Cannabidiol for COVID-19: real-time meta-analysis of 12 studies

@CovidAnalysis, October 2026, Version 22V22 (V1 2022)
Abstract
Meta-analysis using the most serious outcome reported shows 23% [-22‑95%] higher risk, without reaching statistical significance.
All data and sources to reproduce this analysis are in the appendix.
Download PDF
Evolution of COVID-19 clinical evidence Meta-analysis results over time Cannabidiol p=0.39 Vitamin D p<0.0000000001 2020 2021 2022 2023 2024 2025 2026 Lowerrisk Higherrisk c19early.org October 2026 50% 0% -50%
Cannabidiol for COVID-19 — Highlights
Cannabidiol reduces risk with low confidence for ventilation, however increased risk is seen with very high confidence for ICU admission, low confidence for cases, and very low confidence for hospitalization and in pooled analysis.
Real-time updates and corrections with a consistent protocol for 226 treatments. Outcome specific analysis and combined evidence from all studies including treatment delay, a primary confounding factor.
c19early.org
October 2026

Cannabidiol COVID-19 studies

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Kuntzman (DB RCT)77%0.23 · 0.01–4.49death0/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.05death0/452/45
Late treatment32%0.68 · 0.07–6.473/1013/9032% lower risk
Tau² = 1.65, I² = 41.9%, p = 0.75
Nguyen50%0.50 · 0.31–0.82cases26/53148/531 Huang−181%2.81 · 1.04–7.58death13,099 (all patients) Merianos−212%3.12 · 1.87–4.97symp. case94/41620/384 Lehrer−24%1.24 · 1.05–1.45casesn/an/a Shover (PSM)2%0.98 · 0.94–1.04death3/69199/1,762 Mannumbet.. (PSM)56%0.44 · 0.35–0.56death Griffith3%0.97 · 0.82–1.14death Milad−101%2.01 · 1.37–2.94cases74/55436/541 Torrejón-Guirado−154%2.54 · 1.47–4.12cases22/15667/895
Prophylaxis−28%1.28 · 0.79–2.06219/1,726370/4,11328% higher risk
Tau² = 0.46, I² = 96.6%, p = 0.32
All studies−23%1.23 · 0.78–1.95222/1,827373/4,20323% higher risk
Tau² = 0.46, I² = 95.3%, p = 0.3900.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
B
Loading..
Fig. 1. A. Random-effects meta-analysis. This plot shows pooled effects, see the specific outcome analyses for individual outcomes. Analysis validating pooled outcomes for COVID-19 can be found below. Effect extraction is pre-specified, using the most serious outcome reported. For details see the appendix. B. Timeline of results in cannabidiol studies.
Introduction
Fig. 2. SARS-CoV-2 spike protein fibrin binding leads to thromboinflammation and neuropathology, from1.
Immediate treatment recommended
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 injury2-18 and cognitive deficits5,10, cardiovascular complications19-25, DNA damage26-29, organ failure, and death. Even mild untreated infections may result in persistent cognitive deficits30—the spike protein binds to fibrin leading to fibrinolysis-resistant blood clots, thromboinflammation, and neuropathology. Minimizing replication as early as possible is recommended.
Many treatments are expected to modulate infection
SARS-CoV-2 infection and replication involves the complex interplay of 500+ host and viral proteins and other factorsA,31-38, providing many therapeutic targets for which many existing compounds have known activity. Scientists have predicted that over 12,000 compounds may reduce COVID-19 risk39, either by directly minimizing infection or replication, by supporting immune system function, or by minimizing secondary complications.
Analysis
We analyze all significant controlled studies of cannabidiol 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, individual outcomes, peer-reviewed studies, and Randomized Controlled Trials (RCTs).
Treatment timing
Fig. 3 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.
regular treatment to prevent or minimize infectionstreat immediately on symptoms or shortly thereafterlate stage after disease progressionexposed to virusEarly TreatmentProphylaxisTreatment delayLate Treatment
Fig. 3. Treatment stages.
Preclinical Research
An in vitro study supports the efficacy of cannabidiol40.
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 for all stages combined, for Randomized Controlled Trials, for peer-reviewed studies, and for specific outcomes. Table 2 shows results by treatment stage. Fig. 4 plots individual results by treatment stage. Fig. 5, 6, 7, 8, 9, 10, 11, 12, and 13 show forest plots for random-effects meta-analysis of all studies with pooled effects, mortality results, ventilation, ICU admission, hospitalization, recovery, cases, peer reviewed studies, and non-symptomatic vs. symptomatic results.
Table 1. Random-effects meta-analysis for all stages combined, for Randomized Controlled Trials, for peer-reviewed studies, and for specific outcomes. Results show the relative risk with treatment and the 95% confidence interval. * p<0.05  **** p<0.0001.
Relative Risk Studies Patients
All studies1.23 [0.78‑1.95]1210K
Peer-reviewedPeer-reviewed1.46 [1.03‑2.07]*1010K
RCTsRCTs0.68 [0.07‑6.47]3191
Mortality0.82 [0.44‑1.52]610K
VentilationVent.0.75 [0.70‑0.81]****21,831
ICU admissionICU1.26 [1.13‑1.40]****31,841
HospitalizationHosp.1.31 [0.82‑2.10]4101
Cases1.39 [0.92‑2.11]610K
RCT mortality0.22 [0.03‑1.78]2100
Table 2. Random-effects meta-analysis results by treatment stage. Results show the relative risk with treatment and the 95% confidence interval.treatment and the 95% confidence interval. * p<0.05  **** p<0.0001.
Late treatment Prophylaxis
All studies0.68 [0.07‑6.47]0.68
[0.07‑6.47]		1.28 [0.79‑2.06]1.28
[0.79‑2.06]
Peer-reviewedPeer-reviewed1.24 [0.05‑33.04]1.24
[0.05‑33.04]		1.47 [1.03‑2.09]*1.47*
[1.03‑2.09]
RCTsRCTs0.68 [0.07‑6.47]0.68
[0.07‑6.47]
Mortality0.22 [0.03‑1.78]0.22
[0.03‑1.78]		0.92 [0.48‑1.75]0.92
[0.48‑1.75]
VentilationVent.0.75 [0.70‑0.81]****0.75****
[0.70‑0.81]
ICU admissionICU0.43 [0.04‑4.82]0.43
[0.04‑4.82]		1.26 [1.13‑1.40]****1.26****
[1.13‑1.40]
HospitalizationHosp.1.41 [0.22‑9.25]1.41
[0.22‑9.25]		1.38 [0.82‑2.32]1.38
[0.82‑2.32]
Cases1.39 [0.92‑2.11]1.39
[0.92‑2.11]
RCT mortality0.22 [0.03‑1.78]0.22
[0.03‑1.78]
Loading..
Fig. 4. Scatter plot showing the most serious outcome in all studies, and for studies within each stage. Diamonds shows the results of random-effects meta-analysis.
c19early.org
October 2026

