations are largely funded by investments with extreme conflicts of interest for and against specific COVID-19 interventions.
High quality RCTs for novel acute diseases are more challenging, with increased ethical issues due to the urgency of treatment, increased risk due to enrollment delays, and more difficult design with a rapidly evolving evidence base. For COVID-19, the most common site of initial infection is the upper respiratory tract. Immediate treatment is likely to be most successful and may prevent or slow progression to other parts of the body. For a non-prophylaxis RCT, it makes sense to provide treatment in advance and instruct patients to use it immediately on symptoms, just as some governments have done by providing medication kits in advance. Unfortunately, no RCTs have been done in this way. Every treatment RCT to date involves delayed treatment. Among the 66 treatments we have analyzed, 63% of RCTs involve very late treatment 5+ days after onset. No non-prophylaxis COVID-19 RCTs match the potential real-world use of early treatments. They may more accurately represent results for treatments that require visiting a medical facility, e.g., those requiring intravenous administration.
RCTs have a bias against finding an effect for interventions that are widely available — patients that believe they need the intervention are more likely to decline participation and take the intervention. RCTs for ivermectin are more likely to enroll low-risk participants that do not need treatment to recover, making the results less applicable to clinical practice. This bias is likely to be greater for widely known treatments, and may be greater when the risk of a serious outcome is overstated. This bias does not apply to the typical pharmaceutical trial of a new drug that is otherwise unavailable.
Evidence shows that non-RCT trials can also provide reliable results. Concato et al. found that well-designed observational studies do not systematically overestimate the magnitude of the effects of treatment compared to RCTs. Anglemyer et al. summarized reviews comparing RCTs to observational studies and found little evidence for significant differences in effect estimates. Lee et al. showed that only 14% of the guidelines of the Infectious Diseases Society of America were based on RCTs. Evaluation of studies relies on an understanding of the study and potential biases. Limitations in an RCT can outweigh the benefits, for example excessive dosages, excessive treatment delays, or Internet survey bias may have a greater effect on results. Ethical issues may also prevent running RCTs for known effective treatments. For more on issues with RCTs see Deaton, Nichol.
Currently, 44 of the treatments we analyze show statistically significant efficacy or harm, defined as ≥10% decreased risk or >0% increased risk from ≥3 studies. Of the 44 treatments with statistically significant efficacy/harm, 28 have been confirmed in RCTs, with a mean delay of 5.7 months. When considering only low cost treatments, 23 have been confirmed with a delay of 6.9 months. For the 16 unconfirmed treatments, 3 have zero RCTs to date. The point estimates for the remaining 13 are all consistent with the overall results (benefit or harm), with 10 showing >20%. The only treatments showing >10% efficacy for all studies, but <10% for RCTs are sotrovimab and aspirin.
We need to evaluate each trial on its own merits. RCTs for a given medication and disease may be more reliable, however they may also be less reliable. For off-patent medications, very high conflict of interest trials may be more likely to be RCTs, and more likely to be large trials that dominate meta analyses.
Exclusions
To avoid bias in the selection of studies, we analyze all non-retracted studies. Here we show the results after excluding studies with critical issues likely to alter results, non-standard studies, and studies where very minimal detail is currently available. Our bias evaluation is based on analysis of each study and identifying when there is a significant chance that limitations will substantially change the outcome of the study. We believe this can be more valuable than checklist-based approaches such as Cochrane GRADE, which may underemphasize serious issues not captured in the checklists, overemphasize issues unlikely to alter outcomes in specific cases (for example, lack of blinding for an objective mortality outcome, or certain specifics of randomization with a very large effect size), or be subject to bias. However, they can also be very high quality Bryant.
A team of researchers has analyzed the data in ivermectin studies and identified several studies with concerns. Retracted studies are not in this analysis. All other studies that the team has identified are excluded here. For more details see the response section.
Detailed description of issues with Beltran Gonzalez, Bramante, Hayward, López-Medina, Naggie, Reis, Vallejos can be found in the study notes section.
Soto-Becerra is a database analysis covering anyone with ICD-10 COVID-19 codes, which includes asymptomatic PCR+ patients. Therefore many patients in the control group are likely asymptomatic with regards to SARS-CoV-2, but in the hospital for another reason. For those that had symptomatic COVID-19, there is also likely significant confounding by indication. KM curves show that the treatment groups were in more serious condition, with more than the total excess mortality at 30 days occurring on day 1. All treatments are worse than the control group at 30 days, while at the latest followup all treatments show lower mortality than control. The machine learning system used also appears over-parameterized and likely to result in significant overfitting and inaccurate results. There is also no real control group in this study - patients receiving the treatments after 48 hours were put in the control group. Authors also state that outcomes within 24 hours were excluded, however the KM curves show significant mortality at day 1 (only for the treatment groups). Several protocol violations have also been reported in this study Yim. Note that this study provides both 30 day mortality and weighted KM curves up to day 43 for ivermectin, we use the day 43 results as per our protocol. IVERCOR PREP reports prophylaxis results, however only very minimal details are currently available in a news report and an online presentation. Hellwig analyze African countries and COVID-19 cases in October 2020 as a function of whether widespread prophylactic use of ivermectin is used for parasitic infections. Tanioka perform a similar analysis for COVID-19 mortality in January 2021. These studies are excluded because they are not clinical trials. Shahbaznejad had only one death that occurred in a patient that was critically ill at the time of admission and died within the first 24 hours. Galan perform an RCT comparing ivermectin and other treatments with very late stage severe condition hospitalized patients, not showing significant differences between the treatments. Authors were unable to add a control arm due to ethical issues. The closest control comparison we could find is Baqui, which shows 43% hospital mortality in the northern region of Brazil where the study was performed, from which we can estimate the mortality with ivermectin in this study as 47% lower, RR 0.53. Further, the study is restricted to more severe cases, hence the expected mortality, and therefore the benefit of treatment, may be higher. Kishoria restrict inclusion to patients that did not respond to standard treatment, provide no details on the time of the discharge status, and there are very large unadjusted differences in the groups, with over twice as many patients in the ivermectin group with age >40, and all patients over 60 in the ivermectin group. Angkasekwinai does not make sense as reported, for details see c19ivm.org.
Summarizing, the studies excluded are as follows, and the resulting forest plot is shown in Figure 18. The supplementary data shows results after restrictions and exclusions.
Abbas, very minimal patient information, three different results for the recovery outcome, selective omission of the statistically significant recovery p-value, and other inconsistencies.
Ahsan, unadjusted results with no group details.
Beltran Gonzalez, major inconsistencies reported and the data is no longer available Chamie, although the authors state that it is available, and have shared it with an anti-treatment group.
Borody, preliminary report with minimal details.
Buonfrate, significant unadjusted group differences, with 3 times as many patients in the ivermectin arms having the baseline visit in a hospital setting, and arm C having large differences in baseline gender, weight, cough, pyrexia, and anosmia, excessive dose for arm C.
Cadegiani, control group retrospectively obtained from untreated patients in the same population.
Carvallo, concern about potential data issues.
Carvallo (B), concern about potential data issues.
Carvallo (C), minimal details of groups provided.
de Jesús Ascencio-Montiel, unadjusted results with alternate outcome adjusted results showing significant changes with adjustments. Excluded results: death, mechanical ventilation, hospitalization, progression.
de la Rocha, data mismatch, no response from authors.
Elavarasi, unadjusted results with no group details.
Ferreira, unadjusted results with no group details; substantial unadjusted confounding by indication likely.
Hazan (B), study uses a synthetic control arm.
Hellwig, not a typical trial, analysis of African countries that used or did not use ivermectin prophylaxis for parasitic infections.
IVERCOR PREP, minimal details provided.
Kishoria, excessive unadjusted differences between groups.
López-Medina, strong evidence of patients in the control group self-medicating, ivermectin widely used in the population at that time, and the study drug identity was concealed by using the name D11AX22.
Mikamo, very low risk group with almost no progression leaves little room for improvement, unbalanced baseline dyspnea and high symptom scores, design and post-hoc changes favor null result.
Mustafa, unadjusted results with no group details.
Qadeer, minimal baseline details provided.
Ravikirti, exclusion of patients in less severe condition, data/analysis concerns.
Reis, multiple anomalies as per detailed analysis.
Rezai, multiple critical issues, see study page.
Rezai (B), multiple critical issues, see study page.
Roy, no serious outcomes reported and fast recovery in treatment and control groups, there is little room for a treatment to improve results.
Samajdar, minimal details provided; unadjusted results with no group details; results may be significantly affected by survey bias.
Schilling, post-hoc change to exclude patients treated before high viral load, population very low risk, recovering quickly without treatment, high baseline immunity, 2.2x greater baseline antibody negative for the treatment arm.
Soto, substantial unadjusted confounding by indication likely; substantial confounding by time possible due to significant changes in SOC and treatment propensity near the start of the pandemic.
Soto-Becerra, substantial unadjusted confounding by indication likely; includes PCR+ patients that may be asymptomatic for COVID-19 but in hospital for other reasons.
Szente Fonseca, result is likely affected by collinearity across treatments in the model.
Tanioka, not a typical trial, analysis of African countries that used or did not use ivermectin prophylaxis for parasitic infections.
Thairu, significant confounding by time possible due to separation of groups in different time periods.
Zubair, substantial unadjusted confounding by indication likely; unadjusted results with no group details.
Loading..
Figure 18. Random effects meta-analysis excluding studies with significant issues. Effect extraction is pre-specified, using the most serious outcome reported, see the appendix for details.
Heterogeneity
Heterogeneity in COVID-19 studies arises from many factors including:
The time between infection or the onset of symptoms and treatment may critically affect how well a treatment works. For example an antiviral may be very effective when used early but may not be effective in late stage disease, and may even be harmful. Oseltamivir, for example, is generally only considered effective for influenza when used within 0-36 or 0-48 hours McLean, Treanor. Baloxavir studies for influenza also show that treatment delay is critical — Ikematsu et al. report an 86% reduction in cases for post-exposure prophylaxis, Hayden et al. show a 33 hour reduction in the time to alleviation of symptoms for treatment within 24 hours and a reduction of 13 hours for treatment within 24-48 hours, and Kumar et al. report only 2.5 hours improvement for inpatient treatment.
Table 3. Studies of baloxavir for influenza show that early treatment is more effective.
Treatment delayResult
Post exposure prophylaxis86% fewer cases Ikematsu
<24 hours-33 hours symptoms Hayden
24-48 hours-13 hours symptoms Hayden
Inpatients-2.5 hours to improvement Kumar
Figure 19 shows a mixed-effects meta-regression for efficacy as a function of treatment delay in COVID-19 studies from 66 treatments, showing that efficacy declines rapidly with treatment delay. Early treatment is critical for COVID-19.
Loading..
Figure 19. Early treatment is more effective. Meta-regression showing efficacy as a function of treatment delay in COVID-19 studies from 66 treatments.
Details of the patient population including age and comorbidities may critically affect how well a treatment works. For example, many COVID-19 studies with relatively young low-comorbidity patients show all patients recovering quickly with or without treatment. In such cases, there is little room for an effective treatment to improve results (as in López-Medina et al.).
Efficacy may differ significantly depending on the effect measured, for example a treatment may be very effective at reducing mortality, but less effective at minimizing cases or hospitalization. Or a treatment may have no effect on viral clearance while still being effective at reducing mortality.
Efficacy may depend critically on the distribution of SARS-CoV-2 variants encountered by patients. Risk varies significantly across variants Korves, for example the Gamma variant shows significantly different characteristics Faria, Karita, Nonaka, Zavascki. Different mechanisms of action may be more or less effective depending on variants, for example the degree to which TMPRSS2 contributes to viral entry can differ across variants Peacock, Willett.
Effectiveness may depend strongly on the dosage and treatment regimen. Higher dosages have been found to be more successful for ivermectin Babalola. Method of administration may also be critical. Guzzo show that the plasma concentration of ivermectin is much higher when administered with food (Figure 20: geometric mean AUC 2.6 times higher). Many ivermectin studies specify fasting, or they do not specify administration. Fasting administration is expected to reduce effectiveness for COVID-19 due to lower plasma and tissue concentrations. Note that this is different to anthelmintic use in the gastrointestinal tract where fasting is recommended.
Figure 20. Ivermectin plasma concentration is significantly higher when administered with a meal. The graph shows mean plasma concentration (ng/ml) profiles following single oral doses of 30mg (fed and fasted administration), from Guzzo.
The use of other treatments may significantly affect outcomes, including supplements, other medications, or other kinds of treatment such as prone positioning. Treatments may be synergistic Alsaidi, Andreani, De Forni, Fiaschi, Jeffreys, Jitobaom, Jitobaom (B), Ostrov, Said, Thairu, Wan, therefore efficacy may depend strongly on combined treatments.
The quality of medications may vary significantly between manufacturers and production batches, which may significantly affect efficacy and safety. Williams et al. analyze ivermectin from 11 different sources, showing highly variable antiparasitic efficacy across different manufacturers. Xu et al. analyze a treatment from two different manufacturers, showing 9 different impurities, with significantly different concentrations for each manufacturer.
We present both pooled analyses and specific outcome analyses. Notably, pooled analysis often results in earlier detection of efficacy as shown in Figure 21. For many COVID-19 treatments, a reduction in mortality logically follows from a reduction in hospitalization, which follows from a reduction in symptomatic cases, etc. An antiviral tested with a low-risk population may report zero mortality in both arms, however a reduction in severity and improved viral clearance may translate into lower mortality among a high-risk population, and including these results in pooled analysis allows faster detection of efficacy. Trials with high-risk patients may also be restricted due to ethical concerns for treatments that are known or expected to be effective.
Pooled analysis enables using more of the available information. While there is much more information available, for example dose-response relationships, the advantage of the method used here is simplicity and transparency. Note that pooled analysis could hide efficacy, for example a treatment that is beneficial for late stage patients but has no effect on viral replication or early stage disease could show no efficacy in pooled analysis if most studies only examine viral clearance. While we present pooled results, we also present individual outcome analyses, which may be more informative for specific use cases.
Currently, 44 of the treatments we analyze show statistically significant efficacy or harm, defined as ≥10% decreased risk or >0% increased risk from ≥3 studies. 85% of treatments showing statistically significant efficacy/harm with pooled effects have been confirmed with one or more specific outcomes, with a mean delay of 3.7 months. When restricting to RCTs only, 50% of treatments showing statistically significant efficacy/harm with pooled effects have been confirmed with one or more specific outcomes, with a mean delay of 6.1 months.
Loading..
Loading..
Figure 21. The time when studies showed that treatments were effective, defined as statistically significant improvement of ≥10% from ≥3 studies. Pooled results typically show efficacy earlier than specific outcome results. Results from all studies often shows efficacy much earlier than when restricting to RCTs. Results reflect conditions as used in trials to date, these depend on the population treated, treatment delay, and treatment regimen.
The distribution of studies will alter the outcome of a meta analysis. Consider a simplified example where everything is equal except for the treatment delay, and effectiveness decreases to zero or below with increasing delay. If there are many studies using very late treatment, the outcome may be negative, even though early treatment is very effective. This may have a greater effect than pooling different outcomes such as mortality and hospitalization. For example a treatment may have 50% efficacy for mortality but only 40% for hospitalization when used within 48 hours. However efficacy could be 0% when used late.
All meta analyses combine heterogeneous studies, varying in population, variants, and potentially all factors above, and therefore may obscure efficacy by including studies where treatment is less effective. Generally, we expect the estimated effect size from meta analysis to be less than that for the optimal case. Looking at all studies is valuable for providing an overview of all research, important to avoid cherry-picking, and informative when a positive result is found despite combining less-optimal situations. However, the resulting estimate does not apply to specific cases such as early treatment in high-risk populations. While we present results for all studies, we also present treatment time and individual outcome analyses, which may be more informative for specific use cases.
Ivermectin studies vary widely in all the factors above, which makes the consistently positive results even more remarkable. A failure to detect an association after combining heterogeneous studies does not mean the treatment is not effective (it may only work in certain cases), however the reverse is not true — an identified association is valid, although the magnitude of the effect may be larger for more optimal cases, and lower for less optimal cases. Note that trial with a design favoring null results have become common, and are likely to dominate future trials. For example, the Together Trial tested ivermectin in locations known to have a high degree of self-medication, up to 7 days from onset (while claiming to be an early treatment trial), and using low doses compared to clinical recommendations for the dominant variant. The ACTIV-6 trial had a median treatment delay of 6 days and very low risk patients.
Discussion
Publishing is often biased towards positive results, which we would need to adjust for when analyzing the percentage of positive results. For ivermectin, there is strong evidence of a growing negative publication bias.
As Scott Alexander said in November 2021, "if you say anything in favor of ivermectin you will be cast out of civilization and thrown into the circle of social hell reserved for Klan members and 1/6 insurrectionists. All the health officials in the world will shout 'horse dewormer!' at you and compare you to Josef Mengele." In many locations, publishing positive ivermectin results is not conducive to maintaining employment or friendships. This can be seen in the design of recent trials, and the extreme measures taken to avoid presenting statistically significant positive results, as detailed in the study notes below.
One method to evaluate bias is to compare prospective vs. retrospective studies, although this method has become less useful with ivermectin due to the increase of prospective studies with designs favoring null results. Prospective studies are likely to be published regardless of the result, while retrospective studies are more likely to exhibit bias. For example, researchers may perform preliminary analysis with minimal effort and the results may influence their decision to continue. Retrospective studies also provide more opportunities for the specifics of data extraction and adjustments to influence results. Figure 22 shows a scatter plot of results for prospective and retrospective studies. Prospective studies show 60% [49‑69%] improvement in meta analysis, compared to 61% [49‑70%] for retrospective studies, showing no significant difference. Bryant also perform a funnel plot analysis, which they found did not suggest evidence of publication bias.
Loading..
Figure 22. Prospective vs. retrospective studies. The diamonds show the results of random effects meta-analysis.
There is substantial evidence that journals are rejecting and delaying the publication of positive studies, for example by accepting a paper for review, holding it for some time, and then rejecting it without review Jerusalem Post, Kory (B). One group performed prophylaxis and early treatment studies, with only the less positive study being formally published to date IVERCOR PREP, Vallejos, suggesting a negative publication bias. Dr. Eli Schwartz's Biber double blind RCT was rejected without review by The Lancet and Clinical Infectious Diseases Fox. Authors of Efimenko do not plan to submit the very positive results to a journal, and have self-censored the conference publication, providing further evidence of a negative publication bias. Trials with pending and possibly delayed publication often involve researchers that may be restricted due to politics — publishing positive results may be incompatible with continued employment, whereas negative results can receive priority treatment at certain well-known journals, support the positions of employers or funding organizations, and receive substantial press. For more details of censorship and negative publication bias in ivermectin research see Kory (C), Kory (D), Kory (E).
News coverage of ivermectin studies is extremely biased. Only studies with designs favoring null results have received significant press coverage in western media López-Medina, Naggie, Reis, all of which have multiple critical issues as discussed below, but ignored by the press.
Table 4 shows the reported results of physicians that use early treatments for COVID-19, compared to the results for a non-treating physician (this physician reportedly prescribed early treatment for themself, but not for patients medicospelavidacovid19.com.br). The treatments used vary between physicians. Almost all report using ivermectin and/or HCQ, and most use additional treatments in combination. These results are subject to selection and ascertainment bias and more accurate analysis requires details of the patient populations and followup, however results are consistently better across many teams, and consistent with the extensive controlled trial evidence that shows a significant reduction in risk with many early treatments, and improved results with the use of multiple treatments in combination.
Table 4. Physician results with early treatment protocols compared to no early treatment. (*) Dr. Uip reportedly prescribed early treatment for himself, but not for patients medicospelavidacovid19.com.br.
LATE TREATMENT
Physician / TeamLocationPatients HospitalizationHosp. MortalityDeath
Dr. David Uip (*) Brazil 2,200 38.6% (850) Ref. 2.5% (54) Ref.
EARLY TREATMENT - 39 physicians/teams
Physician / TeamLocationPatients HospitalizationHosp. ImprovementImp. MortalityDeath ImprovementImp.
Dr. Roberto Alfonso Accinelli
0/360 deaths for treatment within 3 days
Peru 1,265 0.6% (7) 77.5%
Dr. Mohammed Tarek Alam
patients up to 84 years old
Bangladesh 100 0.0% (0) 100.0%
Dr. Oluwagbenga Alonge Nigeria 310 0.0% (0) 100.0%
Dr. Raja Bhattacharya
up to 88yo, 81% comorbidities
India 148 1.4% (2) 44.9%
Dr. Flavio Cadegiani Brazil 3,450 0.1% (4) 99.7% 0.0% (0) 100.0%
Dr. Alessandro Capucci Italy 350 4.6% (16) 88.2%
Dr. Shankara Chetty South Africa 8,000 0.0% (0) 100.0%
Dr. Deborah Chisholm USA 100 0.0% (0) 100.0%
Dr. Ryan Cole USA 400 0.0% (0) 100.0% 0.0% (0) 100.0%
Dr. Marco Cosentino
vs. 3-3.8% mortality during period; earlier treatment better
Italy 392 6.4% (25) 83.5% 0.3% (1) 89.6%
Dr. Jeff Davis USA 6,000 0.0% (0) 100.0%
Dr. Dhanajay India 500 0.0% (0) 100.0%
Dr. Bryan Tyson & Dr. George Fareed USA 20,000 0.0% (6) 99.9% 0.0% (4) 99.2%
Dr. Raphael Furtado Brazil 170 0.6% (1) 98.5% 0.0% (0) 100.0%
Dr. Heather Gessling USA 1,500 0.1% (1) 97.3%
Dr. Ellen Guimarães Brazil 500 1.6% (8) 95.9% 0.4% (2) 83.7%
Dr. Syed Haider USA 4,000 0.1% (5) 99.7% 0.0% (0) 100.0%
Dr. Mark Hancock USA 24 0.0% (0) 100.0%
Dr. Sabine Hazan USA 1,000 0.0% (0) 100.0%
Dr. Mollie James USA 3,500 1.1% (40) 97.0% 0.0% (1) 98.8%
Dr. Roberta Lacerda Brazil 550 1.5% (8) 96.2% 0.4% (2) 85.2%
Dr. Katarina Lindley USA 100 5.0% (5) 87.1% 0.0% (0) 100.0%
Dr. Ben Marble USA 150,000 0.0% (4) 99.9%
Dr. Edimilson Migowski Brazil 2,000 0.3% (7) 99.1% 0.1% (2) 95.9%
Dr. Abdulrahman Mohana Saudi Arabia 2,733 0.0% (0) 100.0%
Dr. Carlos Nigro Brazil 5,000 0.9% (45) 97.7% 0.5% (23) 81.3%
Dr. Benoit Ochs Luxembourg 800 0.0% (0) 100.0%
Dr. Ortore Italy 240 1.2% (3) 96.8% 0.0% (0) 100.0%
Dr. Valerio Pascua
one death for a patient presenting on the 5th day in need of supplemental oxygen
Honduras 415 6.3% (26) 83.8% 0.2% (1) 90.2%
Dr. Sebastian Pop Romania 300 0.0% (0) 100.0%
Dr. Brian Proctor USA 869 2.3% (20) 94.0% 0.2% (2) 90.6%
Dr. Anastacio Queiroz Brazil 700 0.0% (0) 100.0%
Dr. Didier Raoult France 8,315 2.6% (214) 93.3% 0.1% (5) 97.6%
Dr. Karin Ried
up to 99yo, 73% comorbidities, av. age 63
Turkey 237 0.4% (1) 82.8%
Dr. Roman Rozencwaig
patients up to 86 years old
Canada 80 0.0% (0) 100.0%
Dr. Vipul Shah India 8,000 0.1% (5) 97.5%
Dr. Silvestre Sobrinho Brazil 116 8.6% (10) 77.7% 0.0% (0) 100.0%
Dr. Unknown Brazil 957 1.7% (16) 95.7% 0.2% (2) 91.5%
Dr. Vladimir Zelenko USA 2,200 0.5% (12) 98.6% 0.1% (2) 96.3%
Mean improvement with early treatment protocols 237,521 HospitalizationHosp. 94.1% MortalityDeath 94.7%
Funnel plots have traditionally been used for analyzing publication bias. This is invalid for COVID-19 acute treatment trials — the underlying assumptions are invalid, which we can demonstrate with a simple example. Consider a set of hypothetical perfect trials with no bias. Figure 23 plot A shows a funnel plot for a simulation of 80 perfect trials, with random group sizes, and each patient's outcome randomly sampled (10% control event probability, and a 30% effect size for treatment). Analysis shows no asymmetry (p > 0.05). In plot B, we add a single typical variation in COVID-19 treatment trials — treatment delay. Consider that efficacy varies from 90% for treatment within 24 hours, reducing to 10% when treatment is delayed 3 days. In plot B, each trial's treatment delay is randomly selected. Analysis now shows highly significant asymmetry, p < 0.0001, with six variants of Egger's test all showing p < 0.05 Egger, Harbord, Macaskill, Moreno, Peters, Rothstein, Rücker, Stanley. Note that these tests fail even though treatment delay is uniformly distributed. In reality treatment delay is more complex — each trial has a different distribution of delays across patients, and the distribution across trials may be biased (e.g., late treatment trials may be more common). Similarly, many other variations in trials may produce asymmetry, including dose, administration, duration of treatment, differences in SOC, comorbidities, age, variants, and bias in design, implementation, analysis, and reporting.
Figure 23. Example funnel plot analysis for simulated perfect trials.
Some authors claim that Caly showed that therapeutic concentrations are not easily reached in humans. This is incorrect. The authors explain why their in vitro study cannot be used to determine the effective dose in vivo, and state that the concentration required is very unlikely to be an issue Wagstaff (C). The study used monkey kidney cells (the only choice at the time of the experiments), which lack adaptive immune responses and do not produce interferon. Authors also note that ivermectin accumulates in lung and other tissues, that subsequent experiments with lung cells show many times greater concentrations, and that the average lung concentration shown in modeling studies exceeds the effective level shown in their research. Tissue concentrations of ivermectin can be much higher than plasma concentration Lespine, Lifschitz. Authors note that ivermectin works with the immune system and a 1:1 ratio of drug to virus is unlikely to be required. In Bray, authors reply that "ivermectin's key direct target in mammalian cells is a not a viral component, but a host protein important in intracellular transport; the fact that it is a host-directed agent (HDA) is almost certainly the basis of its broad-spectrum activity against a number of different RNA viruses in vitro. The way a HDA can reduce viral load is by inhibiting a key cellular process that the virus hijacks to enhance infection by suppressing the host antiviral response. Reducing viral load by even a modest amount by using a HDA at low dose early in infection can be the key to enabling the body's immune system to begin to mount the full antiviral response before the infection takes control." In further research, authors note that they find efficacy for prophylactic use, and that smaller repeated doses are more efffective than a single larger dose Wagstaff (C). Moreover, there are now 22 In Vitro studies that support the efficacy of ivermectin for COVID-19 Boschi, Caly, Croci, De Forni, Delandre, Fauquet, García-Aguilar, Jeffreys, Jitobaom, Jitobaom (B), Li, Liu (B), Liu (C), Mody, Mountain Valley MD, Munson, Saha (B), Segatori, Surnar, Yesilbag, Zhang, Zheng.
One theory for the beneficial effect of ivermectin for COVID-19 is related to strongyloides and the use of steroids — control group patients with strongyloides may be at risk due to steroid use, while ivermectin patients are protected. While this mechanism may contribute to efficacy in some cases, it is inconsistent with the data. If this was the case, we would expect to see greater benefit in late stage trials where steroids are used more often, and we would expect to see greater benefit for outcomes that occur after steroids are used. However, we see a very strong opposite effect for treatment time, and we see comparable or stronger efficacy for earlier outcomes.
The theory has gained renewed interest based on a new analysis Bitterman. However, this analysis is confounded by treatment delay, dose, conflicts of interest, and other factors, and the effect disappears when analyzing all studies, all RCTs, or all mortality results, as shown in Figure 24.
Although the first author has responded to the confounders on Twitter, we do not see mention of them in the paper. Author is also aware that the larger sets of all trials, all RCTs, or all mortality results do not show the effect, however we also do not see this mentioned in the paper. These omissions suggest investigator bias. Author claims they could not discuss these issues due to publication delays, however the paper was accepted Jan 31, 2022, and author was aware of the issues months before, for example discussing treatment delay and dose in Nov 2021. These confounders are also basic and not really possible to miss.
The meta analysis for Hashim includes critical patients, however these patients were always allocated to the treatment arm for ethical reasons, therefore including them is not logical and introduces substantial bias. According to the author response, this appears to have been known, suggesting investigator bias. Authors include Shahbaznejad where the only death was a critical patient that died within 24 hours of admission.
Although authors note following PRISMA guidelines, we do not see registration of the protocol or discussion thereof. We note that the current protocol is the result of multiple changes to the original methodology as posted on Twitter: from 3 groups to 2 groups, altering the included studies, and switching from using one source for prevalence estimates to selecting estimate sources on a per study basis, which allows potential bias in the selection. Notably, this resulted in moving the Together Trial (Brazil) into the low prevalence category.
Author's results rely on trials with a very small number of mortality events — the high stronglyoides prevalance group has trials with 1, 3, 4, and 13 events. Authors do mention limitations due to the small number of events and the reliability of strongyloides estimates.
Authors indicate no conflicts of interest, however the first author has been an investigator on a Pfizer trial, which may be NCT04092452, showing completion in January 2022 clinicaltrials.gov, openpaymentsdata.cms.gov.
A
Loading..
B
Loading..
C
Loading..
Figure 24. Mixed-effects meta-regression showing efficacy as a function of strongyloides prevalence. A. all studies. B. all RCTs. C. all mortality results.
The following refers to the first author's analysis posted earlier on Twitter. The author selected 10 of the 101 studies, with 3 in a high strongyloides prevalence group where a greater benefit is seen. This was used to draw strong conclusions about the mechanism of ivermectin efficacy.
There are several limitations to this analysis. One of the 3 studies does not mention steroids in the list of SOC medications, while a second reports 6% usage for the control group. Author has added a fourth paper in a revised grouping with 11 studies.
We perfomed a similar analysis for all studies (except the 2 ecological studies), which shows no significant effect, with the high prevalence group actually showing lower improvement (50% [35‑62%] vs. 65% [57‑72%] for the low prevalence group). Details can be found in the supplementary data. Results are similar when restricting to mortality results or when restricting to RCTs.
Why does the smaller analysis with 11 studies show a greater benefit in high strongyloides prevalence regions? The effect is based on relatively few events - 1, 3, 4, and 13 respectively for the high prevalence group. More importantly, the result is confounded by treatment delay and dose.
Treatment delay. All meta analyses combine heterogeneous studies which results in limitations. For example in pooled analysis we combine hospitalization and mortality. In terms of evaluating efficacy for COVID-19 treatments, reduction in hospitalization reasonably leads to reduction in mortality for high-risk populations. Both are indicators of efficacy, and both are valuable. In the largest series of COVID-19 treatment trials, hospitalization and mortality estimates are very similar. The same does not apply to treatment delay for antivirals. A trial showing efficacy with early treatment provides no information on late treatment, and a trial showing no efficacy with late treatment provides no information on early treatment. Ivermectin, as with many COVID-19 treatments, shows a strong treatment delay relationship — early treatment shows significantly higher efficacy.
The high prevalence group in the 11 study analysis has more early treatment trials, and the low prevalence group has more late treatment trials. The result is confounded by treatment delay, and reflects the greater efficacy of early treatment.
Only one trial in the high prevalence group is classified as late treatment, I-TECH, which was very close to the cutoff. Moreover, of all trials in the 11 trial analysis, this one uses the highest dose.
Dose. The average dosage used in the high prevalence group is about twice the dose in the low prevalence group, and would be close to three times higher if the Together Trial was not moved to the low prevalence group. The result is confounded by dose, and reflects the greater efficacy of higher dosages.
Variants. Efficacy may vary based on variants. Notably, the Gamma variant was most common for one trial in the low prevalance group. This variant shows dramatically different characteristics Zavascki, and clinicians report that significantly higher dosage and/or earlier treatment is required, as may be expected for variants where the peak viral load is significantly higher and/or reached earlier Faria, Nonaka.
Conflicts of interest. Two trials have very high (>$US1B) negative conflicts of interest which may introduce bias towards null effects. The trial in the low prevalance group shows a lower effect size. The trial in the high prevalence group also shows a lower effect size for the primary outcome. This trial shows a larger mortality effect, however with only one event this has very low significance.
Summary. In summary, the greater benefit in high strongyloides prevalence regions is only seen with the small subset of 11 trials and is not seen with all trials, or after restriction to mortality results, or restriction to RCTs. Within the 11 trial sample, all trials except one in the low prevalence group have confounding due to treatment delay and/or low dosage, where a lower effect size is expected. The only remaining trial in the group is unpublished, has an unknown treatment delay (a significant percentage of patients may have been treated very late), has very high negative conflicts of interest, and the Gamma variant was most common, in addition to other issues.
Pharmaceutical drug trials often have conflicts of interest whereby sponsors or trial staff have a financial interest in the outcome being positive. Ivermectin for COVID-19 lacks this because it is off-patent, has many manufacturers, and is very low cost. In contrast, most COVID-19 ivermectin trials have been run by physicians on the front lines with the primary interest of finding the best methods to save human lives and minimize the collateral damage caused by COVID-19. While pharmaceutical companies are careful to run trials under optimal conditions (for example, restricting patients to those most likely to benefit, only including patients that can be treated soon after onset when necessary, ensuring accurate dosing), many ivermectin trials do not represent the optimal conditions for efficacy.
Two ivermectin trials to date involve very large financial conflicts of interest López-Medina, Reis — companies closely involved with the trial or organizers stand to lose billions of dollars if ivermectin efficacy becomes more widely known. The design of these trials favors producing a null outcome as detailed in López-Medina, Reis. Note that biasing an RCT to produce a false positive result is difficult (suppressing adverse events is relatively easy Evans), but biasing a trial to produce a false negative result is very easy — for example, in a trial of an antiviral that works within the first 24 hours of symptom onset, trial organizers only need to avoid treating people within the first 24 hours; or with a disease like COVID-19, organizers only need to select a low-risk population where most people recover quickly without treatment. We note that, even under the very suboptimal designs, these trials produced positive results, although without statistical significance.
Several major trials have been designed in a way that favors finding no effect, with a number of methods including late treatment, selecting low-risk patients, fasting administration, very high conflict of interest medication sourcing, and dosing below current clinical practice. For discussion see Goodkin. One patient reported their experience with one of the remote outpatient ivermectin/fluvoxamine trials: they were offered enrollment 7 days after symptoms (receipt of medication would be even later), were offered $400 to participate, and reportedly target healthy people twitter.com. Details of issues with several trials are included in the study notes.
Other treatments may reduce the efficacy of ivermectin, for example dexamethasone may interfere with the pharmacokinetics of ivermectin, significantly reducing plasma concentration and efficacy Areskog.
The 101 studies are from 91 independent research teams. 5 studies compare against other treatments rather than placebo. Currently ivermectin shows better results than these other treatments, however ivermectin may show greater improvement when compared to placebo. 17 of 101 studies combine treatments, for example ivermectin + doxycycline. The results of ivermectin alone may differ. 4 of 48 RCTs use combined treatment, three with doxycycline, and one with iota-carrageenan. 1 of 101 studies currently has minimal published details available.
Many reviews cover ivermectin for COVID-19, presenting additional background on mechanisms, formulations, and related results, including Adegboro, Al-kuraishy, Babalola (B), Behl, Bray, DiNicolantonio, DiNicolantonio (B), Elkholy, Fordham, Formiga, Heidary, Jagiasi, Jans, Jans (B), Kory, Kory (C), Kory (F), Kory (G), Kow, Lind, Liu, Loo, Low, Santin, Scheim, Schwartz, Semiz, Turkia, Turkia (B), Turkia (C), Vora, Wang, Wehbe, Yagisawa, Yemeke, Zaidi.
Summary statistics from meta analysis necessarily lose information. As with all meta analyses, studies are heterogeneous, with differences in treatment delay, treatment regimen, patient demographics, variants, conflicts of interest, standard of care, and other factors. We provide analyses by specific outcomes and by treatment delay, and we aim to identify key characteristics in the forest plots and summaries. Results should be viewed in the context of study characteristics.
Some analyses classify treatment based on early or late administration, as done here, while others distinguish between mild, moderate, and severe cases. Viral load does not indicate degree of symptoms — for example patients may have a high viral load while being asymptomatic. With regard to treatments that have antiviral properties, timing of treatment is critical — late administration may be less helpful regardless of severity.
Details of treatment delay per patient is often not available. For example, a study may treat 90% of patients relatively early, but the events driving the outcome may come from 10% of patients treated very late. Our 5 day cutoff for early treatment may be too conservative, 5 days may be too late in many cases.
Comparison across treatments is confounded by differences in the studies performed, for example dose, variants, and conflicts of interest. Trials affiliated with special interests may use designs better suited to the preferred outcome.
In some cases, the most serious outcome has very few events, resulting in lower confidence results being used in pooled analysis, however the method is simpler and more transparent. This is less critical as the number of studies increases. Restriction to outcomes with sufficient power may be beneficial in pooled analysis and improve accuracy when there are few studies, however we maintain our pre-specified method to avoid any retrospective changes.
Studies show that combinations of treatments can be highly synergistic and may result in many times greater efficacy than individual treatments alone Alsaidi, Andreani, De Forni, Fiaschi, Jeffreys, Jitobaom, Jitobaom (B), Ostrov, Said, Thairu, Wan. Therefore standard of care may be critical and benefits may diminish or disappear if standard of care does not include certain treatments.
This real-time analysis is constantly updated based on submissions. Accuracy benefits from widespread review and submission of updates and corrections from reviewers. Less popular treatments may receive fewer reviews.
No treatment, vaccine, or intervention is 100% available and effective for all current and future variants. Efficacy may vary significantly with different variants and within different populations. All treatments have potential side effects. Propensity to experience side effects may be predicted in advance by qualified physicians. We do not provide medical advice. Before taking any medication, consult a qualified physician who can compare all options, provide personalized advice, and provide details of risks and benefits based on individual medical history and situations.
Typical meta analyses involve subjective selection criteria, effect extraction rules, and study bias evaluation, which can be used to bias results towards a specific outcome. In order to avoid bias we include all studies and use a pre-specified method to extract results from all studies (we also present results after exclusions). The results to date are overwhelmingly positive, very consistent, and very insensitive to potential selection criteria, effect extraction rules, and/or bias evaluation. Additional meta analyses confirming the effectiveness of ivermectin can be found in Bryant, Hariyanto, Kory, Lawrie, Nardelli, Ragó. Kory et al. also review epidemiological data and provide suggested treatment regimens.
The evidence supporting ivermectin for COVID-19 far exceeds the typical amount of evidence used for the approval of treatments. Lee shows that only 14% of the guidelines of the Infectious Diseases Society of America were based on RCTs. Table 5 and Table 6 compare the amount of evidence for ivermectin compared to that used for other COVID-19 approvals, and that used by WHO for the approval of ivermectin for scabies and strongyloidiasis. Table 7 compares US CDC recommendations for ibuprofen and ivermectin.
Table 5. WHO ivermectin approval status.
IndicationStudiesPatientsStatus
Strongyloidiasis Kory (H)5591Approved
Scabies Kory (H)10852Approved
COVID‑19101142,247 Pending
COVID‑19 RCTs4816,847
Table 6. Evidence base used for other COVID-19 approvals compared with the ivermectin evidence base.
MedicationStudiesPatientsImprovementStatus
Molnupiravir (UK)177550%Approved
Budesonide (UK)11,77917%Approved
Remdesivir (USA EUA)11,06331%Approved
Casiri/imdevimab (USA EUA)179966%Approved
Ivermectin evidence101142,247 61% [53‑67%] Pending
Table 7. Comparison of CDC recommendations Kory (H).
IbuprofenIvermectin
(for scabies)
Ivermectin
(for COVID-19)
Lives saved00>500,000
Deaths per year~450<1<1
CDC recommendedYesYesNo
Based on0 RCTs10 RCTs
852 patients
48 RCTs
16,847 patients
WHO updated their treatment recommendations on 3/30/2021 WHO. For ivermectin they reported a mortality odds ratio of 0.19 [0.09-0.36] based on 7 studies with 1,419 patients. They do not specify which trials they included. The report is inconsistent, with a forest plot that only shows 4 studies with mortality results. WHO's recommendation has not been updated for days.
Despite this extremely positive result, they recommended only using ivermectin in clinical trials. The analysis contains many flaws Kory (I):
Of the 101 studies (48 RCTs), they only included 16.
They excluded all 17 prophylaxis studies (4 RCTs).
There was no protocol for data exclusion.
Trials included in the original UNITAID search protocol were excluded.
They excluded all epidemiological evidence, although WHO has considered such evidence in the past.
They combine early treatment and late treatment studies and do not provide heterogeneity information. As above, early treatment is more successful, so pooling late treatment studies will obscure the effectiveness of early treatment. They chose not to do subgroup analysis by disease severity across trials, although treatment delay is clearly a critical factor in COVID-19 treatment, the analysis is easily done (as above), and it is well known that the studies for ivermectin and many other treatments clearly show greater effectiveness for early treatment.
WHO downgraded the quality of trials compared to the UNITAID systematic review team and a separate international expert guideline group that has long worked with the WHO Bryant.
They disregarded their own guidelines that stipulate quality assessments should be upgraded when there is evidence of a large magnitude effect (which there is), and when there is evidence of a dose-response relationship (which there is). They claim there is no dose-response relationship, while the UNITAID systematic review team found a clear relationship, along with individual studies Babalola.
Their risk of bias assessments do not match the actual risk of bias in studies. For example they classify López-Medina as low risk of bias, however this study has many issues making the results unreliable Covid Analysis, even prompting an open letter from over 170 physicians concluding that the study is fatally flawed Open Letter by 170+ US Doctors. Beltran Gonzalez is also classified as low risk of bias, but is a study with very late stage severe condition high-comorbidity patients. There is a clear treatment delay-response relationship and very late stage treatment is not expected to be as effective as early treatment. Conversely, much higher quality studies were classified as high risk of bias.
Although WHO's analysis is called a "living guideline", it is rarely updated and very out of date. As of May 14, 2021, four of the missing RCTs are known to WHO and labeled "RCTs pending data extraction" COVID-NMA. We added these 4, 4, 2, and one month earlier.
A single person served as Methods Chair, member of the Guidance Support Collaboraton Committee, and member of the Living Systematic Review/NMA team.
Public statements from people involved in the analysis suggest substantial bias. For example, a co-chair reportedly said that "the data available was sparse and likely based on chance" Reuters. The clinical team lead refers to their analysis of ivermectin as "fighting this overuse of unproven therapies ... without evidence of efficacy" Reuters, despite the extensive evidence of efficacy from the 101 studies by 1,129 scientists with 142,247 patients. People involved may be more favorable to late stage treatment of COVID-19, for example the co-chair recommended treating severe COVID-19 with remdesivir Rochwerg.
In summary, although WHO's analysis predicts that over 2 million fewer people would be dead if ivermectin was used from early in the pandemic, they recommend against use outside trials. This appears to be based primarily on excluding the majority of the evidence, and by assigning bias estimates that do not match the actual risk of bias in studies.
Use early in the pandemic was proposed by Kitasato University including the co-discoverer of ivermectin, Dr. Satoshi Ōmura. They requested Merck conduct clinical trials of ivermectin for COVID-19 in Japan, because Merck has priority to submit an application for an expansion of ivermectinʼs indications. Merck declined Yagisawa.
Merck has recommended against ivermectin Merck, however this recommendation has not been updated for days.
They state that there is "no meaningful evidence for clinical activity or clinical efficacy in patients with COVID-19 disease". This is contradicted by numerous studies including Alam, Aref, Aref (B), Babalola, Baguma, Behera, Behera (B), Bernigaud, Budhiraja, Bukhari, Chaccour (B), Chahla, Chahla (B), Chowdhury, de Jesús Ascencio-Montiel, Elalfy, Espitia-Hernandez, Faisal, Ghauri, Hashim, Huvemek, Kerr, Khan, Lima-Morales, Loue, Mahmud, Manomaipiboon, Mayer, Merino, Mohan, Mondal, Morgenstern, Mourya, Okumuş, Ravikirti (B), Seet, Shimizu.
They also claim that there is "a concerning lack of safety data in the majority of studies". Safety analysis is found in Descotes, Errecalde, Guzzo, Kory, Madrid, and safety data can be found in most studies, including Abd-Elsalam, Ahmed, Aref, Babalola, Behera (B), Bhattacharya, Biber, Bukhari, Camprubí, Carvallo (C), Chaccour (B), Chahla (B), Chowdhury, Elalfy, Espitia-Hernandez, George, Ghauri, Gorial, Hazan (B), Huvemek, Khan, Kishoria, Krolewiecki, Lima-Morales, Loue, López-Medina, Mahmud, Mohan, Morgenstern, Mourya, Okumuş, Pott-Junior, Seet, Shahbaznejad, Shouman, Spoorthi, Szente Fonseca, Vallejos, Zubair.
Merck has a number of conflicts of interest:
Merck has committed to give ivermectin away for free "as much as needed, for as long as needed" in the Mectizan® Donation Program Merck (B), to help eliminate river blindness.
Merck has their own new COVID-19 treatments MK-7110 (formerly CD24Fc) Adams and Molnupiravir (MK-4482) Jayk Bernal, Wikipedia. Merck has a ~$US1.2B agreement to supply molnupiravir to the US government, if it receives EUA or approval Khan (B). Over $US10B in near-term orders are expected if approved Genetic Engineering and Biotechnology News.
Ivermectin is off-patent, there are many manufacturers, and Merck is unlikely to be able to compete with low cost manufacturers.
Promoting the use of low cost off-patent medications compared to new products may be undesirable to some shareholders.
Japan requested Merck conduct clinical trials early in the pandemic and they declined. Merck may be reluctant to admit this mistake Yagisawa.
For other concerns regarding Merck's statement and prior actions related to Vioxx, see Scheim (B).
The US FDA recommended against ivermectin on March 5, 2021, however they stated that "The FDA has not reviewed data to support use of ivermectin in COVID-19 patients to treat or to prevent COVID-19". There is still no indication that the FDA has reviewed the clinical trials days later.
The FDA notes that they "received multiple reports of patients who have required medical support and been hospitalized after self-medicating with ivermectin intended for horses". The number of reports was 4 Pfeiffer. For comparison, acetaminophen overdose results in ~33,000 yearly hospitalizations in the USA (~12,000 unintentional) Charilaou. The FDA's recommendation may increase cases of self-medication with animal ivermectin, because it reduces the percentage of prescribing physicians.
They state that "Ivermectin is not an anti-viral", however many studies contradict this Ahmed, Aref, Babalola, Biber, Bukhari, Buonfrate, Caly, Chowdhury, Elalfy, Espitia-Hernandez, Khan, Mahmud, Mohan, Mourya, Okumuş, Qadeer, Rezk, Thairu, including 10 RCTs.
They note that "some initial research is underway", however there had been many studies completed and published prior to the FDA recommendation: Ahmed, Alam, Babalola, Behera, Beltran Gonzalez, Bernigaud, Biber, Budhiraja, Bukhari, Cadegiani, Camprubí, Carvallo (C), Chaccour (B), Chachar, Chahla (B), Chowdhury, Elalfy, Espitia-Hernandez, Ghauri, Gorial, Hashim, Hellwig, Khan, Lima-Morales, López-Medina, Mahmud, Mohan, Okumuş, Podder, Rajter, Ravikirti (B), Shouman, Spoorthi, including 17 RCTs.
Sep 3, 2021: The FDA revised their statement slightly. They removed the false claim that invermectin is not an antiviral, and they removed the statement that they have not reviewed the data. However, there is still nothing to indicate that they have reviewed the clinical trials. Indeed, they state "currently available data do not show ivermectin is effective against COVID-19" and "ivermectin has not been shown to be safe or effective for these indications", which are both false.
Update: NIH has updated the recommendation, based heavily on the Together Trial, while making no mention of the impossible data, blinding, randomization, and protocol failures, or that the co-principal investigator privately reported that "There is a clear signal that IVM works in COVID patients".
NIH has reported that there is insufficient evidence to recommended for or against ivermectin NIH. A table with summaries of 7 studies is provided, dated Dec 16, 2021, and they reference another 23 studies without analysis, however there are 101 studies to date. No quantitative analysis is provided. The NIH recommendation is "insufficient evidence", indicating that they must review new evidence immediately. Lack of updates suggest bias.
The likely members of the panel have been revealed by FOIA requests Yim (B). In the first request, all but two member names were redacted drive.google.com, however all are visible in a second request drive.google.com (B). Major conflicts of interest have been reported trialsitenews.com, Yim (B). 7 of 9 panel members appear to have conflicts of interest. Submit Corrections or Updates
Prof. Adaora Adimora (adimora@med.unc.edu)
Merck: advisory board, consultant, research support covid19treatmentguidelines.nih.gov, trialsitenews.com (B)
Gilead (maker of remdesivir): consultant, research support files.covid19treatmentguidelines.nih.gov, tandfonline.com "Aadimora AA has received consulting fees from Viiv, and Gilead and her institution has received funding from Gilead for her research"
Open Payments shows $62,000 from Merck and $88,000 from Gilead openpaymentsdata.cms.gov (B)
Prof. Roger Bedimo (roger.bedimo@va.gov)
Open Payments shows $149,000 from Merck, $76,000 from Sanofi, and $23,000 from Gilead openpaymentsdata.cms.gov (C)
Prof. Rajesh Gandhi (rgandhi@mgh.harvard.edu)
Gilead: grants, advisory board, personal fees nejm.org, rmed.acponline.org "Dr. Gandhi reports grants and personal fees from Gilead, personal fees from Merck, grants and personal fees from Theratechnologies, grants from ViiV, grants from Janssen."
Merck: advisory board and personal fees files.covid19treatmentguidelines.nih.gov, nejm.org, rmed.acponline.org
Janssen: grants nejm.org
Open Payments shows $45,000 from Merck and $10,000 from Gilead openpaymentsdata.cms.gov (D)
Prof. David Glidden (david.glidden@ucsf.edu)
Prof. Roy Gulick (rgulick@med.cornell.edu)
Merck: grants web.archive.org (deleted from current version medscape.org) "Roy M. Gulick, MD, MPH, has disclosed that he has received grants for clinical research from Abbott, Boehringer Ingelheim, Merck, Pfizer, Schering, and Tibotec, and has received grants for educational activities from Gilead and Monogram. Dr. Gulick has also disclosed that he has served as an ad-hoc advisor or consultant to Abbott, Boehringer Ingelheim, Bristol-Myers Squibb, Gilead, GlaxoSmithKline, Pfizer, Schering, and Tibotec", principal investigator on A5391 with funding in part from Merck actgnetwork.org
Pfizer: grants and ad-hoc advisor or consultant web.archive.org (deleted from current version medscape.org
Gilead: grants and ad-hoc advisor or consultant web.archive.org (deleted from current version medscape.org
GlaxoSmithKline: ad-hoc advisor or consultant web.archive.org (deleted from current version medscape.org
Prof. Susanna Naggie (susanna.naggie@duke.edu)
AbbVie, NIH: grants rmed.acponline.org
Vir Biotechnology: advisory board, stockholder covid19treatmentguidelines.nih.gov
Ivermectin trial grant: $155 million grant after the insufficient evidence recommendation trialsitenews.com (B)
Open Payments shows $2.4 million from AbbVie, $1.2 million from Gilead, $34,000 from Janssen, and $19,000 from Merck openpaymentsdata.cms.gov (E)
Prof. Andrew Pavia (andy.pavia@hsc.utah.edu)
GlaxoSmithKline: consultant rmed.acponline.org
Open Payments shows $14,000 from Pfizer, $5,000 from Merck, and $4,000 from Janssen openpaymentsdata.cms.gov (F)
SARS-CoV-2 infection and replication involves a complex interplay of 50+ host and viral proteins and other factors Lui, Lv, Malone, Murigneux, Niarakis, providing many therapeutic targets. Over 7,000 compounds have been predicted to reduce COVID-19 risk, either by directly minimizing infection or replication, by supporting immune system function, or by minimizing secondary complications. Figure 25 shows an overview of the results for ivermectin in the context of multiple COVID-19 treatments, and Figure 26 shows a plot of efficacy vs. cost for COVID-19 treatments.
Loading..
Figure 25. Scatter plot showing results within the context of multiple COVID-19 treatments. Diamonds shows the results of random effects meta-analysis. 0.6% of 7,092 proposed treatments show efficacy c19early.org (B).
Loading..
Loading..
Figure 26. Efficacy vs. cost for COVID-19 treatments.
Ivermectin is an effective treatment for COVID-19. Treatment is more effective when used early. Meta analysis using the most serious outcome shows 62% [51‑70%] and 85% [77‑90%] lower risk for early treatment and prophylaxis, with similar results for higher quality studies, primary outcomes, peer-reviewed studies, and for RCTs. Statistically significant lower risk is seen for mortality, ventilation, ICU admission, hospitalization, recovery, cases, and viral clearance. All remain significant for higher quality studies. 61 studies from 55 independent teams in 25 different countries show statistically significant improvements. Results are very robust — in worst case exclusion sensitivity analysis 61 of 101 studies must be excluded to avoid finding statistically significant efficacy.
Optimal use of ivermectin may involve synergy with combined treatments, administration taking into account the lipophilic nature, and sublingual, spray, or inhaled formulations for direct treatment to the respiratory tract. Pharmacokinetics show significant inter-individual variability Guzzo. Injectable formulations may reduce variability and provide much faster onset of action Mountain Valley MD. Liposomal formulations show increased antiviral activity and lower cytotoxicity Croci. Synergistic results are seen with polytherapy De Forni, Jitobaom, Jitobaom (B). Efficacy varies depending on the manufacturer Williams, underdosed and contaminated tablets are common Vanhee, and fake tablets with no active ingredient have been reported Yemeke.
As Scott Alexander says: "if you say anything in favor of ivermectin you will be cast out of civilization and thrown into the circle of social hell reserved for Klan members and 1/6 insurrectionists. All the health officials in the world will shout 'horse dewormer!' at you and compare you to Josef Mengele." The extreme politicization means we can only evaluate the data directly.
With 101 controlled studies, 48 RCTs, and extensive supporting evidence, few people have the time and experience to analyze all or most of the evidence. However, the PRINCIPLE trial Hayward provides a single large trial showing efficacy. Despite being perhaps the most biased trial, with extreme bias against showing efficacy in the design, operation, analysis, and reporting, authors failed to eliminate the efficacy and hid the results for 600 days (810 days from the expected announcement time, which resulted in a pause and continuation with even greater bias).
PRINCIPLE showed 36% lower ongoing persistent COVID-19 specific symptoms, p<0.0001 Hayward, and the primary recovery outcome shows superiority of ivermectin with significantly faster recovery and a probability of superiority > 0.999. While authors make an unprecedented claim that the results are not clinically relevant, 2 days faster recovery and 36% lower long COVID are both highly clinically relevant. Moreover, from all 66 treatments we cover, improved recovery is very significantly associated with lower mortality, p < 0.0000000000001. These highly positive results are despite very late treatment, a relatively low-risk population, and poor administration — from analysis of all trials, we expect significantly better results with recommended usage.
Loading..
Figure 27. Improved recovery is associated with lower mortality. Meta-regression across 66 treatments.
Table 8. Summary of widely discussed ivermectin RCTs.
PRINCIPLE36% lower ongoing persistent COVID-19 specific symptoms, p<0.0001 Hayward. The primary outcome shows superiority of ivermectin with significantly faster recovery and a probability of superiority > 0.999.
TOGETHER"There is a clear signal that IVM works in COVID patients.. that would be significant if more patients were added.." - co-principal investigator Reis.
ACTIV‑699%, 98%, 97% superiority for time unwell, progression @14,7 days Naggie (note: the clinical progression results were modified in the journal version, with no explanation for over 510 days, and the 600µg/kg results also differ without explanation).
COVID‑OUT61% lower hospitalization (ivermectin vs. placebo, unreported) Bramante. Authors detail why the hypoxemia results are unusable, however analysis of the data shows that the ER results are similarly uninformative and do not appear to be related to symptoms.
Popp, Roman provide meta analyses that show positive effects without reaching statistical significance. The primary methods used that result in a lack of statistical significance are the exclusion of the majority of the evidence base, and division of the remaining subset. For more details see the study notes.
We use fixed pre-specified effect extraction to avoid bias and to focus on the most clinically relevant results. For comparison, we have also performed analysis using the primary outcome of studies (shown in the supplementary data), with results showing similar effect sizes. Prophylaxis results are very similar with 100% (17 of 17) positive effects. Early treatment shows 84% (32 of 38) positive effects, improved due to the very small event count negative serious outcomes in Krolewiecki, Vallejos, and Buonfrate no longer having priority. Late treatment shows 70% (32 of 46) positive effects, reduced slightly, primarily due to viral clearance results being the primary outcome in some studies, and viral clearance being less successful with late treatment. Overall, the primary outcome analysis shows 80% (81 of 101) positive effects, compared to 82% (83 of 101) in the main protocol analysis.
All meta analyses combine heterogeneous studies, because all studies differ in one or more ways, including patient demographics, treatment delay distribution, effect measured, SARS-CoV-2 variants, and treatment regimens (note that this is different to heterogeneity caused by bias). Combining heterogeneous studies may obscure efficacy - for example if treatment within 24 hours is twice as effective as treatment within 48 hours and we include studies with later treatment; or if a treatment is effective at reducing mortality but has no effect on viral clearance and we include viral clearance studies. Including studies that are further from the optimal treatment situation will reduce the observed effect size. This can be seen in the treatment delay analysis - late treatment is less effective and including late treatment studies lowers the effect size. For any negative meta analysis, we must consider if the treatment is effective but only in a subset of the situations covered by the studies (or a situation not covered by any study, for example few treatments have studies with a treatment delay <= 24 hours).
The team referenced in this article is unreliable. Dr. Kyle Sheldrick posted a schedule A statement as a result of a defamation lawsuit admitting to false and defamatory claims regarding one of the world's most highly published and respected critical care physicians Sheldrick; Dr. Nick Brown has called for trials for crimes against humanity for scientists that "tried to scam the world with their fake treatments" including vitamin D c19early.org (C), twitter.com (B). Gideon Meyerowitz-Katz has posted many false claims detailed below c19ivm.org (B).
Update: authors indicated that their data would be available "soon" as of Sep 14, 2021, however it has not been released over 910 days later twitter.com (C), therefore it is not possible to analyze their methods regarding ivermectin research in detail. However, Dr. Sheldrick posted false and defamatory accusations regarding a highly respected physician that has saved countless lives before and during the pandemic. In this case, detailed methods were published, revealing highly flawed analysis, and a basic misunderstanding of statistics, as detailed by multiple statisticians Fenton, twitter.com (D). Author deleted the blog post. Further, we note that the team's disregard for major issues with the Together Trial, ACTIV-6, López-Medina et al., and Beltran Gonzalez et al. suggest substantial bias.
A BBC article raises questions due to data issues in some studies, based on an analysis from a team of researchers. One of the researchers reports that data in some trials could have been manipulated, while noting that human error can not be ruled out. Others in the team directly accuse authors of malfeasance. Regardless of the cause, concern over these studies is valid. Currently, 2 studies have been retracted, one was withdrawn by a preprint server, and another has been reported as pending retraction, although the journal reports that no retraction is pending. None of these studies are in our analysis.
Existence of some lower quality studies is typical in large evidence bases. The percentage of studies with issues is not greater than reported averages, and is not close to removing evidence of efficacy (and may actually improve evidence as detailed below). We performed an absolute worst case sensitivity analysis, where positive studies are excluded in order of the effect size, with the largest effect first. 60%, or 61 of 101 studies must be excluded to avoid finding statistically significant efficacy (this is in addition to the four papers not in this analysis).
The summary statistics from meta analysis necessarily obscure most of the information in the evidence base. For those that have read all of the research, knowledge of efficacy is supported by extensive additional information, including for example relationships between outcomes within a study, dose-response relationships within and across studies, treatment delay-efficacy relationships within and across studies, variant-efficacy relationships, etc. Notably, removal of Elgazzar, Samaha, and Niaee improve the treatment delay-efficacy and dose-response relationships and may further increase confidence when considering all information.
Concerns about Cadegiani, Carvallo, Carvallo (B), Carvallo (C) have also been reported. All of these studies are excluded in our exclusion analysis.
Studies Prophylaxis Early treatment Late treatment PatientsAuthors
With GMK/BBC exclusionsw/GMK/BBC exclusions 68 84% [73‑91%] 69% [61‑76%] 46% [30‑59%] 124,855 766
RCTs w/GMK/BBC exc.RCTs w/GMK/BBC exc. 37 89% [57‑97%] 66% [54‑75%] 31% [11‑46%] 12,392 468
Percentage improvement with ivermectin treatment after exclusion of all studies reported by this team
We note that, while malfeasance cannot be ruled out, reported concerns may also be caused by typos, data collection errors not affecting analyzed outcomes, and expected results from multiple tests. Authors, without any prior registration or statistical analysis plan, perform thousands of statistical tests across data in the studies and report results without correcting for multiple tests. For example, reporting the occurrence of a 1 in 1,000 event as evidence of randomization failure, while performing more than this number of tests across studies.
This group often dismisses studies based on an arbitrary statistical significance threshold for a specific outcome, a misunderstanding of statistics Amrhein, and indefensible as a pre-filter in meta analysis.
This group has made many claims unsupported by the data. For Niaee, one author claimed the study "made a HUGE difference". It has no effect on early treatment or prophylaxis. For late treatment, which is not recommended, the change was relatively minor. For Elgazzar, the author claimed that it could be "the most consequential medical fraud ever committed". There was almost no difference in our analysis after removing this paper (excluding 1 of 104 studies has very little effect, and the exclusion actually improves the treatment delay-response relationship).
Statements by the group suggest significant bias. The main author first referred to ivermectin as "something else to debunk" in December 2020, and later as a "horse dewormer". Another group member has called for charging scientists that recommend vitamin D with "crimes against humanity".
The group has made claims about all ivermectin evidence based on the existence of some studies with issues. It is inappropriate to generalize about the entire group of 1129 scientists and researchers based on the mistakes or actions of a few individuals.
This group has focused on finding issues in papers reporting large positive effects, which introduces a significant bias. Notably, the few studies that contribute most to minimizing the effects in meta analysis include studies with very high conflicts of interest and many reported protocol violations and data issues, however this group disregards all of these issues.
The article claims "The largest and highest quality ivermectin study published so far is the Together trial" which "found no benefit", however this study has not been published, is one of the lowest quality trials with many documented design, execution, and analysis issues, has extremely high conflicts of interest, there is a history of inaccurate reporting prior to publication for a previous treatment in the same trial, and the trial actually reported 18% lower mortality (not statistically significant).
The article reports that 26 studies were examined, however there are 104 studies, authors have not reported their results for all 26, and authors have not provided their data after repeated requests. Currently they have not even provided a list of the 26 studies.
The group has an excessive focus on RCTs, which have a fundamental bias against finding an effect for interventions like ivermectin that are widely known and easily available — patients that believe they need treatment are more likely to decline participation and take the treatment Yeh (this does not apply to the typical pharmaceutical trial of a new drug that is otherwise unavailable and unfamiliar).
The main author of the group is also against vitamin D. Of the 119 vitamin D COVID-19 treatment studies, author suggests only one trial is worth looking at Murai. This gives us a simple case to examine potential bias. Murai is a small trial providing no statistically significant effects (mortality p = 0.43, other outcomes are positive while also not significant). Author acknowledges that the trial is too small for a conclusion. More importantly, this trial provides no information about whether vitamin D reduces the risk of a serious COVID-19 case, because the patients in this trial already had a serious COVID-19 case (90% already on oxygen treatment at baseline). Author does not mention this. The trial also has poorly matched arms in terms of gender, ethnicity, hypertension, diabetes, and baseline ventilation, all favoring the control group. Further, this study uses an inappropriate form of vitamin D — cholecalciferol. In reality physicians would use calcifediol or calcitriol with late stage treatment, because they avoid a very long delay for conversion. We are unaware of a reason to use cholecalciferol in this case (other than to produce a null result). In summary, author's chosen study is the study providing the least useful information from the 119 vitamin D treatment studies to date, suggesting biased analysis.
We fully support this team's effort to clean up the evidence base. This is extremely valuable and improves the integrity of the evidence base (and the accuracy if done equally for all studies). We hope this or other teams can do the same for all treatments. However the analysis plan should be published, details of all tests should be provided, results should be corrected for multiple testing, results for all studies and tests should be provided, and equal attention should be given to studies with non-statistically significant results, especially those with major reported data issues that have been disregarded by this team (for example data suggesting substantial protocol violations including confounding by time in Reis and control arm use of treatments in López-Medina).
For coverage of other errors in the BBC article, and illumination of the stark contrast between Dr. Lawrie's response to the BBC before publication and what they chose to report, see BiRD Group, Campbell, Elijah, Lawrie (B).
More details can be found in the following response regarding the main author of this group.
On September 6, 2021, author is on video saying "what this web site does is it goes through all of those 10 things, finds the positive one, and only reports the positive one ... they'll pick a number that is the lowest number" youtube.com @46:48. This is false, we report individual outcome analyses, and for pooled results we use the most serious outcome with a detailed protocol. Notably, author knows this is false, having posted the protocol on Twitter a month earlier. Moreover, in most cases there could be no ambiguity on the most serious outcome even without a detailed protocol. At the time of author's statement, the most serious outcome was actually the worst of multiple reported outcomes 26% of the time.
Author indicated their data would be released "soon" on Sep 14, 2021. The data was used for conclusions in the BBC and BMJ along with a message that data transparency is needed. No data has been released over 910 days later twitter.com (C).
TLDR
As a quick way to assess reliability, consider that author describes the Together Trial as "incredibly well-done" and a masterpiece of science". The trial actually reported multiple impossible numbers, has refused to release data despite pledging to do so, has extreme conflicts of interest, and had blinding failure, randomization failure, and multiple protocol violations as detailed below. Author similarly disregards major issues with many other trials, including ACTIV-6, López-Medina et al., and Beltran Gonzalez et al. This suggests substantial bias.
Incorrect, misleading, hyperbolic, and unsupported statements have been made by an influential anti-treatment Twitter personality, journalist, and PhD student known for defending Monsanto Roundup against carcinogenic claims (later settled for $US 11 billion). Author is notable as the only known researcher that reports having read a majority of the 104 (including retracted) studies, but does not find the evidence to be positive. However, their opinion appears to have been formed before reading the studies — they first referred to ivermectin as "something else to debunk". We note that the author has made valuable contributions identifying significant issues with some studies, which has helped to improve the quality of the ivermectin evidence base, and has improved the dose-response and treatment delay-response relationships.
Analysis with GMK's recommended exclusions can be found in the supplementary data, which shows 46% [34‑56%] improvement, p = 0.000000015.
Author has been paid for writing anti-treatment articles, and has also referred to ivermectin as a "horse dewormer". Author has experienced personal tragedy with multiple family members having died of COVID-19, which may introduce a bias against acknowledging errors in treatment advice.
Author's attempt to discredit the scientists performing ivermectin research centers on the false assertion that excluding a small number of lower quality trials results in a negative outcome. It should be clear from the forest plot that this is not possible, but we can be more specific. We perform a worst case sensitivity analysis, where positive studies are excluded in order of the effect size, with the largest effect first. How many studies do we need to exclude before the meta analysis RR has a confidence interval exceeding 1.0? 60%, or 61 of 101 studies must be excluded to avoid finding statistically significant efficacy. As with all data in this paper, this analysis will automatically update as the evidence base evolves. Also note that this is after exclusion of withdrawn papers - one has never been in this analysis, the second was removed on the same day it was withdrawn, and the other two were removed in advance of retraction based on author's notification that retraction is pending (only one has been retracted, the journal for Niaee et al. has reported that no retraction is pending).
Author claims that we include several papers that are already excluded in the 10 exclusion analyses.
Author claims that there is a greater percentage of low quality studies for ivermectin and COVID-19 compared to other treatments. This is unsupported for such a large evidence base, and does not match previous studies.
Author often makes a basic error by equating positive effects that are not statistically significant at a specific level with "no effect", a misunderstanding of statistics Amrhein. For example, if a study reports 50% improvement with a p value of 0.1, we cannot say that the study shows the treatment is ineffective, or in the words of the author shows "no benefit at all". Author repeatedly makes false claims in this way.
On Sep 14, 2021, author indicated that their team had reviewed about 30 ivermectin studies and their data would be available soon, however it has not been released two years later.
Author appears to favor pharmaceutical company affiliated/operated trials. For example, the author has no problem with the lack of IPD for many pharmaceutical affiliated COVID-19 trials that support the author's treatment positions, yet considers the lack of IPD in a positive ivermectin trial to be problematic. Author believes the pharmaceutical affiliated Together Trial is the highest quality trial so far, yet not only have the authors declined to release IPD that they previously pledged to release, there was not even a preprint when GMK made the statement, and the trial has many critical and serious flaws, extremely high conflicts of interest, and a history of inaccurate reporting prior to publication for another treatment arm. GMK has subsequently published a paper with one of the original co-lead's of the Together Trial (who later joined the trial DSMC).
Author has disregarded treatment delay in analysis, which results in incorrect conclusions. For example, author claims that the RECOVERY trial proved that another treatment is not effective, and would provide definitive data if the same was done for ivermectin. The trial provided valuable data on very late use (9 days after symptoms) with an excessively high dose and very late stage patients. However, it did not provide information on early treatment. Oseltamivir, for example, is generally only considered effective for influenza when used within 0-36 or 0-48 hours McLean, Treanor. Paxlovid was tested with a maximum of 3 days from symptom onset (the mean delay is unknown). For ivermectin, author believes the PRINCIPLE trial will provide strong data on efficacy, however this trial includes low risk patients less than 15 days from symptom onset, and may only provide information on late treatment in a low risk population with lower risk variants. Figure 28 shows a mixed-effects meta-regression for efficacy as a function of treatment delay in COVID-19 studies from 66 treatments. Efficacy declines rapidly with treatment delay.
Late treatment results
are not representative
Figure 28. GMK believes that results for treatment delayed 9 days from symptom onset provides definitive information on treatment efficacy. However early treatment is critical for antivirals, as shown with antivirals for other respiratory diseases, and in meta-regression of studies from 66 COVID-19 treatments.
Author has an unwarranted focus on a specific outcome (mortality) and a specific subset of trials (RCTs). This would be reasonable in many cases when sufficient high-quality data is available, however this is not the case for off-patent COVID-19 treatment trials, where RCTs often involve delayed treatment, low-risk patients where mortality is rare, or very high conflicts of interest. Widely accepted and effective (for specific variants) treatments like casirivimab/imdevimab, bamlanivimab, and sotrovimab were all approved without statistically significant mortality benefits. Other outcomes are also important — accelerating viral clearance, and reducing cases, hospitalization, ICU admission, ventilation, etc. are all very valuable, for example reducing serious "long COVID" problems, reducing transmission of the virus, and reducing the burden on the healthcare system. These outcomes are also likely to correlate with reduced mortality among larger or higher-risk populations. We note that there is extensive evidence for the mortality outcome when not restricting to RCTs. RCTs have mostly been run with relatively low risk populations where mortality is low, leading to limited statistical significance. However RCTs are inherently biased towards low mortality and towards not finding an effect in this case — ivermectin is well-known to be beneficial for COVID-19 and is easily available, therefore participants that believe they may be at serious risk are more likely to decline participation in the RCT and take the recommended medications. Patients that do choose to participate are also more likely to have low adherence. This bias of RCTs is likely to be even larger in locations where ivermectin is widely used in the community and very easily obtained, which correlates with the observed RCT results.
Author suggests that we have chosen the wrong outcome in some cases. While mistakes are possible, for example we corrected errors with Espitia-Hernandez et al. and Jain et al., the claims made suggest that the author has not read the studies and/or our protocol carefully. Details are below. We note that the author disregards the existence of the individual outcome analyses and the primary outcome analysis.
Most errors have not been corrected by the author over two years later. Many false, misleading, and defamatory statements continue to be available, highly-ranked in search results, and highly influential. Other errors include:
that excluding Elgazzar et al. completely changes the results and could be "the most consequential medical fraud ever committed". Excluding 1 of 104 studies has very little effect, and the exclusion improves the treatment delay-response relationship.
that Niaee et al. "made a HUGE difference". It has no effect on early treatment or prophylaxis. For late treatment, which is not recommended, the change was minor, and the exclusion improves the treatment delay-response relationship.
making basic errors suggesting very superficial reading of studies, for example claiming the RR in Szente Fonseca is the risk of being treated.
making basic errors suggesting very superficial reading of this paper, for example claiming that a result for prophylaxis studies is based on the number of patients from all studies.
equating a high degree of COVID-19 in a country partially adopting a treatment with a lack of efficacy, disregarding obvious confounding such as heavily affected areas being more likely to adopt treatment (analysis of results in regions or time periods adopting treatment, while not equivalent to controlled studies, is more informative and shows efficacy Chamie-Quintero, Chamie-Quintero (B), Merino, Ontai).
confusing heterogeneity due to dose, treatment delay, etc. and due to bias.
disregarding treatment delay to dilute or obscure effects by including late treatment (author has also used this method with other treatments).
disregarding the existence of specific outcome analyses, RCT analysis, and exclusion-based sensitivity analysis.
suggesting that efficacy over longer periods is not possible because ivermectin has a half-life of "about a day". Author disregards known efficacy for other conditions over much longer periods, and mischaracterizes the half-life. Antiparasitic efficacy can persist for several months after a single dose Canga. Plasma half-life is longer in some studies, and significant plasma concentration can persist for over 2 weeks in some patients Muñoz. More importantly, ivermectin is highly lipophilic and may accumulate in the lung and other tissues where concentrations may be many times higher Chaccour (C), Chiu.
misunderstanding funnel plot analysis and explanations other than selective reporting (and providing no evidence of unreported negative studies, while there is substantial evidence of difficulty publishing positive studies Jerusalem Post, Kory (B)).
suggesting that it is not reasonable to combine evidence from mortality and hospitalization (for example), but happily combining late treatment and early treatment in order to obscure efficacy. If a treatment reduces disease severity requiring hospitalization, reduced mortality in at-risk populations logically follows, whereas lack of efficacy several days after onset can not be extrapolated to early treatment — treatments for a viral infection are often less effective when delayed.
making serious claims about individual studies without contacting authors (for example claiming patients were excluded for reaching the endpoint too quickly in one study, whereas authors report exclusions due to baseline negative status).
author is unaware of different variants, suggesting that results should be identical for treatment at a given delay, even when the predominant variants have markedly different peak viral load, time to peak viral load Faria, Karita, Nonaka, and mortality (for example Gamma vs. non-Gamma aHR 4.73 [1.15-19.41] Zavascki).
The cases where author suggests we have chosen the wrong outcome indicate that the author has not read the studies and/or our protocol carefully:
suggesting that the risk of a good outcome should be selectively used instead of the risk of a bad outcome (author would like to do this when it reduces the effect size). This is similar to using the risk of surviving instead of the risk of death. 99% survival may only be a 4% improvement over 95% survival, but most people would appreciate the 80% lower risk of death.
suggesting that hospitalization time should be used for symptomatic recovery in a study where discharge is based on viral clearance (and only tested weekly).
suggesting that a specific symptom such as cough should be used (author would prefer a less positive result for the study).
suggesting that viral load is more important than symptomatic results.
suggesting that mortality should be used in populations with zero mortality (for low-risk populations with no mortality, reduction in mortality is not possible, this does not mean a reduction in hospitalization, for example, is not valuable).
suggesting that unadjusted results should be used in a study where the adjustments clearly make a significant difference (author wants to cherry-pick unadjusted cough results).
suggesting that, for example, in a study of viral load where all patients recover, it is not valuable if treated patients recover faster (or are less likely to transmit the virus to others).
suggesting that study selected outcomes should have priority rather than using a consistent pre-specified protocol, disregarding the added bias and the fact that this actually improves results for ivermectin (for example the very small event count negative serious outcomes in Krolewiecki, Vallejos, and Buonfrate would no longer have priority).
suggesting that cough is a more important symptom than low SpO2 or fever. Cough can persist for a long time after more serious symptoms resolve, and persistent cough may be caused by many conditions.
suggesting that combined low dose treatment results should be used in a study that had a combined ivermectin/doxycycline arm (single dose ivermectin, 5 days doxycline) and an ivermectin arm with treatment for 5 days.
We note that this personality has an extensive history of incorrect advice, including for example:
claiming that flu is more dangerous than COVID-19
claiming that SARS-CoV-2 is not airborne
claiming that it's impossible to improve immune system functioning
even believing and propagating a made up story that claimed ivermectin overdose was causing gunshot victims to wait at an ER
Author has taken a public position against early treatments for COVID-19 since at least July 2020. Given this longstanding and influential negative position, they may tend to view information with a negative filter and confirmation bias, and may be reluctant to admit errors. They acknowledge not having read all of the studies (and appear to have very superficially read others). They submitted zero feedback to us, suggesting that they know their comments are incorrect or that they have a motivation other than correcting errors. Author claims that they could not contact us, however there are over 50 feedback links throughout this article. We also note that the author is not open to critical feedback and routinely blocks Twitter users correcting mistakes or expressing anything critical on their feed. Reports suggest that the author also pre-emptively blocks people that have not even interacted with them, but are connected to other users reporting on their errors. Author ackowledges using a tool called MegaBlock that blocks all people that liked a specific tweet.
The author is also against vitamin D. Of the 119 vitamin D COVID-19 treatment studies, author suggests only one trial is worth looking at Murai. This gives us a simple case to examine potential bias. Murai is a small trial providing no statistically significant effects (mortality p = 0.43, other outcomes are positive while also not significant). Author acknowledges that the trial is too small for a conclusion. More importantly, this trial provides no information about whether vitamin D reduces the risk of a serious COVID-19 case, because the patients in this trial already had a serious COVID-19 case (90% already on oxygen treatment at baseline). Author does not mention this. The trial also has poorly matched arms in terms of gender, ethnicity, hypertension, diabetes, and baseline ventilation, all favoring the control group. Further, this study uses an inappropriate form of vitamin D — cholecalciferol. In reality physicians would use calcifediol or calcitriol with late stage treatment, because they avoid a very long delay for conversion. We are unaware of a reason to use cholecalciferol in this case (other than to produce a null result). In summary, author's chosen study is one of least useful studies from the 119 studies to date, suggesting biased analysis.
Based on many comments, author appears to focus on superficial criteria such as typesetting and quality of writing. While many of the studies have been performed by non-native English speakers with minimal budgets, this does not imply the researchers are less reliable. Indeed, the author is highly critical of the program used to create a graph, for example, but is unable to see flaws in high budget high conflict of interest trials, even when they prompt >100 scientists to write an open letter requesting retraction Open Letter by 170+ US Doctors.
Two years later, the author has still not contacted us, making content-free comments on Twitter such as calling us "sh*tty". Other individuals pointing out errors with detailed and careful feedback get similar treatment, such as being called a "d*ckhead" and being blocked.
More details can be found in the BBC response.
Feb 2, 2023 update: Scott acknowledges that his analysis was incorrect, that the topic is outside his "expertise and competence level", and that his self-and-GidMK-filtered subset of a 29 study subset of 101 studies actually shows much stronger efficacy than previously claimed. Scott further acknowledges several false and defamatory claims, while standing by others. Scott acknowledges that the strongyloides theory is not very strong. Scott now relies on publication bias to discredit the 617 scientists reporting efficacy, without even reading many of their studies.
For publication bias, author is saying that the statistically significant positive studies were chance events, i.e., there are actually >1,000 ivermectin studies, but there is an extremely strong bias towards publishing only the positive studies.
This is in direct opposition to the evidence on publication bias. High profile journals refuse to publish positive ivermectin studies, while negative ivermectin studies are guaranteed acceptance, widespread press, and fame, just as Scott has received fame for his negative article. As Scott said in November 2021, "if you say anything in favor of ivermectin you will be cast out of civilization and thrown into the circle of social hell reserved for Klan members and 1/6 insurrectionists. All the health officials in the world will shout 'horse dewormer!' at you and compare you to Josef Mengele." Ivermectin has the strongest negative publication bias we have ever seen.
Scott finally points at our overview of efficacy across treatments, claiming that multiple effective treatments is proof of publication bias. As if somehow the existence of several effective treatments is impossible.
Scientists have identified over 2,000 compounds potentially beneficial for COVID-19. It is trivial to disable SARS-CoV-2 in vitro, hundreds of compounds do so, many with existing known and positive safety records. What is the chance that none of these can safely reach the location of SARS-CoV-2 infection? In the nasopharynx, oropharynx? Lung? Other tissues?
Is it really surprising that PVP-I for example, very effective in vitro, can be applied to the nasopharynx/oropharynx in effective concentrations, and is helpful, especially if used before infection spreads further?
Are there really no inputs to the human system that improve functioning of the immune system?
Does Scott really believe that paxlovid, casirivimab/imdevimab, sotrovimab, tixagevimab/cilgavimab, bebtelovimab, ensovibep, molnupiravir, and remdesivir are actually ineffective, all were approved in error because hundreds of other studies were not published?
While Scott dismisses the work of 9,330 scientists reporting statistically significant positive effects of low cost treatments (617 for ivermectin), he appears to have blind faith in trials having perhaps the highest moral and financial conflicts in the history of medicine, while simultaneously missing many critical issues [ACTIV-6, TOGETHER, COVID-OUT, PRINCIPLE] creating unreliable results, and signs of efficacy despite best efforts to avoid it.
Extreme moral conflicts arise for authors and institutions that played a critical role in the suppression of early treatments — they share responsibility for the results of any mistakes that impacted policies. The extreme financial conflicts come from authors and institutions associated with companies where billions of dollars of profit relies on no effective inexpensive treatments existing.
We challenge Scott to do a more serious analysis. For example: examine all of the studies rather than only 29%; directly examine studies rather than relying on a personality that thinks a study treating patients 9 days after onset is conclusive (a study so extremely late that 17% were on ventilation/ECMO); take into account major issues with studies like ACTIV-6, TOGETHER, COVID-OUT, and PRINCIPLE; and review the ~100 additional supporting papers for understanding of the mechanisms of action, censorship, and other issues.
Previous response: For a much better and more thorough analysis of Scott Alexander's essay, including numerous errors, extreme bias, incorrect statistical analysis, contradictory standards, and failure to make corrections, see the Potemkin Argument series. Part 21 illustrates the potentially catastrophic result of humanity's difficulty in objectively creating and evaluating scientific evidence related to critical world problems.
The analysis by SSC / Scott Alexander has a number of major issues. For many other uncorrected errors in SSC's analysis, see doyourownresearch.substack.com, twitter.com (E).
Analysis with SSC's recommended exclusions can be found in the supplementary data.
Update: after exclusions chosen by SSC, exclusions by GMK, excluding all late treatment, and excluding all prophylaxis studies, SSC found the results in Figure 29, showing statistically significant efficacy of ivermectin with p = 0.04. The method for computing this p value is not specified. We used the same event results and performed random-effects inverse variance DerSimonian and Laird meta analysis as shown in Figure 30, finding much higher significance with p = 0.005. We note that the effect extraction appears biased against ivermectin, choosing the excessive dose arm in Buonfrate, and using the post-hoc exclusion in López-Medina. The Krolewiecki treatment count appears to have been scaled for the different group sizes. We do not know where SSC's Mahmud counts are from.
Figure 29. SSC's analysis. SSC reports p = 0.04, with an unspecified method.
Figure 30. Random-effects meta analysis per SSC's chosen results, finding much higher significance.
Author appears to be against all treatments, labeling them all "unorthodox" and "controversial", even those approved by western health authorities, including casirivimab/imdevimab, bamlanivimab, sotrovimab, and paxlovid. Update: author's original article still refers to all treatments we follow as unorthodox and controversial, however they report that they actually recommend fluvoxamine, paxlovid, casirivimab/imdevimab, bamlanivimab/etesevimab, and sotrovimab, and suggest that they support all western health authority approved treatments which additionally includes remdesivir, budesonide, bebtelovimab, tixagevimab/cilgavimab, and molnupiravir. Author also has positive comments for zinc (but reports there is no proof). i.e., author appears to actually support at least 11 of the 66 treatments we follow. We note that the methodology is the same for all treatments.
We encourage the author to at least direct readers to government approved treatments, for which there are several in the author's country, and many more in other countries (including ivermectin). While approved treatments in a specific country may not be as effective (or as inexpensive) as current evidence-based protocols combining multiple treatments, they are better than dismissing everything as "unorthodox". Elimination of COVID-19 is a race against viral evolution. No treatment, vaccine, or intervention is 100% available and effective for all variants — we need to embrace all safe and effective means.
The third-party analysis that author references for the strongyloides theory is confounded by treatment delay and dosage — the high prevalence group has more early treatment trials and a higher average dose, i.e., the analysis reflects the greater efficacy of early treatment and the greater efficacy of higher dosage. More details can be found in the strongyloides section.
Author refers to studies with positive but not statistically significant results as "negative" Mohan, or "[the] original outcome would also have shown ivermectin not working" López-Medina, which are incorrect conclusions Amrhein. Update: author believes this means we abandon statistical significance. We do not know where this comes from — all of our results report confidence intervals, and the first two words of this paper are "statistically significant". What is incorrect is making a negative conclusion based on an insignificant result. For example, if one study reports 50% lower mortality without reaching statistical signifiance, this does not mean that the treatment is useless. Consider if there are 10 studies all reporting ~50% lower mortality, the combined evidence may be strong even if each individual result is not statistically significant.
Author notes that: "if you say anything in favor of ivermectin you will be cast out of civilization and thrown into the circle of social hell reserved for Klan members and 1/6 insurrectionists", suggesting an environment that may bias the information that the author sees, and could unconsciously bias analysis. We note that similar environments influence the design, operation, and publication of some existing (and many upcoming) ivermectin trials.
Author looks at 29 of the 101 studies, which we note is much better than most commenters, but still ignores the majority of studies, including the prophylaxis studies.
The author finds efficacy at p = 0.04 in their analysis of 11 of the 29 studies they looked at. We note that simply looking at the other 72 studies will result in much higher confidence in efficacy. We also note that even at p = 0.04 with 11 independent studies, a rational risk-benefit analysis results in immediate adoption into protocols (pending stronger data with other combinations of treatments), and immediate collection of more data from sources without conflicts of interest.
However, ultimately the author at least partially supports the two prevailing theories that are commonly used by those against treatment. These theories require disregarding extensive contradictory evidence:
The steps required to accept the no-significant-effect outcome are extreme — one needs to find a reason to exclude most of the studies, disregard the strong treatment-delay response relationship, and disregard all prophylaxis studies. Even after this, the result is still positive, just not statistically signficant. This does not support a negative recommendation. Widely accepted and effective (subject to dependence on viral variants) treatments like casirivimab/imdevimab, bamlanivimab, and sotrovimab were all approved without statistically significant mortality benefits.
The steps required to accept the strongyloides-mechanism-only conclusion are also extreme - we need to disregard the majority of outcomes occuring before steroid use, and disregard the strong treatment-delay response relationship which is contradictory. Figure 24 shows analysis by strongyloides prevalence. The third-party analysis referenced by the author is confounded by treatment delay and dosage.
Author seems biased against believing any large effect size. We note that large effect sizes have been seen in several COVID-19 treatments approved by western health authorities, including paxlovid which the author is very positive about, and also that better results may be expected when studies combine multiple effective treaments with complementary mechanisms of action (as physicians that treat COVID-19 early typically do). Update: author confirms this bias but appears to disregard it for paxlovid.
Author is suspicious about a study based on the country of the researchers, and also appears biased against non-native speakers, with comments such as "unreadable" for one paper, compared to "written up very nicely in real English" for another. Update: author confirms being biased against certain countries.
Author calls a physician that has reported zero deaths and 5 hospitalizations with 2,400 COVID-19 patients "a crazy person" that "put his patients on every weird medication he could think of".
Author disregards the dramatically higher mortality for Gamma vs non-Gamma variants (aHR 4.73 [1.15-19.41] Zavascki), instead concluding that higher mortality indicates fraud in one instance, while in another instance assuming that the related confounding by time in the Together Trial is not significant.
Author's review of the 29 studies appears relatively cursory, for example author appears unaware that the ivermectin dosage is very different in the ivermectin + doxycycline arm of Ahmed.
Author appears to accept the analysis and accusations of GMK as correct, however that author is often incorrect.
Author is concerned that we detail problems with López-Medina, while correctly noting that the outcomes in this trial are actually positive and in favor of ivermectin (while not statistically significant in isolation).
Author is concerned that we specifically comment on López-Medina, Reis. We note that it has been others that have focused on these trials — we comment on them because they have received special attention, including being held up as sole evidence overriding all other trials, despite having major issues.
Author claims that nobody can find issues with Vallejos, which suggests that they have not read the study, or our analysis.
This study was withdrawn and was removed from this analysis on the same day. There was no significant change (excluding 1 of 104 studies has very little effect, and the exclusion actually improves the treatment delay-response relationship).
This study was removed from this analysis within an hour of notification that it was pending retraction. There was no significant change in the results, and the exclusion improves the dose-response relationship.
This preprint was censored by the original preprint host. Censors claim that the government treatment program, which used approved medications and saved over 500 people from hospitalization, was unethical. In part they also indicate that studies of "the effects of a medication on a disease outcome" are outside the scope of their site.
The author's response (not provided by the censors) can be found here: twitter.com (F). Author's provide the data and code for the study, and the results have been independently verified.
This paper appears to have been censored at the request of the journal's founding editor pubmed.ncbi.nlm.nih.gov. An external review is mentioned but is not provided, and there is no reply from the authors, or indication that the authors were notified. Conclusions in this study are limited due to the small size, however we should consider all information in the context of the full body of research.
The conference publication for this analysis was self-censored by the authors, not due to any error in the analysis, but because authors believe ivermectin "has proven to be ineffective in clinical trials". This is incorrect, 61 studies show statistically significant positive results for one or more outcomes (34 prospective and 27 retrospective studies, including 24 Randomized Controlled Trials).
Concerns have been raised about Carvallo. There appears to be some valid concerns with potential data issues, and this study is excluded in the exclusion analysis. There is no significant change in results, with only a minor reduction in prophylaxis efficacy to 84% [73‑91%]. However, it is difficult to trust information from the personality reporting the concerns. The author suggests that the study may not have happened at all, claiming for example that the team could not have afforded the medications without funding, and that a busy clinician would not have enough time. However, with just basic checks, the author would know that a drug company has confirmed donating the medications, that they confirmed authorization for the study was received, that the main hospital for the study requested additional supplies, and that the hospital confirmed ethics committee approval. For additional details see O'Reilly. We also note that the combined treatment in this study has been independently shown to be effective, and the complementary mechanisms of action support improved efficacy of the combination Figueroa.
For discussion of all studies see c19ivm.org. A few studies have received special attention, with some considering them to be very strong evidence overriding the other 100 studies. We note limitations of these studies here.
Loading..









Loading..









Loading..









Loading..









Loading..









With only 7% hospitalization, this trial is underpowered. The trial primarily includes low-risk patients that recover quickly without treatment, leaving minimal room for improvement with treatment. 74 patients had symptoms for >= 7 days. Among the 7 patients requiring ventilation, authors note that the earlier requirement in the ivermectin group may be due to those patients having higher severity at baseline. However, authors know the answer to this - it is unclear why it is not reported. There were more adverse events in the placebo group than the ivermectin group, suggesting a possible issue with dispensing or non-trial medication usage. 25+% of patients were hospitalized within 2/3 days for the placebo/treatment groups (Figure S2).
The companion prophylaxis study IVERCOR PREP has reported results in the press and an online presentation lanacion.com.ar, twitter.com (BQ), however these results have not yet been formally published. The prophylaxis study results are very positive and statistically significant, and would be expected to receive priority publication due to the predicted impact on the pandemic and confirmation of previous prophylaxis studies. The lack of formal publication suggests a negative publication bias that may be due to politicization in the authors' location.
Authors pre-specify multivariate analysis but do not present it, however multivariate analysis could significantly change the results. Consider for example if just a few extra patients in the ivermectin group were in severe condition based on baseline SpO2. The lower mean SpO2 in the ivermectin group, and the shorter time to ventilation, are consistent with this being the case. Additionally, there are 14% more male patients in the ivermectin group.
An extremely large percentage of patients (55%) were excluded based on ivermectin use in the last 7 days. However, ivermectin may retain efficacy much longer (for example antiparasitic activity may persist for months Canga). A significant number of patients may also misrepresent their prior and future usage — the population is clearly aware of ivermectin, and patients with progressing disease may be motivated to take it, knowing that they may be in the control group. Another report states that 12,000 patients were excluded for recent use of ivermectin scidev.net).
RCTs have a fundamental bias against finding an effect for interventions that are widely available — patients that believe they need treatment are more likely to decline participation and take the intervention Yeh, i.e., RCTs are more likely to enroll low-risk participants that do not need treatment to recover (this does not apply to the typical pharmaceutical trial of a new drug that is otherwise unavailable). This trial was run in a community where ivermectin was very widely known and used.
For other issues see trialsitenews.com (K).
Another study reports results on a larger group of patients in the same hospital, showing ivermectin mortality RR 0.81 [0.53-1.24] Guzman.
Questions have been raised about this study and the early termination of the study and discontinuation of treatments, because the hospital statistics show a dramatically lower (~75%) case fatality rate during the period of the study web.archive.org (N) (data from gob.mx).
DateCasesDeathsCFR
3/20202150%
4/20204125%
5/20201318%
6/20203725%
7/20206558%
8/2020792329%
9/2020541222%
10/2020622134%
11/2020802633%
12/2020411332%
Several other inconsistencies have been reported Chamie.
Although the data from this study is reported to be available and has been shared with an anti-treatment group, independent researchers have been unable to obtain the data for verification Chamie, twitter.com (BR).
This meta analysis is designed to exclude most studies. Authors select a small subset of studies, with a majority of results based on only 1 or 2 studies. Authors split up studies which dilutes the effects and results in a lack of statistical significance for most outcomes. Authors perform 16+ meta analyses with very few studies in each analysis, and do not combine the evidence from all studies. However, we can consider the probability of the observed results across all outcomes.
Authors find positive results for 11 of 12 primary efficacy outcomes with events, or 16 of 18 including secondary outcomes. One of the primary outcomes and two of the secondary outcomes show statistically significant improvements in isolation. If we assume independence, the probability that 11+ of 12 primary efficacy outcomes were positive for an ineffective treatment is p = 0.003. For 16+ of 18 outcomes we get p = 0.0007. This simple analysis does not take into account the magnitude of positive effects, or the dependence due to some studies contributing multiple outcomes, however observation suggests that a full analysis of the combined evidence is likely to show efficacy.
The study is entirely retrospective in the current version. The protocol is dated April 20, 2021, and the most recent study included is from March 9, 2021. The protocol was modified after publication in order to include a close to null result (Beltran Gonzalez "patients discharged without respiratory deterioration or death at 28 days"), so the current protocol is dated July 28, 2021.
Authors excluded many studies by requiring results at a specific time, for example mortality, ventilation, etc. required results at exactly 28 days. Authors excluded all prophylaxis studies by requiring results at exactly 14 days.
Studies comparing with other medications were excluded, however these studies confirm efficacy of ivermectin. The only case where they could overstate the efficacy of ivermectin is if the other medication was harmful. There is some evidence of this for excessive dosage/very late stage use, however that does not apply to any of the studies here.
Studies using combined treatment were excluded, even when it is known that the other components have minimal or no effect. 3 of 4 RCTs with combined treatment use doxycycline in addition, which was shown to have no significant effect in Butler (B). Other studies were excluded by requiring PCR confirmation.
Authors are inconsistent regarding active comparators. They state that hydroxychloroquine “does not work”, yet excluded trials comparing ivermectin to a drug they hold to be inactive. On the other hand, remdesivir was an acceptable comparator, although it is considered to be effective standard of care in some locations Fordham.
Authors fail to recognize that Risk of Bias (RoB) domains such as blinding are far less important for the objective outcome of mortality.
Authors include Beltran Gonzalez as "moderate" COVID-19, however patients in this study were in severe condition (baseline SatO2 83).
Fordham summarizes several problems:
unsupported assertions of adverse reactions to ivermectin, and the outdated claim that unsafe dosing would be needed to be effective;
a demand for PCR or antigen testing, without analysis of reliability and not universally available even in developed countries at the start of the pandemic;
contradictions in the exclusion criteria, including placebo and approved SoC comparators, but rejecting hydroxychloroquine, though held to be ineffective (and an approved SoC in some jurisdictions);
inclusion of “deemed active” comparators whilst excluding “potentially active” ones;
exclusion of combination therapies, though the norm among practising clinicians;
the rejection of other than RCTs when the objective is a “complete evidence profile”;
arbitrary time-points for outcome measures, excluding non-compliant trials;
fragmentation of data by location of care under varying hospitalisation criteria;
the resulting focus on a small fraction of the available clinical evidence, with most comparisons based on single studies with no meta-analysis possible;
a resulting inpatient mortality comparison with fewer patients than a June 2020 confounder-matched study;
no conclusion on the headline mortality outcome, when multiple lines of evidence from elsewhere (including the WHO) point to significant mortality advantage.
Cochrane was reputable in the past, but is now controlled by pharmaceutical interests. For example, see the news related to the expulsion of founder Dr. Gøtzsche and the associated mass resignation of board members in protest blogs.bmj.com, bmj.com, en.x-mol.com. For another example of bias see ebm.bmj.com.
The BiRD group gave the following early comment: "Yesterday’s Cochrane review surprisingly doesn’t take a pragmatic approach comparing ivermectin versus no ivermectin, like in the majority of other existing reviews. It uses a granular approach similar to WHO’s and the flawed Roman et al paper, splitting studies up and thereby diluting effects. Consequently, the uncertain conclusions add nothing to the evidence base. A further obfuscation of the evidence on ivermectin and an example of research waste. Funding conflicts of interests of the authors and of the journal concerned should be examined."
For dicussion of issues added in the updated version see Popp (B).
Please submit updates and corrections at the bottom of this page.Please submit updates and corrections at https://c19ivm.org/meta.html.
3/9: Discussion updates.
2/29: We added Hayward.
2/23: RCT discussion updates.
2/12: We added Oranu.
1/24/2024: We updated the introduction.
12/27: We added Mikamo.
9/23: Preclinical updates.
8/10: Preclinical updates.
7/25: We added Osati.
7/17: We updated the introduction.
6/11: We added Llenas-García.
6/6: We added Wada.
4/21: We added Munir.
4/18: We updated Desort-Henin to the published version.
2/23: We updated Schilling to the journal version.
1/6/2023: We added Desort-Henin, Sarojvisut.
12/20: We updated the discussion of heterogeneity and RCTs.
12/15: We added the ACTIV-6 600µg/kg arm Naggie (B).
12/9: We updated de la Rocha to the journal version.
10/27: We added Ochoa-Jaramillo.
10/21: We updated the ACTIV-6 trial to the journal version.
9/23: We added Aref (B).
9/9: We added Qadeer.
8/18: We added Bramante.
7/26: We added Schilling.
6/24: We added Mirahmadizadeh.
6/16: We updated the ACTIV-6 analysis.
6/16: We added Rezai, Rezai (B).
6/12: We added Naggie.
6/1: We updated the Together Trial analysis.
5/30: We added George.
5/30: We updated the Together Trial analysis.
5/27: We added de la Rocha.
4/17: We added a section on preclinical research.
4/16: We added discussion of the NIH recommendation.
4/9: We updated the Together Trial analysis.
4/8: We added Ravikirti.
4/5: We added preprint discussion based on Zeraatkar, and updated the Together Trial analysis.
4/2: We updated the Together Trial analysis.
3/30: We updated Reis to the journal version.
3/21: Strongyloides discussion updates.
3/3: We updated Beltran Gonzalez to the journal version.
3/2: We added Soto.
2/28: We added Efimenko.
2/25: We added Thairu.
2/23: We updated Mayer to the journal version.
2/18: We updated Lim to the journal version.
2/2: We added Manomaipiboon.
1/21: We added Zubair.
1/17: We added an explanation of why funnel plot analysis is not valid in this case.
1/16: We added RCT viral clearance analysis and corrected missing symptomatic case results in the case analysis.
1/15: We updated Kerr to the journal version.
1/15: We corrected hospitalization group sizes in Buonfrate.
1/13: We added Abbas, Baguma.
1/11: We updated Kerr to the latest results, and added discussion of Beltran Gonzalez.
1/7/2022: We updated Buonfrate to the journal version, and we updated Kerr to the latest results.
12/31: We added Shimizu.
12/29: We added Mustafa.
12/26: We updated Kerr to the revised version of the paper.
12/16: We added Jamir.
12/11: We added Kerr.
12/8: We added analysis of the number of independent research groups reporting statistically significant positive results.
12/5: We added Ferreira.
12/5: We added Rezk.
12/3: A note on Bernigaud: continuity correction uses the reciprocal of the contrasting arm Sweeting, as detailed in the appendix. We previously limited the size of the control group when showing the total number of patients, however this was confusing for people that did not read the details, as discussed below. The full group size has always been used when computing the RR.
12/1: Strongyloides discussion updates.
11/30: We corrected Ghauri to use the event counts.
11/24: We added Ozer.
11/24: SSC discussion updates.
11/21: Strongyloides discussion updates.
11/20: Strongyloides discussion updates.
11/19: We added analysis by strongyloides prevalence, and updated it to match the revised classification used in the comparable analysis.
11/19: We added additional exclusion analyses in the supplementary data.
11/18: We incorrectly included López-Medina as a study not reporting use of steroids, however they report 6% usage in the control group.
11/18: We added Samajdar.
11/16: Discussion updates.
11/12: We now show the number of studies reporting statistically significant results for any outcome, primary outcomes, and the most serious outcome.
11/9: Discussion updates.
11/5: We added discussion of strongyloides, comparison with the recent molnupiravir approval, and notes on recruitment for remote outpatient delayed treatment trials.
11/3: We added Lim.
11/3: Discussion updates.
10/29: Discussion updates including GMK vitamin D analysis.
10/28: Discussion updates.
10/26: We updated the GMK response.
10/24: We added additional exclusion analyses for individual outcomes.
10/21: We added Borody.
10/19: Discussion updates.
10/18: Ghauri was updated to the journal version.
10/16: We added a summary plot for all results.
10/13: We added primary outcome analysis and additional exclusion analyses. Niaee et al. has been reported as pending retraction and has been removed. 10/27 update: the journal has reported that this is incorrect — no retraction is pending.
10/11: Discussion updates. Niaee et al. exclusion. Updates to the study notes including discussion of Vallejos et al. and additional issues in the Together Trial. Discussion of inherent bias in RCTs for widely available interventions.
10/8: Discussion updates.
10/7: Samaha et al. has been reported as pending retraction and has been removed. There was no significant change in the results.
10/4: Merck discussion updates.
9/29: We corrected a display error causing a few points to be missing in Figure 3.
9/27: We added Mayer.
9/24: We added a graph of variants over time for the Together Trial discussion and corrected outcome discussion for Popp et al.
9/22: Discussion updates.
9/20: Discussion updates.
9/18: We added Buonfrate, and updated discussion of the Together Trial.
9/17: We added study notes.
9/15: Discussion updates.
9/14: FDA discussion updates.
9/9: We added sensitivity analysis to compute the minimum number of studies that need to be excluded in order to avoid showing efficacy. Discussion updates.
9/7: Discussion updates.
9/6: We corrected Espitia-Hernandez to use the reported recovery time and added missing recovery and viral clearance results.
9/3: We updated discussion and excluded Carvallo et al. in the exclusion analysis.
8/27: We updated Morgenstern (B) with the journal version of the article.
8/26: We updated Mohan with the journal version of the article.
8/16: We updated Reis with event counts.
8/15: We updated discussion and made the abstract more concise.
8/12: We added Elavarasi, Reis.
8/8: We updated discussion in the responses.
8/6: We updated Behera (B) with the journal version of the article.
8/5: We added Mondal.
8/4: We added discussion of the FDA recommendation.
8/3: We added discussion in the responses section.
8/2: We added analysis restricted to serious outcomes and analysis restricted to recovery, and we added discussion in the responses section.
7/31: We added discussion in the responses section related to in vitro evidence and therapeutic concentrations.
7/29: We added discussion in the responses section.
7/20: We updated Hashim with the journal version of the article.
7/16: We updated Ravikirti (B) with the journal version of the article.
7/15: Elgazzar et al. was withdrawn by the preprint server and has been removed.
7/9: We added Hazan (B).
7/8: We updated Cadegiani to the journal version.
7/6: We previously limited the size of the control group for Bernigaud when calculating the total number of patients, however this was confusing for many people that did not read the details. We now show the original counts and note the larger size of the control group in the text.
7/3: We added Vallejos.
7/2: We updated Niaee et al. to the journal version.
6/21: We added more information to the abstract.
6/19: We updated Bryant to the journal version.
6/19: Beltran Gonzalez was incorrectly included in the peer-reviewed analysis.
6/18: We added Krolewiecki.
6/15: We added Aref.
6/7: We added Hariyanto.
6/5: We added Ahsan.
6/2: We added Abd-Elsalam.
5/31: Biber was updated to the preprint.
5/26: Samaha et al. was updated to the journal version.
5/18: We added analysis of Merck's recommendation.
5/17: We added Szente Fonseca.
5/15: We updated the discussion of the WHO analysis.
5/13: We updated Mahmud to the journal version.
5/10: We added Faisal.
5/10: We added additional information in the abstract.
5/8: We added Merino.
5/7: We updated Shahbaznejad to the journal version, which includes additional outcomes not reported earlier.
5/6: We updated Chahla to the Research Square preprint.
5/6: We added a comparison of CDC recommendations.
5/6: We added mechanical ventilation and ICU admission analysis.
5/6: We updated discussion based on peer review including discussion of heterogeneity, exclusion based sensitivity analysis, and search criteria.
5/5: We updated Okumuş to the journal paper.
5/5: We previously limited the size of the control group in Bernigaud to be the same as the treatment group for calculation of the total number of patients. This is now also reflected and noted in the forest plots.
5/4: We added Loue.
4/30: We added analysis of the WHO meta analysis and updated Kory to the journal version.
4/28: We added the WHO meta analysis results for comparison.
4/27: We added analysis restricted to hospitalization results and a comparison with the evidence base used in the approval of other COVID-19 treatments.
4/26: We added notes on heterogeneity.
4/25: We updated Biber to the latest results reported at the International Ivermectin for Covid Conference.
4/18: We updated Morgenstern to the preprint.
4/16: We added Morgenstern.
4/14: We added Seet.
4/10: We added Kishoria.
4/9: We corrected a duplicate entry for Bukhari.
4/7: We identified studies where minimal detail is currently available in the forest plots.
4/5: We added Mourya.
4/4: We added event counts to the forest plots.
3/31: We updated Chahla (B) to the preprint.
3/30: We added Chahla.
3/28: We highlighted and added discussion for studies that use combined treatments.
3/26: We added Tanioka.
3/25: We added Huvemek.
3/17: We added Nardelli.
3/12: We added Bryant, Roy.
3/10: We added Pott-Junior.
3/6: We added Chowdhury and we identify studies that compare with another treatment.
3/5: We added discussion of pooled effects (we show both pooled effects and individual outcome results).
3/4: We added López-Medina, and we added more information in the abstract.
3/3: We updated the graphs to indicate the time period for the dosage column, now showing the dosage over one month for prophylaxis and over four days for other studies.
3/2: We updated IVERCOR PREP with the latest results Vallejos (B).
2/27: We added analysis restricted to peer reviewed studies.
2/24: We added a comparison of the evidence base and WHO approval status for the use of ivermectin with scabies and COVID-19. We updated Okumuş with the Research Square preprint.
2/23: We added Beltran Gonzalez.
2/18: We updated Babalola to the journal version of the paper.
2/17: We added Elalfy, and we added analysis restricted to viral clearance outcomes, and mortality results restricted to RCTs.
2/16: We updated Behera to the journal version of the paper.
2/15: We added Behera (B).
2/14: We added analysis restricted to COVID-19 case outcomes, and we added additional results in the abstract.
2/12: We added Biber.
2/11: We added more details on the analysis of prospective vs. retrospective studies.
2/10: We added Lima-Morales.
2/5: We updated Bukhari to the preprint.
2/2: We added Mohan.
1/26: We updated Shouman with the journal version of the article.
1/25: We updated IVERCOR PREP with the recently released results.
1/19: We added Shahbaznejad and Samaha et al. Chaccour (B) was updated to the journal version of the paper.
1/17: We added Bukhari.
1/16: We moved the analysis with exclusions to the main text, and added additional commentary.
1/15: We added the effect measured for each study in the forest plots.
1/12: We added Okumuş.
1/11: We added Chahla (B).
1/10: We put all prophylaxis studies in a single group.
1/9: We added Ravikirti (B). Due to the much larger size of the control group in Bernigaud, we limited the size of the control group to be the same as the treatment group for calculation of the total number of patients.
1/7: We added direct links to the study details in the chronological plots.
1/6: We added Babalola.
1/5: We added direct links to the study details in the forest plots.
1/2/2021: We added dosage information and we added the number of patients to the forest plots.
12/31: We added additional details about the studies in the appendix.
12/29: We added meta analysis excluding late treatment.
12/27: We added the total number of authors and patients.
12/26: We added Carvallo (B), IVERCOR PREP.
12/17: We added Alam.
12/16: We added Ghauri.
12/11: We added Soto-Becerra.
12/7: We added Chaccour (B).
12/2: We added Ahmed.
11/26/2020: Initial revision.
We perform ongoing searches of PubMed, medRxiv, Europe PMC, ClinicalTrials.gov, The Cochrane Library, Google Scholar, Research Square, ScienceDirect, Oxford University Press, the reference lists of other studies and meta-analyses, and submissions to the site c19ivm.org, which regularly receives submissions of studies upon publication. Search terms are ivermectin and COVID-19 or SARS-CoV-2. Automated searches are performed twice daily, with all matches reviewed for inclusion. All studies regarding the use of ivermectin for COVID-19 that report a comparison with a control group are included in the main analysis. Sensitivity analysis is performed, excluding studies with major issues, epidemiological studies, and studies with minimal available information. This is a living analysis and is updated regularly.
We extracted effect sizes and associated data from all studies. If studies report multiple kinds of effects then the most serious outcome is used in pooled analysis, while other outcomes are included in the outcome specific analyses. For example, if effects for mortality and cases are both reported, the effect for mortality is used, this may be different to the effect that a study focused on. If symptomatic results are reported at multiple times, we used the latest time, for example if mortality results are provided at 14 days and 28 days, the results at 28 days have preference. Mortality alone is preferred over combined outcomes. Outcomes with zero events in both arms are not used, the next most serious outcome with one or more events is used. For example, in low-risk populations with no mortality, a reduction in mortality with treatment is not possible, however a reduction in hospitalization, for example, is still valuable. Clinical outcomes are considered more important than viral test status. When basically all patients recover in both treatment and control groups, preference for viral clearance and recovery is given to results mid-recovery where available. After most or all patients have recovered there is little or no room for an effective treatment to do better, however faster recovery is valuable. If only individual symptom data is available, the most serious symptom has priority, for example difficulty breathing or low SpO2 is more important than cough. When results provide an odds ratio, we compute the relative risk when possible, or convert to a relative risk according to Zhang (B). Reported confidence intervals and p-values were used when available, using adjusted values when provided. If multiple types of adjustments are reported propensity score matching and multivariable regression has preference over propensity score matching or weighting, which has preference over multivariable regression. Adjusted results have preference over unadjusted results for a more serious outcome when the adjustments significantly alter results. When needed, conversion between reported p-values and confidence intervals followed Altman, Altman (B), and Fisher's exact test was used to calculate p-values for event data. If continuity correction for zero values is required, we use the reciprocal of the opposite arm with the sum of the correction factors equal to 1 Sweeting. Results are expressed with RR < 1.0 favoring treatment, and using the risk of a negative outcome when applicable (for example, the risk of death rather than the risk of survival). If studies only report relative continuous values such as relative times, the ratio of the time for the treatment group versus the time for the control group is used. Calculations are done in Python (3.12.2) with scipy (1.12.0), pythonmeta (1.26), numpy (1.26.4), statsmodels (0.14.1), and plotly (5.19.0).
Forest plots are computed using PythonMeta Deng with the DerSimonian and Laird random effects model (the fixed effect assumption is not plausible in this case) and inverse variance weighting. Results are presented with 95% confidence intervals. Heterogeneity among studies was assessed using the I2 statistic. Mixed-effects meta-regression results are computed with R (4.1.2) using the metafor (3.0-2) and rms (6.2-0) packages, and using the most serious sufficiently powered outcome. Forest plots show simplified dosages for comparison, these are the total dose in the first four days for treatment, and the monthly dose for prophylaxis, for a 70kg person. For full dosage details see below. For all statistical tests, a p-value less than 0.05 was considered statistically significant. Grobid 0.8.0 is used to parse PDF documents.
We have classified studies as early treatment if most patients are not already at a severe stage at the time of treatment (for example based on oxygen status or lung involvement), and treatment started within 5 days of the onset of symptoms. If studies contain a mix of early treatment and late treatment patients, we consider the treatment time of patients contributing most to the events (for example, consider a study where most patients are treated early but late treatment patients are included, and all mortality events were observed with late treatment patients). We note that a shorter time may be preferable. Antivirals are typically only considered effective when used within a shorter timeframe, for example 0-36 or 0-48 hours for oseltamivir, with longer delays not being effective McLean, Treanor.
We received no funding, this research is done in our spare time. We have no affiliations with any pharmaceutical companies or political parties.
Note that the size of the control group in Bernigaud is significantly larger than the treatment group. We previously limited the size to be the same as that of the treatment group for calculation of the number of patients, however this was confusing to many people that did not read the details.
A summary of study results is below. Please submit updates and corrections at the bottom of this page.
A summary of study results is below. Please submit updates and corrections at https://c19ivm.org/meta.html.
Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. Only the first (most serious) outcome is used in pooled analysis, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.
Abbas, 12/31/2021, Double Blind Randomized Controlled Trial, placebo-controlled, China, peer-reviewed, 3 authors, study period May 2021 - August 2021, dosage 300μg/kg days 1-5, excluded in exclusion analyses: very minimal patient information, three different results for the recovery outcome, selective omission of the statistically significant recovery p-value, and other inconsistencies. risk of death, 4.0% higher, RR 1.04, p = 1.00, treatment 1 of 99 (1.0%), control 1 of 103 (1.0%).
deterioration of 2 or more points, 40.5% lower, RR 0.59, p = 0.54, treatment 4 of 99 (4.0%), control 7 of 103 (6.8%), NNT 36.
escalation of care, 14.9% lower, RR 0.85, p = 0.82, treatment 9 of 99 (9.1%), control 11 of 103 (10.7%), NNT 63.
fever during study, 17.9% lower, RR 0.82, p = 0.58, treatment 15 of 99 (15.2%), control 19 of 103 (18.4%), NNT 30.
risk of no recovery, 35.6% lower, RR 0.64, p = 0.04, treatment 26 of 99 (26.3%), control 42 of 103 (40.8%), NNT 6.9, primary outcome.
recovery time, 30.8% lower, relative time 0.69, p = 0.08, treatment 99, control 103, primary outcome.
Ahmed, 12/2/2020, Double Blind Randomized Controlled Trial, Bangladesh, peer-reviewed, mean age 42.0, 15 authors, average treatment delay 3.83 days, dosage 12mg days 1-5, the ivermectin + doxycycline group took only a single dose of ivermectin. risk of unresolved symptoms, 85.0% lower, RR 0.15, p = 0.09, treatment 0 of 17 (0.0%), control 3 of 19 (15.8%), NNT 6.3, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), day 7, fever, ivermectin (5 days), primary outcome.
risk of unresolved symptoms, 62.7% lower, RR 0.37, p = 0.35, treatment 1 of 17 (5.9%), control 3 of 19 (15.8%), NNT 10, day 7, fever, ivermectin (1 day) + doxycycline.
risk of no viral clearance, 75.6% lower, HR 0.24, p = 0.03, treatment 11 of 22 (50.0%), control 20 of 23 (87.0%), NNT 2.7, adjusted per study, inverted to make HR<1 favor treatment, day 7, ivermectin (5 days).
risk of no viral clearance, 56.5% lower, HR 0.43, p = 0.22, treatment 16 of 23 (69.6%), control 20 of 23 (87.0%), NNT 5.8, adjusted per study, inverted to make HR<1 favor treatment, day 7, ivermectin (1 day) + doxycycline.
risk of no viral clearance, 63.0% lower, HR 0.37, p = 0.02, treatment 5 of 22 (22.7%), control 14 of 23 (60.9%), NNT 2.6, adjusted per study, inverted to make HR<1 favor treatment, day 14, ivermectin (5 days).
risk of no viral clearance, 41.2% lower, HR 0.59, p = 0.19, treatment 9 of 23 (39.1%), control 14 of 23 (60.9%), NNT 4.6, adjusted per study, inverted to make HR<1 favor treatment, day 14, ivermectin (1 day) + doxycycline.
time to viral-, 23.6% lower, relative time 0.76, p = 0.02, treatment 22, control 23, ivermectin (5 days).
time to viral-, 9.4% lower, relative time 0.91, p = 0.27, treatment 23, control 23, ivermectin (1 day) + doxycycline.
Aref, 6/15/2021, Randomized Controlled Trial, Egypt, peer-reviewed, 7 authors, study period February 2021 - March 2021, dosage not specified, trial NCT04716569 (history). relative duration of fever, 63.2% lower, relative time 0.37, p < 0.001, treatment 57, control 57, primary outcome.
relative duration of dyspnea, 56.4% lower, relative time 0.44, p < 0.001, treatment 57, control 57.
relative duration of anosmia, 68.8% lower, relative time 0.31, p < 0.001, treatment 57, control 57.
relative duration of cough, 64.3% lower, relative time 0.36, p < 0.001, treatment 57, control 57.
risk of no viral clearance, 78.6% lower, RR 0.21, p = 0.004, treatment 3 of 57 (5.3%), control 14 of 57 (24.6%), NNT 5.2.
time to viral-, 35.7% lower, relative time 0.64, p < 0.001, treatment 57, control 57.
Babalola, 1/6/2021, Double Blind Randomized Controlled Trial, Nigeria, peer-reviewed, baseline oxygen required 8.3%, 10 authors, study period May 2020 - November 2020, dosage 12mg or 6mg q84h for two weeks, this trial compares with another treatment - results may be better when compared to placebo. adjusted risk of viral+ at day 5, 63.9% lower, RR 0.36, p = 0.11, treatment 40, control 20, adjusted per study, inverted to make RR<1 favor treatment.
relative ∆SpO2 (unadjusted), 41.5% better, RR 0.59, p = 0.07, treatment 38, control 18, figure 3.
risk of no viral clearance, 58.0% lower, HR 0.42, p = 0.01, treatment 20, control 20, inverted to make HR<1 favor treatment, 12mg - Cox proportional hazard model.
risk of no viral clearance, 40.5% lower, HR 0.60, p = 0.12, treatment 20, control 20, inverted to make HR<1 favor treatment, 6mg - Cox proportional hazard model.
time to viral-, 49.2% lower, relative time 0.51, p = 0.02, treatment 20, control 20, 12mg, primary outcome.
time to viral-, 34.4% lower, relative time 0.66, p = 0.08, treatment 20, control 20, 6mg.
Biber, 2/12/2021, Double Blind Randomized Controlled Trial, placebo-controlled, Israel, peer-reviewed, 10 authors, study period 12 May, 2020 - 31 October, 2020, average treatment delay 4.0 days, dosage 12mg days 1-3, 15mg for patients ≥70kg, trial NCT04429711 (history). risk of hospitalization, 70.2% lower, RR 0.30, p = 0.34, treatment 1 of 47 (2.1%), control 3 of 42 (7.1%), NNT 20.
risk of no viral clearance, 61.6% lower, RR 0.38, p = 0.02, treatment 8 of 47 (17.0%), control 17 of 42 (40.5%), NNT 4.3, adjusted per study, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, Ct>30, multivariable, day 8.
risk of no viral clearance, 39.0% lower, RR 0.61, p = 0.09, treatment 13 of 47 (27.7%), control 21 of 42 (50.0%), NNT 4.5, adjusted per study, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, Ct>30, multivariable, day 6, primary outcome.
risk of no viral clearance, 73.0% lower, RR 0.27, p = 0.008, treatment 3 of 23 (13.0%), control 14 of 29 (48.3%), NNT 2.8, culture viability.
risk of no viral clearance, 70.2% lower, RR 0.30, p = 0.14, treatment 2 of 47 (4.3%), control 6 of 42 (14.3%), NNT 10.0, non-infectious samples (Ct>30 or non-viable culture), day 10.
risk of no viral clearance, 82.1% lower, RR 0.18, p = 0.01, treatment 2 of 47 (4.3%), control 10 of 42 (23.8%), NNT 5.1, non-infectious samples (Ct>30 or non-viable culture), day 8.
risk of no viral clearance, 75.6% lower, RR 0.24, p = 0.02, treatment 3 of 47 (6.4%), control 11 of 42 (26.2%), NNT 5.0, non-infectious samples (Ct>30 or non-viable culture), day 6.
risk of no viral clearance, 65.1% lower, RR 0.35, p = 0.05, treatment 4 of 28 (14.3%), control 9 of 22 (40.9%), NNT 3.8, non-infectious samples (Ct>30 or non-viable culture), day 4.
risk of no viral clearance, 51.9% lower, RR 0.48, p = 0.08, treatment 7 of 47 (14.9%), control 13 of 42 (31.0%), NNT 6.2, Ct>30, day 10.
risk of no viral clearance, 57.9% lower, RR 0.42, p = 0.02, treatment 8 of 47 (17.0%), control 17 of 42 (40.5%), NNT 4.3, Ct>30, day 8.
risk of no viral clearance, 44.7% lower, RR 0.55, p = 0.049, treatment 13 of 47 (27.7%), control 21 of 42 (50.0%), NNT 4.5, Ct>30, day 6.
risk of no viral clearance, 31.9% lower, RR 0.68, p = 0.16, treatment 13 of 28 (46.4%), control 15 of 22 (68.2%), NNT 4.6, Ct>30, day 4.
Borody, 10/19/2021, retrospective, Australia, preprint, 2 authors, study period 1 June, 2021 - 30 September, 2021, dosage 24mg days 1-10, this trial uses multiple treatments in the treatment arm (combined with zinc and doxycycline) - results of individual treatments may vary, excluded in exclusion analyses: preliminary report with minimal details. risk of death, 92.3% lower, RR 0.08, p = 0.03, treatment 0 of 600 (0.0%), control 6 of 600 (1.0%), NNT 100, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of hospitalization, 92.9% lower, RR 0.07, p < 0.001, treatment 5 of 600 (0.8%), control 70 of 600 (11.7%), NNT 9.2, primary outcome.
Bramante, 8/18/2022, Double Blind Randomized Controlled Trial, placebo-controlled, USA, peer-reviewed, 3 authors, average treatment delay 4.6 days, dosage 430μg/kg days 1-3, this trial compares with another treatment - results may be better when compared to placebo, trial NCT04510194 (history) (COVID-OUT). risk of death, 197.1% higher, RR 2.97, p = 1.00, treatment 1 of 408 (0.2%), control 0 of 396 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), day 28.
risk of death/hospitalization, 26.7% lower, RR 0.73, p = 0.66, treatment 4 of 406 (1.0%), control 5 of 394 (1.3%), NNT 352, odds ratio converted to relative risk.
risk of progression, 36.8% higher, RR 1.37, p = 0.33, treatment 23 of 406 (5.7%), control 16 of 394 (4.1%), odds ratio converted to relative risk, combined ER, hospitalization, death.
risk of progression, 3.7% higher, RR 1.04, p = 0.78, treatment 105 of 407 (25.8%), control 96 of 391 (24.6%), odds ratio converted to relative risk, combined hypoxemia, ER, hospitalization, death, primary outcome.
risk of hospitalization, 60.8% lower, RR 0.39, p = 0.28, treatment 2 of 206 (1.0%), control 5 of 202 (2.5%), NNT 66, IVM vs. placebo.
risk of hospitalization, 74.6% lower, RR 0.25, p = 0.37, treatment 1 of 137 (0.7%), control 4 of 139 (2.9%), NNT 47, IVM vs. placebo, ≤5 days from onset.
risk of hospitalization, 70.1% lower, RR 0.30, p = 0.12, treatment 3 of 406 (0.7%), control 5 of 202 (2.5%), NNT 58, IVM and IVM+MF vs. placebo.
risk of hospitalization, 41.8% lower, RR 0.58, p = 0.50, treatment 3 of 406 (0.7%), control 5 of 394 (1.3%), NNT 189, IVM and IVM+MF vs. placebo and MF.
Bukhari, 1/16/2021, Randomized Controlled Trial, Pakistan, preprint, 10 authors, study period 15 March, 2020 - 15 June, 2020, dosage 12mg single dose, trial NCT04392713 (history). risk of no viral clearance, 82.4% lower, RR 0.18, p < 0.001, treatment 4 of 41 (9.8%), control 25 of 45 (55.6%), NNT 2.2, day 7, primary outcome.
risk of no viral clearance, 38.7% lower, RR 0.61, p < 0.001, treatment 24 of 41 (58.5%), control 43 of 45 (95.6%), NNT 2.7, day 3.
Buonfrate, 9/6/2021, Double Blind Randomized Controlled Trial, Italy, peer-reviewed, 18 authors, study period 31 July, 2020 - 8 June, 2021, average treatment delay 4.0 days, dosage 1200μg/kg days 1-5, arm B 600µg/kg, arm C 1200µg/kg, trial NCT04438850 (history) (COVER), excluded in exclusion analyses: significant unadjusted group differences, with 3 times as many patients in the ivermectin arms having the baseline visit in a hospital setting, and arm C having large differences in baseline gender, weight, cough, pyrexia, and anosmia, excessive dose for arm C. risk of hospitalization, 210.7% higher, RR 3.11, p = 0.47, treatment 1 of 28 (3.6%), control 0 of 31 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), arm B.
risk of hospitalization, 610.0% higher, RR 7.10, p = 0.11, treatment 3 of 30 (10.0%), control 0 of 31 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), arm C, very high dose, poorly tolerated with low compliance.
relative change in viral load, RR 0.80, p = 0.59, treatment mean 2.5 (±2.2) n=28, control mean 2.0 (±4.4) n=29, day 7, arm B, primary outcome.
relative change in viral load, RR 0.69, p = 0.07, treatment mean 2.9 (±1.6) n=30, control mean 2.0 (±2.1) n=29, day 7, arm C, primary outcome.
Cadegiani, 11/4/2020, prospective, Brazil, peer-reviewed, 4 authors, average treatment delay 2.9 days, dosage 200μg/kg days 1-3, this trial uses multiple treatments in the treatment arm (combined with AZ, nitazoxanide (82), HCQ (22), spironolactone (66), dutasteride (4)) - results of individual treatments may vary, excluded in exclusion analyses: control group retrospectively obtained from untreated patients in the same population. risk of death, 78.3% lower, RR 0.22, p = 0.50, treatment 0 of 110 (0.0%), control 2 of 137 (1.5%), NNT 68, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), control group 1.
risk of mechanical ventilation, 94.2% lower, RR 0.06, p = 0.005, treatment 0 of 110 (0.0%), control 9 of 137 (6.6%), NNT 15, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), control group 1.
risk of hospitalization, 98.0% lower, RR 0.02, p < 0.001, treatment 0 of 110 (0.0%), control 27 of 137 (19.7%), NNT 5.1, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), control group 1.
Carvallo (C), 9/15/2020, prospective, Argentina, peer-reviewed, mean age 55.7, 3 authors, dosage 36mg days 1, 8, dose varied depending on patient condition - mild 24mg, moderate 36mg, severe 48mg, this trial uses multiple treatments in the treatment arm (combined with dexamethasone, enoxaparin, and aspirin) - results of individual treatments may vary, excluded in exclusion analyses: minimal details of groups provided. risk of death, 85.4% lower, RR 0.15, p = 0.08, treatment 1 of 32 (3.1%), control 3 of 14 (21.4%), NNT 5.5, moderate/severe patients, the only treatment death was a patient already in the ICU before treatment, primary outcome.
Chaccour (B), 12/7/2020, Double Blind Randomized Controlled Trial, Spain, peer-reviewed, 23 authors, study period 31 July, 2020 - 11 September, 2020, average treatment delay 1.0 days, dosage 400μg/kg single dose, trial NCT04390022 (history). risk of symptoms, 96.0% lower, OR 0.04, p < 0.05, treatment 12, control 12, logistic regression, chance of presenting any symptom, RR approximated with OR.
viral load, 94.6% lower, relative load 0.05, p < 0.01, treatment 12, control 12, day 7 mid-recovery, average of gene E and gene N, data in supplementary appendix.
risk of no viral clearance, 8.0% lower, RR 0.92, p = 1.00, treatment 12, control 12, primary outcome.
Chahla, 3/30/2021, Cluster Randomized Controlled Trial, Argentina, peer-reviewed, 9 authors, study period September 2020 - January 2021, dosage 24mg days 1, 8, 15, 22, trial NCT04784481 (history). risk of no discharge, 86.9% lower, RR 0.13, p = 0.004, treatment 2 of 110 (1.8%), control 20 of 144 (13.9%), NNT 8.3, adjusted per study, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, logistic regression, multivariable, primary outcome.
Chowdhury, 7/14/2020, Randomized Controlled Trial, Bangladesh, peer-reviewed, 6 authors, study period 2 May, 2020 - 5 June, 2020, dosage 200μg/kg single dose, this trial compares with another treatment - results may be better when compared to placebo, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary, trial NCT04434144 (history). risk of hospitalization, 80.6% lower, RR 0.19, p = 0.23, treatment 0 of 60 (0.0%), control 2 of 56 (3.6%), NNT 28, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of no recovery, 46.4% lower, RR 0.54, p < 0.001, treatment 27 of 60 (45.0%), control 47 of 56 (83.9%), NNT 2.6, mid-recovery day 5.
recovery time, 15.2% lower, relative time 0.85, p = 0.07, treatment 60, control 56.
risk of no viral clearance, 80.6% lower, RR 0.19, p = 0.23, treatment 0 of 60 (0.0%), control 2 of 56 (3.6%), NNT 28, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), primary outcome.
time to viral-, 4.3% lower, relative time 0.96, p = 0.23, treatment 60, control 56.
de Jesús Ascencio-Montiel, 1/24/2022, retrospective, Mexico, peer-reviewed, 10 authors, dosage 6mg days 1-2, this trial uses multiple treatments in the treatment arm (combined with AZ, acetaminophen, aspirin) - results of individual treatments may vary. risk of death/hospitalization, 59.0% lower, RR 0.41, p < 0.001, treatment 7,898, control 20,150, adjusted per study, multivariable, primary outcome.
risk of death/hospitalization, 71.0% lower, RR 0.29, p < 0.001, treatment 5,557, control 12,526, adjusted per study, with phone call followup, multivariable.
risk of death, 15.0% lower, RR 0.85, p = 0.16, treatment 101 of 7,898 (1.3%), control 303 of 20,150 (1.5%), NNT 445, unadjusted, excluded in exclusion analyses: unadjusted results with alternate outcome adjusted results showing significant changes with adjustments.
risk of mechanical ventilation, 9.1% lower, RR 0.91, p = 0.51, treatment 77 of 7,898 (1.0%), control 216 of 20,150 (1.1%), NNT 1031, unadjusted, excluded in exclusion analyses: unadjusted results with alternate outcome adjusted results showing significant changes with adjustments.
risk of hospitalization, 47.6% lower, RR 0.52, p < 0.001, treatment 485 of 7,898 (6.1%), control 2,360 of 20,150 (11.7%), NNT 18, unadjusted, excluded in exclusion analyses: unadjusted results with alternate outcome adjusted results showing significant changes with adjustments.
risk of progression, 41.8% lower, RR 0.58, p < 0.001, treatment 435 of 7,898 (5.5%), control 1,906 of 20,150 (9.5%), NNT 25, unadjusted, ER, excluded in exclusion analyses: unadjusted results with alternate outcome adjusted results showing significant changes with adjustments.
de la Rocha, 5/23/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Mexico, peer-reviewed, 21 authors, study period 1 July, 2020 - 29 January, 2021, dosage 12mg days 1-3, trial NCT04407507 (history), excluded in exclusion analyses: data mismatch, no response from authors. risk of progression to serious adverse events, 186.7% higher, RR 2.87, p = 1.00, treatment 1 of 30 (3.3%), control 0 of 26 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm).
viral load, 2.4% lower, relative load 0.98, p = 0.64, treatment mean 33.74 (±4.77) n=30, control mean 32.94 (±7.74) n=26, day 14.
viral load, 7.8% lower, relative load 0.92, p = 0.04, treatment mean 30.64 (±3.74) n=30, control mean 28.25 (±4.21) n=26, mid-recovery, day 5.
Elalfy, 2/16/2021, retrospective, Egypt, peer-reviewed, 15 authors, dosage 18mg days 1, 4, 7, 10, 13, <90kg 18mg, 90-120kg 24mg, >120kg 30mg, this trial uses multiple treatments in the treatment arm (combined with nitazoxanide, ribavirin, and zinc) - results of individual treatments may vary. risk of no viral clearance, 86.9% lower, RR 0.13, p < 0.001, treatment 7 of 62 (11.3%), control 44 of 51 (86.3%), NNT 1.3, day 15, primary outcome.
risk of no viral clearance, 58.1% lower, RR 0.42, p < 0.001, treatment 26 of 62 (41.9%), control 51 of 51 (100.0%), NNT 1.7, day 7.
Espitia-Hernandez, 8/15/2020, retrospective, Mexico, peer-reviewed, mean age 45.1, 5 authors, dosage 6mg days 1-2, 8-9, this trial uses multiple treatments in the treatment arm (combined with azithromycin and cholecalciferol) - results of individual treatments may vary. recovery time, 70.0% lower, relative time 0.30, p < 0.001, treatment 28, control 7.
risk of viral+ at day 10, 97.2% lower, RR 0.03, p < 0.001, treatment 0 of 28 (0.0%), control 7 of 7 (100.0%), NNT 1.0, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), primary outcome.
Faisal, 5/10/2021, Randomized Controlled Trial, Pakistan, peer-reviewed, 3 authors, study period 5 April, 2020 - 30 May, 2020, dosage 12mg days 1-5. risk of no recovery, 68.4% lower, RR 0.32, p = 0.005, treatment 6 of 50 (12.0%), control 19 of 50 (38.0%), NNT 3.8, 6-8 days, mid-recovery, primary outcome.
risk of no recovery, 27.3% lower, RR 0.73, p = 0.11, treatment 24 of 50 (48.0%), control 33 of 50 (66.0%), NNT 5.6, 3-5 days.
risk of no recovery, 75.0% lower, RR 0.25, p = 0.09, treatment 2 of 50 (4.0%), control 8 of 50 (16.0%), NNT 8.3, 9-10 days.
Ghauri, 12/15/2020, retrospective, Pakistan, peer-reviewed, 6 authors, dosage 12mg days 1-6. risk of fever, 92.2% lower, RR 0.08, p = 0.04, treatment 0 of 37 (0.0%), control 7 of 53 (13.2%), NNT 7.6, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), day 14.
risk of fever, 86.4% lower, RR 0.14, p < 0.001, treatment 2 of 37 (5.4%), control 21 of 53 (39.6%), NNT 2.9, day 10.
risk of fever, 55.7% lower, RR 0.44, p < 0.001, treatment 13 of 37 (35.1%), control 42 of 53 (79.2%), NNT 2.3, day 7.
risk of fever, 42.2% lower, RR 0.58, p < 0.001, treatment 21 of 37 (56.8%), control 52 of 53 (98.1%), NNT 2.4, day 5.
Kamal, 9/1/2020, Randomized Controlled Trial, trial NCT04425707 (history). Estimated 100 patient RCT with results unknown and over 3 years late.
Krolewiecki, 6/18/2021, Randomized Controlled Trial, Argentina, peer-reviewed, 23 authors, study period 18 May, 2020 - 9 September, 2020, average treatment delay 3.5 days, dosage 600μg/kg days 1-5, trial NCT004381884 (history). risk of mechanical ventilation, 151.9% higher, RR 2.52, p = 1.00, treatment 1 of 27 (3.7%), control 0 of 14 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm).
risk of progression, 3.7% higher, RR 1.04, p = 1.00, treatment 2 of 27 (7.4%), control 1 of 14 (7.1%).
viral decay rate, RR 0.34, p = 0.09, treatment 20, control 14, relative mean viral decay rate (corrigendum table 2, all treatment patients vs. all control patients), primary outcome.
Loue, 4/17/2021, retrospective quasi-randomized (patient choice), France, peer-reviewed, 2 authors, dosage 200μg/kg single dose. risk of death, 70.0% lower, RR 0.30, p = 0.34, treatment 1 of 10 (10.0%), control 5 of 15 (33.3%), NNT 4.3.
risk of severe case, 55.0% lower, RR 0.45, p = 0.11, treatment 3 of 10 (30.0%), control 10 of 15 (66.7%), NNT 2.7.
López-Medina, 3/4/2021, Double Blind Randomized Controlled Trial, Colombia, peer-reviewed, median age 37.0, 19 authors, study period 15 July, 2020 - 30 November, 2020, average treatment delay 5.0 days, dosage 300μg/kg days 1-5, excluded in exclusion analyses: strong evidence of patients in the control group self-medicating, ivermectin widely used in the population at that time, and the study drug identity was concealed by using the name D11AX22. risk of death, 66.8% lower, RR 0.33, p = 0.50, treatment 0 of 200 (0.0%), control 1 of 198 (0.5%), NNT 198, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of escalation of care, 60.8% lower, RR 0.39, p = 0.11, treatment 4 of 200 (2.0%), control 10 of 198 (5.1%), NNT 33, odds ratio converted to relative risk.
risk of escalation of care with post-hoc <12h exclusion, 34.3% lower, RR 0.66, p = 0.52, treatment 4 of 200 (2.0%), control 6 of 198 (3.0%), NNT 97, odds ratio converted to relative risk.
risk of deterioration by >= 2 points on an 8-point scale, 43.1% lower, RR 0.57, p = 0.37, treatment 4 of 200 (2.0%), control 7 of 198 (3.5%), NNT 65, odds ratio converted to relative risk, primary outcome.
risk of fever post randomization, 24.8% lower, RR 0.75, p = 0.38, treatment 16 of 200 (8.0%), control 21 of 198 (10.6%), NNT 38, odds ratio converted to relative risk.
risk of unresolved symptoms at day 21, 15.3% lower, RR 0.85, p = 0.53, treatment 36 of 200 (18.0%), control 42 of 198 (21.2%), NNT 31, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, Cox proportional-hazard model.
lack of resolution of symptoms, 6.5% lower, HR 0.93, p = 0.53, treatment 200, control 198, inverted to make HR<1 favor treatment, post-hoc primary outcome.
Mahmud, 10/9/2020, Double Blind Randomized Controlled Trial, Bangladesh, peer-reviewed, 15 authors, study period 1 June, 2020 - 30 August, 2020, average treatment delay 4.0 days, dosage 12mg single dose, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary, trial NCT04523831 (history). risk of death, 85.7% lower, HR 0.14, p = 0.25, treatment 0 of 183 (0.0%), control 3 of 183 (1.6%), NNT 61, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of progression, 57.0% lower, HR 0.43, p < 0.001, treatment 16 of 183 (8.7%), control 32 of 180 (17.8%), NNT 11, adjusted per study, Cox regression.
risk of no recovery, 94.0% lower, HR 0.06, p < 0.001, treatment 72 of 183 (39.3%), control 100 of 180 (55.6%), NNT 6.2, adjusted per study, day 7, Cox regression.
risk of no recovery, 38.5% lower, RR 0.61, p = 0.005, treatment 40 of 183 (21.9%), control 64 of 180 (35.6%), NNT 7.3, day 11.
risk of no recovery, 96.0% lower, HR 0.04, p < 0.001, treatment 42 of 183 (23.0%), control 67 of 180 (37.2%), NNT 7.0, adjusted per study, day 12, Cox regression.
time to recovery, 27.0% lower, HR 0.73, p = 0.003, treatment 183, control 180, Cox regression, primary outcome.
risk of no viral clearance, 39.0% lower, HR 0.61, p = 0.002, treatment 14 of 183 (7.7%), control 36 of 180 (20.0%), NNT 8.1, adjusted per study, Cox regression.
Manomaipiboon, 2/2/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Thailand, peer-reviewed, mean age 48.6, 8 authors, study period 10 October, 2021 - 15 December, 2021, dosage 12mg days 1-5, trial NCT05076253 (history). risk of no recovery, 43.5% lower, RR 0.57, p = 0.26, treatment 3 of 36 (8.3%), control 6 of 36 (16.7%), NNT 12, adjusted per study, odds ratio converted to relative risk, resolution of symptoms, Table S2, day 28.
recovery time, 15.3% lower, RR 0.85, p = 0.56, treatment 36, control 36, inverted to make RR<1 favor treatment, time to resolution of symptoms.
risk of no viral clearance, 5.0% lower, RR 0.95, p = 1.00, treatment 19 of 36 (52.8%), control 20 of 36 (55.6%), NNT 36, day 14, primary outcome.
risk of no viral clearance, 3.3% lower, RR 0.97, p = 1.00, treatment 29 of 36 (80.6%), control 30 of 36 (83.3%), NNT 36, day 7.
Mayer, 9/23/2021, retrospective, Argentina, peer-reviewed, 14 authors, dosage 540μg/kg days 1-5, mean prescribed dose. risk of death, 55.1% lower, RR 0.45, p < 0.001, treatment 3,266, control 17,966, adjusted per study, odds ratio converted to relative risk, Figure 3, multivariable.
risk of ICU admission, 65.9% lower, RR 0.34, p < 0.001, treatment 3,266, control 17,966, adjusted per study, odds ratio converted to relative risk, Figure 3, multivariable.
risk of death, 27.6% lower, RR 0.72, p = 0.03, treatment 3,266, control 17,966, odds ratio converted to relative risk, unadjusted.
risk of ICU admission, 26.0% lower, RR 0.74, p = 0.13, treatment 3,266, control 17,966, odds ratio converted to relative risk, unadjusted.
Merino, 5/3/2021, retrospective quasi-randomized (patients receiving kit), population-based cohort, Mexico, preprint, 7 authors, dosage 6mg bid days 1-2. risk of hospitalization, 74.4% lower, RR 0.26, p < 0.001, model 7, same time period, patients receiving kit.
risk of hospitalization, 68.4% lower, RR 0.32, p < 0.001, model 1, different time periods, administrative rule.
Mikamo, 9/26/2022, Double Blind Randomized Controlled Trial, placebo-controlled, multiple countries, peer-reviewed, 19 authors, study period 12 November, 2021 - 7 August, 2022, dosage 300μg/kg days 1-3, trial NCT05056883 (history), excluded in exclusion analyses: very low risk group with almost no progression leaves little room for improvement, unbalanced baseline dyspnea and high symptom scores, design and post-hoc changes favor null result. COVID-19 pneumonia, 205.0% higher, RR 3.05, p = 0.49, treatment 1 of 502 (0.2%), control 0 of 527 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), Table S8.
use of therapeutic agents, oxygen, transfer, hospitalization, death, 214.9% higher, RR 3.15, p = 0.36, treatment 3 of 502 (0.6%), control 1 of 527 (0.2%).
risk of no improvement, 4.0% higher, HR 1.04, p = 0.62, treatment 502, control 527, 168hr, improving trend, primary outcome.
risk of no recovery, 4.0% lower, HR 0.96, p = 0.72, treatment 502, control 527, 168hr, clinical resolution, Figure S3.
risk of no recovery, 4.0% lower, HR 0.96, p = 0.64, treatment 502, control 527, 240hr, clinical resolution, Figure S3.
Mirahmadizadeh, 6/23/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Iran, peer-reviewed, 12 authors, study period 9 April, 2021 - 20 May, 2021, dosage 24mg single dose, 12mg and 24mg arms. risk of mechanical ventilation, 66.9% lower, RR 0.33, p = 0.37, treatment 1 of 131 (0.8%), control 3 of 130 (2.3%), NNT 65, 24mg.
risk of mechanical ventilation, 33.3% lower, RR 0.67, p = 1.00, treatment 2 of 130 (1.5%), control 3 of 130 (2.3%), NNT 130, 12mg.
risk of hospitalization, 45.9% lower, RR 0.54, p = 0.22, treatment 6 of 131 (4.6%), control 11 of 130 (8.5%), NNT 26, 24mg, primary outcome.
risk of hospitalization, 27.3% lower, RR 0.73, p = 0.63, treatment 8 of 130 (6.2%), control 11 of 130 (8.5%), NNT 43, 12mg, primary outcome.
risk of no recovery, 38.9% lower, RR 0.61, p = 0.27, treatment 8 of 131 (6.1%), control 13 of 130 (10.0%), NNT 26, day 28, 24mg.
risk of no recovery, 30.8% lower, RR 0.69, p = 0.50, treatment 9 of 130 (6.9%), control 13 of 130 (10.0%), NNT 32, day 28, 12mg.
Mohan, 2/2/2021, Double Blind Randomized Controlled Trial, India, peer-reviewed, 27 authors, study period 28 July, 2020 - 29 September, 2020, average treatment delay 5.0 days, dosage 400μg/kg single dose, 200μg/kg also tested, trial CTRI/2020/06/026001 (RIVET-COV). risk of no discharge at day 14, 62.5% lower, RR 0.38, p = 0.27, treatment 2 of 40 (5.0%), control 6 of 45 (13.3%), NNT 12, 24mg.
risk of clinical worsening, 32.5% lower, RR 0.68, p = 0.72, treatment 3 of 40 (7.5%), control 5 of 45 (11.1%), NNT 28, 24mg.
risk of no viral clearance, 23.8% lower, RR 0.76, p = 0.18, treatment 21 of 40 (52.5%), control 31 of 45 (68.9%), NNT 6.1, day 5, 24mg, primary outcome.
risk of no viral clearance, 10.3% lower, RR 0.90, p = 0.65, treatment 20 of 36 (55.6%), control 26 of 42 (61.9%), NNT 16, day 7, 24mg.
Mourya, 4/1/2021, retrospective, India, peer-reviewed, 5 authors, dosage 12mg days 1-7. risk of no viral clearance, 89.4% lower, RR 0.11, p < 0.001, treatment 5 of 50 (10.0%), control 47 of 50 (94.0%), NNT 1.2, primary outcome.
Ravikirti (B), 1/9/2021, Double Blind Randomized Controlled Trial, India, peer-reviewed, 11 authors, study period 1 August, 2020 - 31 October, 2020, average treatment delay 6.1 days, dosage 12mg days 1, 2. risk of death, 88.7% lower, RR 0.11, p = 0.12, treatment 0 of 55 (0.0%), control 4 of 57 (7.0%), NNT 14, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of mechanical ventilation, 79.3% lower, RR 0.21, p = 0.10, treatment 1 of 55 (1.8%), control 5 of 57 (8.8%), NNT 14.
risk of ICU admission, 13.6% lower, RR 0.86, p = 0.80, treatment 5 of 55 (9.1%), control 6 of 57 (10.5%), NNT 70.
risk of no hospital discharge, 88.7% lower, RR 0.11, p = 0.12, treatment 0 of 55 (0.0%), control 4 of 57 (7.0%), NNT 14, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of no viral clearance, 11.6% higher, RR 1.12, p = 0.35, treatment 42 of 55 (76.4%), control 39 of 57 (68.4%).
Reis, 8/6/2021, Double Blind Randomized Controlled Trial, Brazil, peer-reviewed, 27 authors, study period 23 March, 2021 - 6 August, 2021, dosage 400μg/kg days 1-3, impossible data, see notes, trial NCT04727424 (history) (TOGETHER), excluded in exclusion analyses: multiple anomalies as per detailed analysis. risk of death, 12.0% lower, RR 0.88, p = 0.68, treatment 21 of 679 (3.1%), control 24 of 679 (3.5%), NNT 226.
risk of mechanical ventilation, 23.0% lower, RR 0.77, p = 0.38, treatment 19 of 679 (2.8%), control 25 of 679 (3.7%), NNT 113.
risk of hospitalization, 17.0% lower, RR 0.83, p = 0.19, treatment 79 of 679 (11.6%), control 95 of 679 (14.0%), NNT 42.
extended ER observation or hospitalization, 10.0% lower, RR 0.90, p = 0.42, treatment 100 of 679 (14.7%), control 111 of 679 (16.3%), NNT 62, primary outcome.
viral clearance, no change, RR 1.00, p = 1.00, treatment 106 of 142 (74.6%), control 123 of 165 (74.5%), day 7.
viral clearance, 31.6% higher, RR 1.32, p = 0.46, treatment 148, control 170, inverted to make RR<1 favor treatment, day 3.
Rezai, 6/16/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Iran, peer-reviewed, mean age 35.5, 29 authors, study period 19 February, 2021 - 30 August, 2021, dosage 400μg/kg days 1-3, trial IRCT20111224008507N4, excluded in exclusion analyses: multiple critical issues, see study page. risk of death, 4.9% higher, RR 1.05, p = 1.00, treatment 1 of 268 (0.4%), control 1 of 281 (0.4%).
risk of ICU admission, 9.0% higher, RR 1.09, p = 0.95, treatment 268, control 281.
risk of hospitalization, 36.0% higher, RR 1.36, p = 0.41, treatment 268, control 281.
risk of no recovery, 2.0% higher, RR 1.02, p = 0.49, treatment 268, control 281, inverted to make RR<1 favor treatment, post-hoc primary outcome.
risk of no recovery, 13.2% lower, RR 0.87, p = 0.09, treatment mean 3.87 (±0.18) n=268, control mean 4.46 (±0.18) n=281, cough.
risk of no recovery, 16.7% lower, RR 0.83, p = 0.81, treatment mean 2.5 (±0.51) n=268, control mean 3.0 (±0.92) n=281, tachypnea.
risk of no viral clearance, 23.5% higher, RR 1.23, p = 0.16, treatment 268, control 281, inverted to make RR<1 favor treatment.
Roy, 3/12/2021, retrospective, database analysis, India, preprint, 5 authors, dosage not specified, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary, excluded in exclusion analyses: no serious outcomes reported and fast recovery in treatment and control groups, there is little room for a treatment to improve results. relative time to clinical response of wellbeing, 5.6% lower, relative time 0.94, p = 0.87, treatment 14, control 15, primary outcome.
Schilling, 7/19/2022, Randomized Controlled Trial, Thailand, peer-reviewed, median age 27.0, 38 authors, study period 30 September, 2021 - 18 April, 2022, average treatment delay 2.0 days, dosage 600μg/kg days 1-7, trial NCT05041907 (history) (PLATCOV), excluded in exclusion analyses: post-hoc change to exclude patients treated before high viral load, population very low risk, recovering quickly without treatment, high baseline immunity, 2.2x greater baseline antibody negative for the treatment arm. risk of hospitalization, 66.7% lower, RR 0.33, p = 1.00, treatment 0 of 45 (0.0%), control 1 of 45 (2.2%), NNT 45, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of progression, 85.7% lower, RR 0.14, p = 0.24, treatment 0 of 45 (0.0%), control 3 of 45 (6.7%), NNT 15, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), hospitalization or progression to COVID-19 rhabdomyolysis.
relative clearance rate, 9.1% worse, RR 1.09, p = 0.36, treatment 45, control 45, primary outcome.
Szente Fonseca, 10/31/2020, retrospective, Brazil, peer-reviewed, mean age 50.6, 10 authors, average treatment delay 4.6 days, dosage 12mg days 1-2, excluded in exclusion analyses: result is likely affected by collinearity across treatments in the model. risk of hospitalization, 13.9% higher, RR 1.14, p = 0.53, treatment 340, control 377, adjusted per study, odds ratio converted to relative risk, control prevalence approximated with overall prevalence, primary outcome.
Vallejos, 7/2/2021, Double Blind Randomized Controlled Trial, Argentina, peer-reviewed, 29 authors, study period 19 August, 2020 - 22 February, 2021, average treatment delay 4.0 days, dosage 12mg days 1-2, <80kg 12mg, 80-110kg 18mg, >110kg 24mg. risk of death, 33.5% higher, RR 1.33, p = 0.70, treatment 4 of 250 (1.6%), control 3 of 251 (1.2%), odds ratio converted to relative risk.
risk of mechanical ventilation, 33.5% higher, RR 1.33, p = 0.70, treatment 4 of 250 (1.6%), control 3 of 251 (1.2%), odds ratio converted to relative risk.
risk of hospitalization, 33.0% lower, RR 0.67, p = 0.23, treatment 14 of 250 (5.6%), control 21 of 251 (8.4%), NNT 36, odds ratio converted to relative risk, primary outcome.
risk of no viral clearance, 5.0% higher, RR 1.05, p = 0.55, treatment 137 of 250 (54.8%), control 131 of 251 (52.2%), inverted to make RR<1 favor treatment, odds ratio converted to relative risk, day 3.
risk of no viral clearance, 26.8% higher, RR 1.27, p = 0.29, treatment 38 of 250 (15.2%), control 30 of 251 (12.0%), inverted to make RR<1 favor treatment, odds ratio converted to relative risk, day 12.
Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. Only the first (most serious) outcome is used in pooled analysis, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.
Abd-Elsalam, 6/2/2021, Randomized Controlled Trial, Egypt, peer-reviewed, 16 authors, study period March 2020 - October 2020, dosage 12mg days 1-3. risk of death, 25.0% lower, RR 0.75, p = 0.70, treatment 3 of 82 (3.7%), control 4 of 82 (4.9%), NNT 82, odds ratio converted to relative risk, logistic regression, primary outcome.
risk of mechanical ventilation, no change, RR 1.00, p = 1.00, treatment 3 of 82 (3.7%), control 3 of 82 (3.7%).
hospitalization time, 19.6% lower, relative time 0.80, p = 0.09, treatment 82, control 82.
Ahsan, 4/29/2021, retrospective, Pakistan, peer-reviewed, 10 authors, dosage 150μg/kg days 1-2, 150-200µg/kg, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary, excluded in exclusion analyses: unadjusted results with no group details. risk of death, 50.0% lower, RR 0.50, p = 0.03, treatment 17 of 110 (15.5%), control 17 of 55 (30.9%), NNT 6.5.
Aref (B), 9/19/2022, Randomized Controlled Trial, placebo-controlled, Egypt, peer-reviewed, 9 authors, study period 1 March, 2021 - 30 April, 2021, dosage 2 puffs of 70 μg/mL nasal ivermectin, trial NCT04951362 (history). recovery time, 74.0% lower, relative time 0.26, p < 0.001, treatment 49, control 47, anosmia.
Baguma, 12/28/2021, retrospective, Uganda, preprint, 16 authors, study period March 2020 - October 2021, dosage not specified. risk of death, 96.8% lower, RR 0.03, p = 0.31, treatment 7, control 474, adjusted per study, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, multivariable, control prevalance approximated with overall prevalence.
Beltran Gonzalez, 2/23/2021, Double Blind Randomized Controlled Trial, Mexico, peer-reviewed, mean age 53.8, 13 authors, average treatment delay 7.0 days, dosage 12mg single dose, 18mg for patients >80kg, excluded in exclusion analyses: major inconsistencies reported and the data is no longer available, although the authors state that it is available, and have shared it with an anti-treatment group. risk of death, 14.4% lower, RR 0.86, p = 1.00, treatment 5 of 36 (13.9%), control 6 of 37 (16.2%), NNT 43.
risk of respiratory deterioration or death, 8.6% lower, RR 0.91, p = 1.00, treatment 8 of 36 (22.2%), control 9 of 37 (24.3%), NNT 48.
risk of no hospital discharge, 37.0% higher, RR 1.37, p = 0.71, treatment 4 of 36 (11.1%), control 3 of 37 (8.1%).
hospitalization time, 20.0% higher, relative time 1.20, p = 0.43, treatment 36, control 37, primary outcome.
Budhiraja, 11/18/2020, retrospective, India, preprint, 12 authors, dosage not specified. risk of death, 99.1% lower, RR 0.009, p = 0.04, treatment 0 of 34 (0.0%), control 103 of 942 (10.9%), NNT 9.1, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), unadjusted, primary outcome.
Camprubí, 11/11/2020, retrospective, Spain, peer-reviewed, 9 authors, average treatment delay 12.0 days, dosage 200μg/kg single dose. risk of mechanical ventilation, 40.0% lower, RR 0.60, p = 0.67, treatment 3 of 13 (23.1%), control 5 of 13 (38.5%), NNT 6.5.
risk of ICU admission, 33.3% lower, RR 0.67, p = 1.00, treatment 2 of 13 (15.4%), control 3 of 13 (23.1%), NNT 13, ICU at day 8.
risk of no improvement at day 8, 33.3% higher, RR 1.33, p = 1.00, treatment 4 of 13 (30.8%), control 3 of 13 (23.1%).
risk of no viral clearance, 25.0% higher, RR 1.25, p = 1.00, treatment 5 of 13 (38.5%), control 4 of 13 (30.8%), tests done between days 3-5, primary outcome.
Chachar, 9/30/2020, Randomized Controlled Trial, India, peer-reviewed, 6 authors, dosage 36mg, 12mg stat, 12mg after 12 hours, 12mg after 24 hours. risk of no recovery at day 7, 10.0% lower, RR 0.90, p = 0.50, treatment 9 of 25 (36.0%), control 10 of 25 (40.0%), NNT 25, primary outcome.
Efimenko, 2/28/2022, retrospective, propensity score matching, USA, peer-reviewed, 6 authors, study period 1 January, 2020 - 11 July, 2021, dosage not specified, this trial compares with another treatment - results may be better when compared to placebo. risk of death, 69.2% lower, OR 0.31, p < 0.001, treatment 1,072, control 40,536, propensity score matching, RR approximated with OR.
Elavarasi, 8/12/2021, retrospective, India, preprint, 26 authors, dosage not specified, excluded in exclusion analyses: unadjusted results with no group details. risk of death, 19.6% lower, RR 0.80, p = 0.12, treatment 48 of 283 (17.0%), control 311 of 1,475 (21.1%), NNT 24, unadjusted.
Ferreira, 11/26/2021, retrospective, Brazil, peer-reviewed, 5 authors, study period 12 March, 2020 - 8 July, 2020, average treatment delay 7.0 days, dosage not specified, excluded in exclusion analyses: unadjusted results with no group details; substantial unadjusted confounding by indication likely. risk of death, 5.0% higher, RR 1.05, p = 1.00, treatment 3 of 21 (14.3%), control 11 of 81 (13.6%).
risk of death/intubation, 54.3% higher, RR 1.54, p = 0.37, treatment 6 of 21 (28.6%), control 15 of 81 (18.5%).
risk of death/intubation/ICU, 62.4% higher, RR 1.62, p = 0.27, treatment 8 of 21 (38.1%), control 19 of 81 (23.5%).
George, 5/27/2022, Randomized Controlled Trial, India, peer-reviewed, 15 authors, study period June 2020 - February 2021, dosage 24mg single dose, trial CTRI/2020/05/025068. risk of death, 30.4% lower, RR 0.70, p = 0.55, treatment 5 of 35 (14.3%), control 8 of 39 (20.5%), NNT 16, 24mg.
risk of death, 2.6% higher, RR 1.03, p = 1.00, treatment 8 of 38 (21.1%), control 8 of 39 (20.5%), 12mg.
recovery time, 18.7% lower, relative time 0.81, p = 0.37, treatment mean 4.82 (±4.35) n=35, control mean 5.93 (±5.93) n=39, 24mg.
recovery time, 6.2% lower, relative time 0.94, p = 0.78, treatment mean 5.56 (±5.42) n=38, control mean 5.93 (±5.93) n=39, 12mg.
risk of progression, 33.1% lower, RR 0.67, p = 0.41, treatment 6 of 35 (17.1%), control 10 of 39 (25.6%), NNT 12, 24mg.
risk of progression, 17.9% lower, RR 0.82, p = 0.79, treatment 8 of 38 (21.1%), control 10 of 39 (25.6%), NNT 22, 12mg.
Gorial, 7/8/2020, retrospective, Iraq, preprint, 9 authors, dosage 200μg/kg single dose, trial NCT04343092 (history). risk of death, 71.0% lower, RR 0.29, p = 1.00, treatment 0 of 16 (0.0%), control 2 of 71 (2.8%), NNT 36, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
hospitalization time, 42.0% lower, relative time 0.58, p < 0.001, treatment 16, control 71.
risk of no recovery, 71.0% lower, RR 0.29, p = 1.00, treatment 0 of 16 (0.0%), control 2 of 71 (2.8%), NNT 36, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), primary outcome.
Hashim, 10/26/2020, Single Blind Randomized Controlled Trial, Iraq, peer-reviewed, 7 authors, study period 1 July, 2020 - 14 October, 2020, dosage 200μg/kg days 1-2, some patients received a third dose on day 8, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary, trial NCT04591600 (history). risk of death, 91.7% lower, RR 0.08, p = 0.03, treatment 0 of 59 (0.0%), control 6 of 70 (8.6%), NNT 12, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), excluding non-randomized critical patients, primary outcome.
risk of death, 67.1% lower, RR 0.33, p = 0.16, treatment 2 of 70 (2.9%), control 6 of 70 (8.6%), NNT 18, odds ratio converted to relative risk, including critical patients that were always allocated to treatment.
risk of progression, 83.1% lower, RR 0.17, p = 0.07, treatment 1 of 59 (1.7%), control 7 of 70 (10.0%), NNT 12, excluding non-randomized critical patients.
risk of progression, 57.4% lower, RR 0.43, p = 0.20, treatment 3 of 70 (4.3%), control 7 of 70 (10.0%), NNT 18, odds ratio converted to relative risk, including critical patients that were always allocated to treatment.
recovery time, 40.7% lower, relative time 0.59, p < 0.001, treatment 70, control 70.
Hayward, 2/29/2024, Randomized Controlled Trial, placebo-controlled, United Kingdom, preprint, 21 authors, study period 23 June, 2021 - 1 July, 2022, trial ISRCTN86534580 (PRINCIPLE). risk of death/hospitalization, 1.0% higher, HR 1.01, p = 0.97, treatment 34 of 2,157 (1.6%), control 27 of 1,806 (1.5%), concurrent and eligible, primary outcome.
risk of mechanical ventilation, 151.4% higher, RR 2.51, p = 0.63, treatment 3 of 2,149 (0.1%), control 1 of 1,801 (0.1%).
risk of ICU admission, 319.0% higher, RR 4.19, p = 0.23, treatment 5 of 2,149 (0.2%), control 1 of 1,801 (0.1%).
time to sustained recovery, 16.0% lower, HR 0.84, p < 0.001, treatment 2,157, control 1,806, inverted to make HR<1 favor treatment, concurrent and eligible.
early sustained recovery, 23.1% lower, HR 0.77, p < 0.001, treatment 2,154, control 1,805, inverted to make HR<1 favor treatment, concurrent and eligible.
sustained alleviation, 17.4% lower, HR 0.83, p < 0.001, treatment 1,826, control 1,535, inverted to make HR<1 favor treatment, concurrent and eligible.
alleviation of all symptoms, 16.0% lower, HR 0.84, p < 0.001, treatment 2,154, control 1,805, inverted to make HR<1 favor treatment, concurrent and eligible.
first reported recovery, 13.0% lower, HR 0.87, p < 0.001, treatment 2,157, control 1,806, inverted to make HR<1 favor treatment, concurrent and eligible, primary outcome.
no recovery at 3/6/12 months, 28.0% lower, HR 0.72, p = 0.02, treatment 94 of 1,941 (4.8%), control 109 of 1,624 (6.7%), NNT 54.
risk of no recovery, 17.6% lower, RR 0.82, p = 0.001, treatment 417 of 1,848 (22.6%), control 420 of 1,533 (27.4%), NNT 21, day 365.
risk of PASC, 36.3% lower, RR 0.64, p < 0.001, treatment 1,886, control 1,567, all symptoms combined.
risk of PASC, 63.2% higher, RR 1.63, p = 1.00, treatment 2 of 1,507 (0.1%), control 1 of 1,230 (0.1%), ongoing/persistent, fever.
risk of PASC, 72.7% lower, RR 0.27, p = 0.33, treatment 1 of 1,819 (0.1%), control 3 of 1,489 (0.2%), NNT 683, ongoing/persistent, cough.
risk of PASC, 50.1% lower, RR 0.50, p = 0.04, treatment 15 of 1,886 (0.8%), control 25 of 1,567 (1.6%), NNT 125, ongoing/persistent, dyspnea.
risk of PASC, 35.3% higher, RR 1.35, p = 0.74, treatment 5 of 1,808 (0.3%), control 3 of 1,468 (0.2%), ongoing/persistent, chest pain.
risk of PASC, 25.2% lower, RR 0.75, p = 0.36, treatment 21 of 1,831 (1.1%), control 23 of 1,501 (1.5%), NNT 259, ongoing/persistent, smell.
risk of PASC, 58.7% lower, RR 0.41, p = 0.42, treatment 2 of 1,821 (0.1%), control 4 of 1,503 (0.3%), NNT 640, ongoing/persistent, diarrhoea.
risk of PASC, 70.2% lower, RR 0.30, p < 0.001, treatment 10 of 1,739 (0.6%), control 27 of 1,400 (1.9%), NNT 74, ongoing/persistent, headache.
risk of PASC, 29.8% lower, RR 0.70, p = 0.25, treatment 23 of 1,739 (1.3%), control 27 of 1,433 (1.9%), NNT 178, ongoing/persistent, muscle ache.
risk of PASC, 47.1% lower, RR 0.53, p = 0.03, treatment 19 of 1,724 (1.1%), control 30 of 1,441 (2.1%), NNT 102, ongoing/persistent, generally unwell.
risk of PASC, 20.1% lower, RR 0.80, p = 0.19, treatment 66 of 1,876 (3.5%), control 69 of 1,567 (4.4%), NNT 113, ongoing/persistent, fatigue.
risk of PASC, 28.5% lower, RR 0.72, p < 0.001, treatment 1,513, control 1,238, all symptoms combined.
risk of PASC, 39.9% lower, RR 0.60, p = 0.01, treatment 46 of 1,435 (3.2%), control 51 of 1,136 (4.5%), relatedness (yes + unsure) 9.1% (treatment) 11.2% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, cough.
risk of PASC, 41.6% lower, RR 0.58, p < 0.001, treatment 96 of 1,513 (6.3%), control 106 of 1,238 (8.6%), relatedness (yes + unsure) 14.4% (treatment) 18.5% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, shortness of breath.
risk of PASC, 22.2% lower, RR 0.78, p = 0.27, treatment 39 of 1,426 (2.7%), control 36 of 1,117 (3.2%), relatedness (yes + unsure) 7.6% (treatment) 8.4% (control) , NNT 205, adjusted per study and for relatedness, moderate/major symptoms at 12 months, chest pain.
risk of PASC, 38.0% lower, RR 0.62, p = 0.02, treatment 46 of 1,427 (3.2%), control 44 of 1,140 (3.9%), relatedness (yes + unsure) 7.7% (treatment) 10.3% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, palpitations.
risk of PASC, 11.5% higher, RR 1.12, p = 0.52, treatment 77 of 1,403 (5.5%), control 57 of 1,131 (5.0%), relatedness (yes + unsure) 13.2% (treatment) 12.9% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, smell.
risk of PASC, 36.4% lower, RR 0.64, p = 0.02, treatment 51 of 1,375 (3.7%), control 51 of 1,116 (4.6%), relatedness (yes + unsure) 1.1% (treatment) 1.4% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, taste.
risk of PASC, 22.1% lower, RR 0.78, p = 0.37, treatment 26 of 1,290 (2.0%), control 25 of 1,057 (2.4%), relatedness (yes + unsure) 4.8% (treatment) 5.3% (control) , NNT 286, adjusted per study and for relatedness, moderate/major symptoms at 12 months, ear ache.
risk of PASC, 9.6% higher, RR 1.10, p = 0.73, treatment 37 of 1,325 (2.8%), control 25 of 1,089 (2.3%), relatedness (yes + unsure) 6.7% (treatment) 7.4% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, sore throat.
risk of PASC, 34.8% lower, RR 0.65, p = 0.12, treatment 24 of 1,308 (1.8%), control 27 of 1,077 (2.5%), relatedness (yes + unsure) 6.0% (treatment) 6.9% (control) , NNT 149, adjusted per study and for relatedness, moderate/major symptoms at 12 months, hoarse voice.
risk of PASC, 16.6% higher, RR 1.17, p = 0.42, treatment 60 of 1,341 (4.5%), control 46 of 1,082 (4.3%), relatedness (yes + unsure) 8.3% (treatment) 7.4% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, tinnitus.
risk of PASC, 34.7% lower, RR 0.65, p = 0.29, treatment 13 of 1,382 (0.9%), control 12 of 1,139 (1.1%), relatedness (yes + unsure) 2.2% (treatment) 3.0% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, vomiting.
risk of PASC, 46.5% higher, RR 1.46, p = 0.12, treatment 44 of 1,439 (3.1%), control 25 of 1,175 (2.1%), relatedness (yes + unsure) 5.9% (treatment) 5.8% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, abdominal pain.
risk of PASC, 1.6% lower, RR 0.98, p = 0.95, treatment 33 of 1,416 (2.3%), control 24 of 1,167 (2.1%), relatedness (yes + unsure) 4.4% (treatment) 5.1% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, diarrhoea.
risk of PASC, 43.8% higher, RR 1.44, p = 0.23, treatment 30 of 1,417 (2.1%), control 17 of 1,160 (1.5%), relatedness (yes + unsure) 4.5% (treatment) 4.6% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, reduced appetite.
risk of PASC, 89.8% higher, RR 1.90, p = 0.27, treatment 11 of 1,375 (0.8%), control 4 of 1,129 (0.4%), relatedness (yes + unsure) 1.8% (treatment) 2.2% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, weight loss.
risk of PASC, 41.3% lower, RR 0.59, p < 0.001, treatment 89 of 1,287 (6.9%), control 94 of 973 (9.7%), relatedness (yes + unsure) 13.2% (treatment) 16.2% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, anxiety.
risk of PASC, 38.6% lower, RR 0.61, p < 0.001, treatment 95 of 1,298 (7.3%), control 92 of 992 (9.3%), relatedness (yes + unsure) 13.7% (treatment) 17.4% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, depression.
risk of PASC, 49.0% lower, RR 0.51, p < 0.001, treatment 129 of 1,376 (9.4%), control 143 of 1,105 (12.9%), relatedness (yes + unsure) 19.6% (treatment) 27.3% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, brain fog.
risk of PASC, 32.9% lower, RR 0.67, p = 0.08, treatment 37 of 1,187 (3.1%), control 33 of 874 (3.8%), relatedness (yes + unsure) 7.1% (treatment) 9.2% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, confusion.
risk of PASC, 41.9% lower, RR 0.58, p < 0.001, treatment 78 of 1,278 (6.1%), control 81 of 964 (8.4%), relatedness (yes + unsure) 12.5% (treatment) 15.7% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, headache.
risk of PASC, 45.4% lower, RR 0.55, p = 0.001, treatment 53 of 1,254 (4.2%), control 54 of 955 (5.7%), relatedness (yes + unsure) 11.5% (treatment) 15.8% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, dizziness.
risk of PASC, 73.5% lower, RR 0.27, p = 0.10, treatment 3 of 1,105 (0.3%), control 3 of 802 (0.4%), relatedness (yes + unsure) 0.4% (treatment) 1.1% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, fainting.
risk of PASC, 38.7% lower, RR 0.61, p = 0.02, treatment 42 of 1,200 (3.5%), control 39 of 894 (4.4%), relatedness (yes + unsure) 7.9% (treatment) 10.7% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, numbness.
risk of PASC, 42.1% lower, RR 0.58, p < 0.001, treatment 123 of 1,288 (9.5%), control 121 of 982 (12.3%), relatedness (yes + unsure) 13.9% (treatment) 18.5% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, sleeping problems.
risk of PASC, 11.8% lower, RR 0.88, p = 0.37, treatment 99 of 1,180 (8.4%), control 88 of 967 (9.1%), relatedness (yes + unsure) 15.1% (treatment) 16.1% (control) , NNT 141, adjusted per study and for relatedness, moderate/major symptoms at 12 months, body pains.
risk of PASC, 32.2% lower, RR 0.68, p < 0.001, treatment 136 of 1,220 (11.1%), control 146 of 1,034 (14.1%), relatedness (yes + unsure) 16.1% (treatment) 19.0% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, joint pains.
risk of PASC, 30.2% lower, RR 0.70, p < 0.001, treatment 205 of 1,398 (14.7%), control 209 of 1,181 (17.7%), relatedness (yes + unsure) 24.1% (treatment) 28.3% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, fatigue.
risk of PASC, 31.9% lower, RR 0.68, p = 0.006, treatment 91 of 1,208 (7.5%), control 89 of 997 (8.9%), relatedness (yes + unsure) 14.5% (treatment) 18.1% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, weakness.
risk of PASC, 35.9% lower, RR 0.64, p < 0.001, treatment 99 of 1,211 (8.2%), control 107 of 1,002 (10.7%), relatedness (yes + unsure) 14.3% (treatment) 17.4% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, generally unwell.
risk of PASC, 9.2% lower, RR 0.91, p = 0.81, treatment 16 of 1,066 (1.5%), control 11 of 855 (1.3%), relatedness (yes + unsure) 3.0% (treatment) 3.8% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, fever.
risk of PASC, 42.4% lower, RR 0.58, p = 0.21, treatment 10 of 1,065 (0.9%), control 11 of 853 (1.3%), relatedness (yes + unsure) 3.0% (treatment) 3.7% (control) , adjusted per study and for relatedness, moderate/major symptoms at 12 months, rashes.
risk of PASC, 10.0% lower, RR 0.90, p = 0.62, treatment 47 of 1,090 (4.3%), control 41 of 873 (4.7%), NNT 260, adjusted per study, moderate/major symptoms at 12 months, other.
Hazan (B), 7/7/2021, retrospective, USA, peer-reviewed, 7 authors, average treatment delay 9.2 days, dosage 12mg days 1, 4, 8, this trial uses multiple treatments in the treatment arm (combined with doxycycline, zinc, vitamin D, vitamin C) - results of individual treatments may vary, trial NCT04949230 (history), excluded in exclusion analyses: study uses a synthetic control arm. risk of death, 86.2% lower, RR 0.14, p = 0.04, NNT 6.9, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of hospitalization, 93.5% lower, RR 0.07, p = 0.001, NNT 3.3, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), primary outcome.
Huvemek, 3/25/2021, Double Blind Randomized Controlled Trial, Bulgaria, preprint, 1 author, average treatment delay 7.0 days, dosage 400μg/kg days 1-3, trial EudraCT2020-002091-12. risk of no improvement, 31.6% lower, RR 0.68, p = 0.28, treatment 13 of 50 (26.0%), control 19 of 50 (38.0%), NNT 8.3, patients with improvement on WHO scale, day 7.
risk of no improvement, 31.8% lower, RR 0.68, p = 0.21, treatment 15 of 50 (30.0%), control 22 of 50 (44.0%), NNT 7.1, patients with improvement on WHO scale, day 6.
risk of no improvement, 36.0% lower, RR 0.64, p = 0.10, treatment 16 of 50 (32.0%), control 25 of 50 (50.0%), NNT 5.6, patients with improvement on WHO scale, day 5.
risk of no improvement, 34.5% lower, RR 0.66, p = 0.07, treatment 19 of 50 (38.0%), control 29 of 50 (58.0%), NNT 5.0, patients with improvement on WHO scale, day 4.
Jamir, 12/13/2021, retrospective, India, peer-reviewed, 6 authors, study period June 2020 - October 2020, dosage not specified. risk of death, 53.0% higher, RR 1.53, p = 0.13, treatment 32 of 76 (42.1%), control 69 of 190 (36.3%), adjusted per study, multivariable Cox regression.
Khan, 9/24/2020, retrospective, Bangladesh, peer-reviewed, median age 35.0, 8 authors, dosage 12mg single dose. risk of death, 87.1% lower, RR 0.13, p = 0.02, treatment 1 of 115 (0.9%), control 9 of 133 (6.8%), NNT 17.
risk of ICU admission, 89.5% lower, RR 0.11, p = 0.007, treatment 1 of 115 (0.9%), control 11 of 133 (8.3%), NNT 14.
risk of progression, 83.5% lower, RR 0.17, p < 0.001, treatment 3 of 115 (2.6%), control 21 of 133 (15.8%), NNT 7.6.
risk of no recovery, 87.1% lower, RR 0.13, p = 0.02, treatment 1 of 115 (0.9%), control 9 of 133 (6.8%), NNT 17.
hospitalization time, 40.0% lower, relative time 0.60, p < 0.001, treatment 115, control 133.
time to viral-, 73.3% lower, relative time 0.27, p < 0.001, treatment 115, control 133.
Kishoria, 8/31/2020, Randomized Controlled Trial, India, peer-reviewed, 7 authors, dosage 12mg single dose, excluded in exclusion analyses: excessive unadjusted differences between groups. risk of no hospital discharge, 7.5% higher, RR 1.08, p = 1.00, treatment 11 of 19 (57.9%), control 7 of 13 (53.8%).
risk of no viral clearance, 7.5% higher, RR 1.08, p = 1.00, treatment 11 of 19 (57.9%), control 7 of 13 (53.8%), day 3, primary outcome.
risk of no viral clearance, 220.0% higher, RR 3.20, p = 0.45, treatment 1 of 5 (20.0%), control 0 of 6 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), day 5.
Lim, 11/3/2021, Randomized Controlled Trial, Malaysia, peer-reviewed, 26 authors, study period 31 May, 2021 - 9 October, 2021, average treatment delay 5.1 days, dosage 400μg/kg days 1-5, trial NCT04920942 (history) (I-TECH). risk of death, 69.0% lower, RR 0.31, p = 0.09, treatment 3 of 241 (1.2%), control 10 of 249 (4.0%), NNT 36.
risk of death, 75.2% lower, RR 0.25, p = 0.02, treatment 3 of 52 (5.8%), control 10 of 43 (23.3%), NNT 5.7, among patients progressing to severe cases (mostly before treatment ended).
risk of mechanical ventilation, 59.0% lower, RR 0.41, p = 0.17, treatment 4 of 241 (1.7%), control 10 of 249 (4.0%), NNT 42.
risk of ICU admission, 22.0% lower, RR 0.78, p = 0.79, treatment 6 of 241 (2.5%), control 8 of 249 (3.2%), NNT 138.
risk of progression, 31.1% lower, RR 0.69, p = 0.29, treatment 14 of 241 (5.8%), control 21 of 249 (8.4%), NNT 38, death/IMV/NIV/high flow (WHO severe cases).
risk of progression, 25.0% higher, RR 1.25, p = 0.25, treatment 52 of 241 (21.6%), control 43 of 249 (17.3%), primary outcome.
hospitalization time, 5.5% higher, relative time 1.05, p = 0.38, treatment 241, control 249.
risk of no recovery, 2.5% higher, RR 1.02, p = 0.86, treatment 116 of 241 (48.1%), control 116 of 247 (47.0%), day 5.
Lima-Morales, 2/10/2021, prospective, Mexico, peer-reviewed, 9 authors, average treatment delay 7.2 days, dosage 12mg single dose, this trial uses multiple treatments in the treatment arm (combined with azithromycin, montelukast, and aspirin) - results of individual treatments may vary. risk of death, 77.7% lower, RR 0.22, p < 0.001, treatment 15 of 481 (3.1%), control 52 of 287 (18.1%), NNT 6.7, adjusted per study, odds ratio converted to relative risk, multivariate.
risk of mechanical ventilation, 51.9% lower, RR 0.48, p = 0.15, treatment 8 of 434 (1.8%), control 11 of 287 (3.8%), NNT 50.
risk of hospitalization, 67.4% lower, RR 0.33, p < 0.001, treatment 44 of 481 (9.1%), control 89 of 287 (31.0%), NNT 4.6, adjusted per study, odds ratio converted to relative risk, multivariate.
risk of no recovery, 58.6% lower, RR 0.41, p < 0.001, treatment 75 of 481 (15.6%), control 118 of 287 (41.1%), NNT 3.9, adjusted per study, inverted to make RR<1 favor treatment, odds ratio converted to relative risk, recovery at day 14 after symptoms, multivariate.
Llenas-García, 5/10/2023, retrospective, Spain, peer-reviewed, 11 authors, study period 23 February, 2020 - 14 March, 2021, average treatment delay 6.0 days, dosage 200μg/kg single dose. risk of death, 16.7% lower, RR 0.83, p = 0.82, treatment 10 of 96 (10.4%), control 12 of 96 (12.5%), NNT 48.
risk of oxygen therapy, 18.4% lower, RR 0.82, p = 0.37, treatment 31 of 96 (32.3%), control 38 of 96 (39.6%), NNT 14.
risk of progression, 23.0% lower, OR 0.77, p = 0.52, treatment 96, control 96, adjusted per study, in-hospital mortality or respiratory support, multivariable, primary outcome, RR approximated with OR.
risk of ICU admission, 4.0% higher, OR 1.04, p = 0.92, treatment 96, control 96, adjusted per study, multivariable, RR approximated with OR.
Munir, 4/21/2023, retrospective, Pakistan, peer-reviewed, 3 authors, study period March 2021 - March 2022, dosage not specified. risk of death, 48.2% lower, OR 0.52, p = 0.13, treatment 92, control 908, adjusted per study, multivariable, RR approximated with OR.
Mustafa, 12/29/2021, retrospective, Pakistan, peer-reviewed, 7 authors, dosage varies, excluded in exclusion analyses: unadjusted results with no group details. risk of death, 63.7% lower, RR 0.36, p = 0.09, treatment 3 of 73 (4.1%), control 42 of 371 (11.3%), NNT 14.
Naggie, 6/12/2022, Double Blind Randomized Controlled Trial, placebo-controlled, USA, peer-reviewed, 28 authors, study period 23 June, 2021 - 4 February, 2022, average treatment delay 6.0 days, dosage 600μg/kg days 1-6, trial NCT04885530 (history) (ACTIV-6). risk of death, 200.3% higher, RR 3.00, p = 0.50, treatment 1 of 602 (0.2%), control 0 of 604 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), 600µg/kg, day 28.
risk of death, 194.7% higher, RR 2.95, p = 1.00, treatment 1 of 817 (0.1%), control 0 of 774 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), 400µg/kg, day 28.
risk of death/hospitalization, 151.7% higher, RR 2.52, p = 0.29, treatment 5 of 600 (0.8%), control 2 of 604 (0.3%), 600µg/kg, day 28.
risk of hospitalization, 5.3% higher, RR 1.05, p = 1.00, treatment 10 of 817 (1.2%), control 9 of 774 (1.2%), 400µg/kg, day 28.
risk of progression, 68.4% lower, RR 0.32, p = 0.36, treatment 1 of 817 (0.1%), control 3 of 774 (0.4%), NNT 377, 400µg/kg, aggravated C19 pneumonia, eTable 2.
risk of progression, no change, RR 1.00, p = 0.53, treatment 34 of 600 (5.7%), control 36 of 604 (6.0%), NNT 341, adjusted per study, urgent or emergency care visits, hospitalizations, or death, 600µg/kg.
risk of progression, 20.0% higher, RR 1.20, p = 0.32, treatment 32 of 817 (3.9%), control 28 of 774 (3.6%), adjusted per study, urgent or emergency care visits, hospitalizations, or death, 400µg/kg.
clinical progression, 61.0% higher, OR 1.61, p = 0.07, treatment 600, control 604, mid-recovery, 600µg/kg, day 7, RR approximated with OR.
clinical progression, 114.0% higher, OR 2.14, p = 0.04, treatment 600, control 604, 600µg/kg, day 14, RR approximated with OR.
clinical progression, 161.0% higher, OR 2.61, p = 0.04, treatment 600, control 604, 600µg/kg, day 28, RR approximated with OR.
clinical progression, 24.0% lower, OR 0.76, p = 0.07, treatment 817, control 774, mid-recovery, 400µg/kg, day 7, RR approximated with OR.
clinical progression, 27.0% lower, OR 0.73, p = 0.05, treatment 817, control 774, 400µg/kg, day 14, RR approximated with OR.
clinical progression, 10.0% lower, OR 0.90, p = 0.57, treatment 817, control 774, 400µg/kg, day 28, RR approximated with OR.
time to recovery, 2.0% lower, HR 0.98, p = 0.72, treatment 600, control 604, inverted to make HR<1 favor treatment, 600µg/kg, post-hoc primary outcome.
risk of limited activity, 30.5% lower, RR 0.70, p = 0.14, treatment 30 of 805 (3.7%), control 41 of 765 (5.4%), NNT 61, eFigure 2, 400µg/kg, day 14.
time to recovery, 6.5% lower, HR 0.93, p = 0.18, treatment 817, control 774, inverted to make HR<1 favor treatment, 400µg/kg, post-hoc primary outcome.
time to recovery, 46.2% lower, HR 0.54, p = 0.02, treatment 39, control 51, inverted to make HR<1 favor treatment, 400µg/kg, severe, recovery.
Ochoa-Jaramillo, 10/21/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Colombia, peer-reviewed, 8 authors, study period 10 December, 2020 - 9 December, 2021, average treatment delay 8.8 days, dosage 400μg/kg single dose, trial NCT04602507 (history). risk of death, 57.0% lower, HR 0.43, p = 0.35, treatment 2 of 37 (5.4%), control 4 of 38 (10.5%), NNT 20, Cox proportional hazards.
risk of mechanical ventilation, 34.0% higher, HR 1.34, p = 0.62, treatment 7 of 37 (18.9%), control 5 of 38 (13.2%), Cox proportional hazards.
risk of ICU admission, 37.0% higher, HR 1.37, p = 0.52, treatment 8 of 37 (21.6%), control 6 of 38 (15.8%), Cox proportional hazards, primary outcome.
Okumuş, 1/12/2021, Double Blind Randomized Controlled Trial, Turkey, peer-reviewed, 15 authors, study period May 2020 - September 2020, dosage 200μg/kg days 1-5, 36-50kg - 9mg, 51-65kg - 12mg, 66-79kg - 15mg, >80kg 200μg/kg, trial NCT04646109 (history). risk of death, 33.3% lower, RR 0.67, p = 0.55, treatment 6 of 30 (20.0%), control 9 of 30 (30.0%), NNT 10.
risk of no improvement at day 10, 42.9% lower, RR 0.57, p = 0.18, treatment 8 of 30 (26.7%), control 14 of 30 (46.7%), NNT 5.0.
risk of no improvement at day 5, 15.8% lower, RR 0.84, p = 0.60, treatment 16 of 30 (53.3%), control 19 of 30 (63.3%), NNT 10, primary outcome.
risk of no viral clearance, 80.0% lower, RR 0.20, p = 0.02, treatment 2 of 16 (12.5%), control 5 of 8 (62.5%), NNT 2.0, day 10.
Osati, 7/16/2023, retrospective, Tanzania, preprint, median age 60.0, 22 authors, study period 26 March, 2021 - 30 July, 2022. risk of death, 31.5% lower, OR 0.68, p = 0.02, treatment 448, control 849, adjusted per study, inverted to make OR<1 favor treatment, multivariable, RR approximated with OR.
Ozer, 11/23/2021, prospective, USA, peer-reviewed, 12 authors, dosage 200μg/kg days 1, 3. risk of death, 75.0% lower, RR 0.25, p = 0.09, treatment 2 of 60 (3.3%), control 8 of 60 (13.3%), NNT 10.0, PSM.
risk of mechanical ventilation, 12.6% lower, RR 0.87, p = 0.20, treatment 3 of 60 (5.0%), control 2 of 60 (3.3%), adjusted per study, odds ratio converted to relative risk, propensity score matching, multivariable.
ventilation time, 83.3% lower, relative time 0.17, p = 0.002, treatment 60, control 60.
risk of ICU admission, 48.7% lower, RR 0.51, p = 0.42, treatment 6 of 60 (10.0%), control 3 of 60 (5.0%), adjusted per study, odds ratio converted to relative risk, propensity score matching, multivariable.
ICU time, 70.6% lower, relative time 0.29, p < 0.001, treatment 60, control 60.
hospitalization time, 9.0% higher, relative time 1.09, p = 0.09, treatment 60, control 60, adjusted per study, propensity score matching, multivariable.
Podder, 9/3/2020, Randomized Controlled Trial, Bangladesh, peer-reviewed, 4 authors, study period 1 May, 2020 - 31 July, 2020, average treatment delay 7.0 days, dosage 200μg/kg single dose. recovery time from enrollment, 16.1% lower, relative time 0.84, p = 0.34, treatment 32, control 30.
Pott-Junior, 3/9/2021, Randomized Controlled Trial, Brazil, peer-reviewed, 10 authors, study period 1 July, 2020 - 1 December, 2020, average treatment delay 8.0 days, dosage 200μg/kg single dose, dose varies in three arms 100, 200, 400μg/kg, trial NCT04431466 (history). risk of mechanical ventilation, 85.2% lower, RR 0.15, p = 0.25, treatment 1 of 27 (3.7%), control 1 of 4 (25.0%), NNT 4.7.
risk of ICU admission, 85.2% lower, RR 0.15, p = 0.25, treatment 1 of 27 (3.7%), control 1 of 4 (25.0%), NNT 4.7.
relative improvement in Ct value, 0.8% better, RR 0.99, p = 1.00, treatment 27, control 3.
risk of no viral clearance, 11.1% higher, RR 1.11, p = 1.00, treatment 10 of 27 (37.0%), control 1 of 3 (33.3%), primary outcome.
Qadeer, 8/31/2022, prospective, Pakistan, peer-reviewed, median age 55.4, 6 authors, study period 1 November, 2020 - 30 May, 2021, dosage 12mg days 1-5, excluded in exclusion analyses: minimal baseline details provided. risk of no viral clearance, 58.3% lower, RR 0.42, p < 0.001, treatment 35 of 105 (33.3%), control 84 of 105 (80.0%), NNT 2.1, mid-recovery, day 10.
risk of no viral clearance, 20.0% lower, RR 0.80, p < 0.001, treatment 84 of 105 (80.0%), control 105 of 105 (100.0%), NNT 5.0, day 7.
risk of no viral clearance, 98.6% lower, RR 0.01, p < 0.001, treatment 0 of 105 (0.0%), control 35 of 105 (33.3%), NNT 3.0, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), day 14.
Rajter, 10/13/2020, retrospective, propensity score matching, USA, peer-reviewed, 6 authors, dosage 200μg/kg single dose. risk of death, 46.0% lower, RR 0.54, p = 0.045, treatment 13 of 98 (13.3%), control 24 of 98 (24.5%), NNT 8.9, adjusted per study, odds ratio converted to relative risk, PSM.
risk of death, 66.9% lower, RR 0.33, p = 0.03, treatment 26 of 173 (15.0%), control 27 of 107 (25.2%), NNT 9.8, adjusted per study, odds ratio converted to relative risk, multivariate, primary outcome.
risk of mechanical ventilation, 63.6% lower, RR 0.36, p = 0.10, treatment 4 of 98 (4.1%), control 11 of 98 (11.2%), NNT 14, matched cohort excluding intubated at baseline.
Ravikirti, 4/6/2022, retrospective, India, preprint, 7 authors, study period 1 April, 2021 - 15 May, 2021, dosage varies, excluded in exclusion analyses: exclusion of patients in less severe condition, data/analysis concerns. risk of death, 2.8% lower, RR 0.97, p = 0.82, treatment 53 of 171 (31.0%), control 254 of 794 (32.0%), NNT 100, odds ratio converted to relative risk.
Rezai (B), 6/16/2022, Double Blind Randomized Controlled Trial, placebo-controlled, Iran, peer-reviewed, mean age 53.8, 29 authors, study period 19 February, 2021 - 14 August, 2021, average treatment delay 7.18 days, dosage 400μg/kg days 1-3, trial IRCT20111224008507N5, excluded in exclusion analyses: multiple critical issues, see study page. risk of death, 30.8% lower, RR 0.69, p = 0.36, treatment 13 of 311 (4.2%), control 18 of 298 (6.0%), NNT 54.
risk of mechanical ventilation, 50.0% lower, RR 0.50, p = 0.07, treatment 311, control 298.
risk of ICU admission, 16.0% lower, RR 0.84, p = 0.47, treatment 311, control 298.
hospitalization time, 11.5% higher, relative time 1.11, p = 0.009, treatment mean 7.98 (±4.4) n=311, control mean 7.16 (±3.2) n=298.
deterioration, 12.7% higher, RR 1.13, p = 0.74, treatment 20 of 311 (6.4%), control 17 of 298 (5.7%).
risk of no recovery, 24.2% lower, RR 0.76, p = 0.02, treatment 311, control 298, inverted to make RR<1 favor treatment, post-hoc primary outcome.
risk of no recovery, 64.0% lower, RR 0.36, p = 0.06, treatment 5 of 145 (3.4%), control 10 of 105 (9.5%), NNT 16, day 7, cough.
risk of no recovery, 76.0% lower, RR 0.24, p = 0.38, day 7, tachypnea.
Rezk, 10/30/2021, prospective, Egypt, peer-reviewed, 4 authors, dosage 36mg days 1, 3, 6. risk of death, 80.0% lower, RR 0.20, p = 0.50, treatment 0 of 160 (0.0%), control 2 of 160 (1.2%), NNT 80, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of progression, 55.6% lower, RR 0.44, p = 0.06, treatment 8 of 160 (5.0%), control 18 of 160 (11.2%), NNT 16, 2 weeks, including deaths.
risk of no recovery, 33.4% lower, RR 0.67, p = 0.27, treatment 14 of 145 (9.7%), control 20 of 138 (14.5%), NNT 21, 4 weeks, more patients were lost to followup in the control group.
time to viral-, 27.3% lower, relative time 0.73, p = 0.01, treatment 160, control 160.
Sarojvisut, 12/12/2022, Randomized Controlled Trial, Thailand, peer-reviewed, 8 authors, study period 1 October, 2021 - 31 May, 2022, dosage 400μg/kg days 1-5, trial TCTR20220427005. risk of ICU admission, 103.8% higher, RR 2.04, p = 0.62, treatment 2 of 157 (1.3%), control 1 of 160 (0.6%).
risk of no improvement, 103.8% higher, RR 2.04, p = 0.62, treatment 2 of 157 (1.3%), control 1 of 160 (0.6%).
recovery time, 3.8% lower, relative time 0.96, p = 0.63, treatment 157, control 160.
Shahbaznejad, 1/19/2021, Double Blind Randomized Controlled Trial, Iran, peer-reviewed, 8 authors, average treatment delay 6.29 days, dosage 200μg/kg single dose, trial IRCT20111224008507N3. risk of death, 197.1% higher, RR 2.97, p = 1.00, treatment 1 of 35 (2.9%), control 0 of 34 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), patient died within 24 hours of admission.
risk of mechanical ventilation, 94.3% higher, RR 1.94, p = 1.00, treatment 2 of 35 (5.7%), control 1 of 34 (2.9%).
recovery time, 31.6% lower, relative time 0.68, p = 0.048, treatment 35, control 34, duration of dsypnea.
recovery time, 19.2% lower, relative time 0.81, p = 0.02, treatment 35, control 34, duration of all symptoms, primary outcome.
hospitalization time, 15.5% lower, relative time 0.85, p = 0.02, treatment 35, control 34.
Shimizu, 12/31/2021, retrospective, Japan, peer-reviewed, 11 authors, study period December 2020 - May 2021, dosage 200μg/kg days 1, 14. risk of death, 99.9% lower, HR 0.001, p < 0.001, treatment 0 of 39 (0.0%), control 8 of 49 (16.3%), NNT 6.1, adjusted per study, Cox proportional hazard regression.
ventilator free days, 47.9% lower, OR 0.52, p = 0.03, treatment 39, control 49, adjusted per study, inverted to make OR<1 favor treatment, proportional odds logistic regression, primary outcome, RR approximated with OR.
ventilation time, 38.5% lower, relative time 0.62, p < 0.001, treatment 39, control 49.
ICU free days, 42.8% lower, OR 0.57, p = 0.06, treatment 39, control 49, adjusted per study, inverted to make OR<1 favor treatment, proportional odds logistic regression, RR approximated with OR.
ICU time, 37.5% lower, relative time 0.62, p < 0.001, treatment 39, control 49.
GI complications while ventilated, 77.9% lower, RR 0.22, p = 0.03, treatment 39, control 49, adjusted per study, Cox proportional hazard regression.
Soto, 3/2/2022, retrospective, Peru, peer-reviewed, median age 58.0, 10 authors, study period April 2020 - August 2020, dosage not specified, excluded in exclusion analyses: substantial unadjusted confounding by indication likely; substantial confounding by time possible due to significant changes in SOC and treatment propensity near the start of the pandemic. risk of death, 41.0% higher, HR 1.41, p = 0.001, treatment 280 of 484 (57.9%), control 374 of 934 (40.0%), adjusted per study, multivariable.
Soto-Becerra, 10/8/2020, retrospective, database analysis, Peru, preprint, median age 59.4, 4 authors, study period 1 April, 2020 - 19 July, 2020, dosage 200μg/kg single dose, excluded in exclusion analyses: substantial unadjusted confounding by indication likely; includes PCR+ patients that may be asymptomatic for COVID-19 but in hospital for other reasons. risk of death, 17.1% lower, HR 0.83, p = 0.01, treatment 92 of 203 (45.3%), control 1,438 of 2,630 (54.7%), NNT 11, IVM vs. control day 43 (last day available) weighted KM from figure 3, per the pre-specified rules, the last available day mortality results have priority.
risk of death, 39.0% higher, HR 1.39, p = 0.16, treatment 47 of 203 (23.2%), control 401 of 2,630 (15.2%), adjusted per study, day 30, Table 2, IVM wHR, primary outcome.
Spoorthi, 11/14/2020, prospective, India, peer-reviewed, 2 authors, dosage not specified, this trial uses multiple treatments in the treatment arm (combined with doxycycline) - results of individual treatments may vary. recovery time, 21.1% lower, relative time 0.79, p = 0.03, treatment 50, control 50.
hospitalization time, 15.5% lower, relative time 0.84, p = 0.01, treatment 50, control 50.
Thairu, 2/25/2022, retrospective, Nigeria, peer-reviewed, mean age 41.7, 6 authors, study period April 2021 - November 2021, dosage 200μg/kg days 1-5, excluded in exclusion analyses: significant confounding by time possible due to separation of groups in different time periods. risk of death, 87.9% lower, RR 0.12, p = 0.12, treatment 0 of 21 (0.0%), control 4 of 26 (15.4%), NNT 6.5, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), propensity score matching.
risk of death, 93.0% lower, RR 0.07, p = 0.007, treatment 0 of 61 (0.0%), control 4 of 26 (15.4%), NNT 6.5, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), all patients.
time to discharge, 54.6% lower, relative time 0.45, p < 0.001, treatment 61, control 26, propensity score matching.
risk of no viral clearance, 94.8% lower, RR 0.05, p = 0.001, treatment 0 of 21 (0.0%), control 10 of 26 (38.5%), NNT 2.6, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), propensity score matching, day 21.
risk of no viral clearance, 95.2% lower, RR 0.05, p < 0.001, treatment 1 of 21 (4.8%), control 26 of 26 (100.0%), NNT 1.1, propensity score matching, day 14.
risk of no viral clearance, 28.6% lower, RR 0.71, p = 0.005, treatment 15 of 21 (71.4%), control 26 of 26 (100.0%), NNT 3.5, propensity score matching, day 5.
Wada, 5/22/2023, Double Blind Randomized Controlled Trial, placebo-controlled, Japan, peer-reviewed, 26 authors, study period August 2020 - October 2021, average treatment delay 6.6 days, dosage 200μg/kg single dose, trial NCT04703205 (history). risk of progression, 19.0% lower, RR 0.81, p = 0.46, treatment 19 of 106 (17.9%), control 23 of 106 (21.7%), NNT 26, adjusted per study, odds ratio converted to relative risk.
risk of progression, 27.1% lower, RR 0.73, p = 0.47, treatment 7 of 28 (25.0%), control 9 of 29 (31.0%), NNT 17, adjusted per study, odds ratio converted to relative risk, pneumonia for patients w/o pneumonia at baseline.
risk of oxygen therapy, 14.3% higher, RR 1.14, p = 0.46, treatment 22 of 106 (20.8%), control 19 of 106 (17.9%), adjusted per study, odds ratio converted to relative risk.
improvement, 23.5% worse, OR 1.23, p = 0.61, treatment 106, control 106, adjusted per study, inverted to make OR<1 favor treatment, RR approximated with OR.
risk of no recovery, 60.0% lower, RR 0.40, p = 0.17, treatment 4 of 107 (3.7%), control 10 of 107 (9.3%), NNT 18, day 15, dyspnea.
risk of no recovery, 20.0% lower, RR 0.80, p = 1.00, treatment 4 of 107 (3.7%), control 5 of 107 (4.7%), NNT 107, day 15, headache.
risk of no recovery, no change, RR 1.00, p = 1.00, treatment 7 of 107 (6.5%), control 7 of 107 (6.5%), day 15, sore throat.
risk of no recovery, 18.2% lower, RR 0.82, p = 0.81, treatment 9 of 107 (8.4%), control 11 of 107 (10.3%), NNT 53, day 15, nasal discharge.
risk of no recovery, 3.8% higher, RR 1.04, p = 1.00, treatment 27 of 107 (25.2%), control 26 of 107 (24.3%), day 15, cough.
risk of no recovery, no change, RR 1.00, p = 1.00, treatment 18 of 107 (16.8%), control 18 of 107 (16.8%), day 15, sputum.
risk of no recovery, 20.0% lower, RR 0.80, p = 1.00, treatment 4 of 107 (3.7%), control 5 of 107 (4.7%), NNT 107, day 15, diarhhea.
risk of no recovery, 66.7% higher, RR 1.67, p = 0.72, treatment 5 of 107 (4.7%), control 3 of 107 (2.8%), day 15, myalgia.
risk of no recovery, 1000.0% higher, RR 11.00, p = 0.06, treatment 5 of 107 (4.7%), control 0 of 107 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm), day 15, arthralgia.
risk of no viral clearance, 4.2% higher, HR 1.04, p = 0.79, treatment 106, control 106, inverted to make HR<1 favor treatment, Kaplan–Meier, primary outcome.
Zubair, 1/18/2022, retrospective, Pakistan, peer-reviewed, 8 authors, study period October 2020 - February 2021, dosage 12mg single dose, excluded in exclusion analyses: substantial unadjusted confounding by indication likely; unadjusted results with no group details. risk of death, 9.0% higher, RR 1.09, p = 1.00, treatment 5 of 90 (5.6%), control 5 of 98 (5.1%), unadjusted.
hospitalization time, 8.0% higher, relative time 1.08, p = 0.40, treatment 90, control 98, unadjusted, Table 3, mean number of days.
Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. Only the first (most serious) outcome is used in pooled analysis, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.
Alam, 12/15/2020, prospective, Bangladesh, peer-reviewed, 13 authors, dosage 12mg monthly. risk of case, 90.6% lower, RR 0.09, p < 0.001, treatment 4 of 58 (6.9%), control 44 of 60 (73.3%), NNT 1.5.
Behera (B), 2/15/2021, prospective, India, peer-reviewed, 14 authors, dosage 300μg/kg days 1, 4. risk of case, 83.0% lower, RR 0.17, p < 0.001, treatment 45 of 2,199 (2.0%), control 133 of 1,147 (11.6%), NNT 10, two doses, primary outcome.
Behera, 11/3/2020, retrospective, India, peer-reviewed, 13 authors, dosage 300μg/kg days 1, 4. risk of case, 53.8% lower, RR 0.46, p < 0.001, treatment 41 of 117 (35.0%), control 145 of 255 (56.9%), NNT 4.6, adjusted per study, odds ratio converted to relative risk, model 2 2+ doses conditional logistic regression.
Bernigaud, 11/28/2020, retrospective, France, peer-reviewed, 12 authors, dosage 200μg/kg days 1, 8, 15, 400μg/kg days 1, 8, 15, two different dosages. risk of death, 99.4% lower, RR 0.006, p = 0.08, treatment 0 of 69 (0.0%), control 150 of 3,062 (4.9%), NNT 20, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of case, 55.1% lower, RR 0.45, p = 0.01, treatment 7 of 69 (10.1%), control 692 of 3,062 (22.6%), NNT 8.0.
Carvallo, 11/17/2020, prospective, Argentina, peer-reviewed, 4 authors, dosage 12mg weekly, this trial uses multiple treatments in the treatment arm (combined with iota-carrageenan) - results of individual treatments may vary, excluded in exclusion analyses: concern about potential data issues. risk of case, 99.9% lower, RR 0.001, p < 0.001, treatment 0 of 788 (0.0%), control 237 of 407 (58.2%), NNT 1.7, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
Carvallo (B), 10/19/2020, prospective, Argentina, preprint, 1 author, dosage 1mg days 1-14, this trial uses multiple treatments in the treatment arm (combined with iota-carrageenan) - results of individual treatments may vary, trial NCT04425850 (history), excluded in exclusion analyses: concern about potential data issues. risk of case, 96.3% lower, RR 0.04, p < 0.001, treatment 0 of 131 (0.0%), control 11 of 98 (11.2%), NNT 8.9, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
Chahla (B), 1/11/2021, Randomized Controlled Trial, Argentina, peer-reviewed, 11 authors, study period 15 October, 2020 - 31 December, 2020, dosage 12mg weekly, this trial uses multiple treatments in the treatment arm (combined with iota-carrageenan) - results of individual treatments may vary, trial NCT04701710 (history). risk of moderate/severe case, 95.2% lower, RR 0.05, p = 0.002, treatment 0 of 117 (0.0%), control 10 of 117 (8.5%), NNT 12, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), moderate/severe COVID-19.
risk of case, 84.0% lower, RR 0.16, p = 0.004, treatment 4 of 117 (3.4%), control 25 of 117 (21.4%), NNT 5.6, adjusted per study, odds ratio converted to relative risk, all cases, primary outcome.
Desort-Henin, 1/5/2023, Double Blind Randomized Controlled Trial, placebo-controlled, Bulgaria, preprint, 5 authors, study period March 2022 - October 2022, dosage 200μg/kg day 1, 100μg/kg days 2-28, trial NCT05305560 (history) (SAIVE). risk of case with high viral load, 96.0% lower, RR 0.04, p < 0.001, treatment 4 of 200 (2.0%), control 99 of 199 (49.7%), NNT 2.1.
risk of case, 71.6% lower, RR 0.28, p < 0.001, treatment 30 of 200 (15.0%), control 105 of 199 (52.8%), NNT 2.6, primary outcome.
Hellwig, 11/28/2020, retrospective, ecological study, multiple countries, peer-reviewed, 2 authors, dosage 200μg/kg, dose varied, typically 150-200μg/kg, excluded in exclusion analyses: not a typical trial, analysis of African countries that used or did not use ivermectin prophylaxis for parasitic infections. risk of case, 78.0% lower, RR 0.22, p < 0.02, African countries, PCTI vs. no PCT, relative cases per capita.
IVERCOR PREP, 12/20/2020, retrospective, Argentina, preprint, 1 author, dosage 12mg weekly, excluded in exclusion analyses: minimal details provided. risk of case, 73.4% lower, RR 0.27, p < 0.001, treatment 13 of 389 (3.3%), control 61 of 486 (12.6%), NNT 11.
Kerr, 12/11/2021, retrospective, propensity score matching, Brazil, peer-reviewed, 9 authors, study period July 2020 - December 2020, dosage 200μg/kg days 1, 2, 16, 17, 0.2mg/kg/day for 2 days every 15 days. risk of death, 70.0% lower, RR 0.30, p < 0.001, treatment 25 of 3,034 (0.8%), control 79 of 3,034 (2.6%), NNT 56, adjusted per study, multivariate linear regression, propensity score matching.
risk of hospitalization, 67.0% lower, RR 0.33, p < 0.001, treatment 44 of 3,034 (1.5%), control 99 of 3,034 (3.3%), adjusted per study, multivariate linear regression, propensity score matching.
risk of case, 44.5% lower, RR 0.56, p < 0.001, treatment 4,197 of 113,845 (3.7%), control 3,034 of 45,716 (6.6%), NNT 34.
Mondal, 5/31/2021, retrospective, India, peer-reviewed, 11 authors, dosage not specified. risk of symptomatic case, 87.9% lower, RR 0.12, p = 0.006, treatment 128, control 1,342, adjusted per study, odds ratio converted to relative risk, control prevalence approximated with overall prevalence, multivariable, primary outcome.
Moraes, 4/30/2021, Randomized Controlled Trial, this trial compares with another treatment - results may be better when compared to placebo, trial NCT04384458 (history). Estimated 400 patient RCT with results unknown and over 2 years late.
Morgenstern, 4/16/2021, retrospective, propensity score matching, Dominican Republic, peer-reviewed, 16 authors, dosage 200μg/kg weekly, trial NCT04832945 (history). risk of hospitalization, 80.0% lower, RR 0.20, p = 0.50, treatment 0 of 271 (0.0%), control 2 of 271 (0.7%), NNT 136, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm), PSM.
risk of case, 74.0% lower, RR 0.26, p = 0.008, treatment 5 of 271 (1.8%), control 18 of 271 (6.6%), NNT 21, adjusted per study, PSM, multivariate Cox regression, primary outcome.
PREVENT-COVID, 6/24/2023, Double Blind Randomized Controlled Trial, placebo-controlled, trial NCT05060666 (history) (PREVENT-COVID). Estimated 412 patient RCT with results unknown and over 2 years late.
Samajdar, 11/17/2021, retrospective, India, peer-reviewed, 9 authors, study period 1 September, 2020 - 31 December, 2020, dosage not specified, excluded in exclusion analyses: minimal details provided; unadjusted results with no group details; results may be significantly affected by survey bias. risk of case, 79.8% lower, RR 0.20, p < 0.001, treatment 12 of 164 (7.3%), control 29 of 81 (35.8%), NNT 3.5, odds ratio converted to relative risk, physician survey.
risk of case, 48.6% lower, RR 0.51, p = 0.03, treatment 11 of 109 (10.1%), control 39 of 200 (19.5%), NNT 11, odds ratio converted to relative risk, combined ivermectin or HCQ in community.
Seet, 4/14/2021, Cluster Randomized Controlled Trial, Singapore, peer-reviewed, 15 authors, study period 13 May, 2020 - 31 August, 2020, dosage 12mg single dose, 200µg/kg, maximum 12mg, this trial compares with another treatment - results may be better when compared to placebo, trial NCT04446104 (history). risk of symptomatic case, 49.8% lower, RR 0.50, p < 0.001, treatment 32 of 617 (5.2%), control 64 of 619 (10.3%), NNT 19.
risk of case, 5.8% lower, RR 0.94, p = 0.61, treatment 398 of 617 (64.5%), control 433 of 619 (70.0%), NNT 18, adjusted per study, odds ratio converted to relative risk, model 6, primary outcome.
Shouman, 8/28/2020, Randomized Controlled Trial, Egypt, peer-reviewed, 8 authors, study period 1 June, 2020 - 28 July, 2020, dosage 18mg days 1, 3, dose varies depending on weight - 40-60kg: 15mg, 60-80kg: 18mg, >80kg: 24mg, trial NCT04422561 (history). risk of symptomatic case, 91.3% lower, RR 0.09, p < 0.001, treatment 15 of 203 (7.4%), control 59 of 101 (58.4%), NNT 2.0, adjusted per study, inverted to make RR<1 favor treatment, multivariate, primary outcome.
risk of severe case, 92.9% lower, RR 0.07, p = 0.002, treatment 1 of 203 (0.5%), control 7 of 101 (6.9%), NNT 16, unadjusted.
Tanioka, 3/26/2021, retrospective, ecological study, multiple countries, preprint, 3 authors, dosage 200μg/kg, dose varied, typically 150-200μg/kg, excluded in exclusion analyses: not a typical trial, analysis of African countries that used or did not use ivermectin prophylaxis for parasitic infections. risk of death, 88.2% lower, RR 0.12, p = 0.002, relative mean mortality per million.
covid19criticalcare.com, covid19criticalcare.com/wp-content/uploads/2022/03/Statement-on-Together-Trial-WSJ-Article-Mar-18-1.pdf.covid19criticalcare.com/wp-content/uploads/2022/03/Statement-on-Together-Trial-WSJ-Article-Mar-18-1.pdf.
doyourownresearch.substack.com (R), doyourownresearch.substack.com/p/the-story-of-a-real-activ-6-patient.doyourownresearch.substack.com/p/the-story-of-a-real-activ-6-patient.
Please send us corrections, updates, or comments. c19early involves the extraction of 100,000+ datapoints from thousands of papers. Community updates help ensure high accuracy. Vaccines and treatments are complementary. All practical, effective, and safe means should be used based on risk/benefit analysis. No treatment, vaccine, or intervention is 100% available and effective for all current and future variants. We do not provide medical advice. Before taking any medication, consult a qualified physician who can provide personalized advice and details of risks and benefits based on your medical history and situation. FLCCC and WCH provide treatment protocols.
  or use drag and drop   
Submit