Cannabidiol COVID-19 studies

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Kuntzman (DB RCT)77%0.23 · 0.01–4.49death0/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.05death0/452/45
Late treatment32%0.68 · 0.07–6.473/1013/9032% lower risk
Tau² = 1.65, I² = 41.9%, p = 0.75
Nguyen50%0.50 · 0.31–0.82cases26/53148/531 Huang−181%2.81 · 1.04–7.58death13,099 (all patients) Merianos−212%3.12 · 1.87–4.97symp. case94/41620/384 Lehrer−24%1.24 · 1.05–1.45casesn/an/a Shover (PSM)2%0.98 · 0.94–1.04death3/69199/1,762 Mannumbet.. (PSM)56%0.44 · 0.35–0.56death Griffith3%0.97 · 0.82–1.14death Milad−101%2.01 · 1.37–2.94cases74/55436/541 Torrejón-Guirado−154%2.54 · 1.47–4.12cases22/15667/895
Prophylaxis−28%1.28 · 0.79–2.06219/1,726370/4,11328% higher risk
Tau² = 0.46, I² = 96.6%, p = 0.32
All studies−23%1.23 · 0.78–1.95222/1,827373/4,20323% higher risk
Tau² = 0.46, I² = 95.3%, p = 0.3900.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 5. Random-effects meta-analysis for all studies. This plot shows pooled effects, see the specific outcome analyses for individual outcomes. Analysis validating pooled outcomes for COVID-19 can be found below. Effect extraction is pre-specified, using the most serious outcome reported. For details see the appendix.
c19early.org
October 2026

Cannabidiol COVID-19 mortality results

StudyImprovementRR · 95% CITreatmentControlRelative Risk
Kuntzman (DB RCT)77%0.23 · 0.01–4.490/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.050/452/45
Late treatment78%0.22 · 0.03–1.780/523/4878% lower risk
Tau² = 0.00, I² = 0.0%, p = 0.15
Huang−181%2.81 · 1.04–7.5813,099 (all patients) Shover (PSM)2%0.98 · 0.94–1.043/69199/1,762 Mannumbet.. (PSM)56%0.44 · 0.35–0.56 Griffith3%0.97 · 0.82–1.14
Prophylaxis8%0.92 · 0.48–1.753/69199/1,7628% lower risk
Tau² = 0.33, I² = 95.7%, p = 0.8
All studies18%0.82 · 0.44–1.523/121202/1,81018% lower risk
Tau² = 0.32, I² = 93.0%, p = 0.5400.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 6. Random-effects meta-analysis for mortality results.
c19early.org
October 2026

Cannabidiol COVID-19 mechanical ventilation results

StudyImprovementRR · 95% CITreatmentControlRelative Risk
Shover (PSM)5%0.95 · 0.91–0.993/69292/1,762 Mannumbet.. (PSM)25%0.75 · 0.70–0.81
Prophylaxis25%0.75 · 0.70–0.813/69292/1,76225% lower risk
Tau² = 0.00, I² = 0.0%, p < 0.0001
All studies25%0.75 · 0.70–0.813/69292/1,76225% lower risk
Tau² = 0.00, I² = 0.0%, p < 0.000100.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 7. Random-effects meta-analysis for ventilation.
c19early.org
October 2026

Cannabidiol COVID-19 ICU results

StudyImprovementRR · 95% CITreatmentControlRelative Risk
Kuntzman (DB RCT)57%0.43 · 0.04–4.821/71/3
Late treatment57%0.43 · 0.04–4.821/71/357% lower risk
Tau² = 0.00, I² = 0.0%, p = 0.5
Shover (PSM)9%0.91 · 0.85–0.998/69543/1,762 Griffith−27%1.27 · 1.14–1.41
Prophylaxis−26%1.26 · 1.13–1.408/69543/1,76226% higher risk
Tau² = 0.00, I² = 0.0%, p < 0.0001
All studies−26%1.26 · 1.13–1.409/76544/1,76526% higher risk
Tau² = 0.00, I² = 0.0%, p < 0.000100.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 8. Random-effects meta-analysis for ICU admission.
c19early.org
October 2026

Cannabidiol COVID-19 hospitalization results

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Kuntzman (DB RCT)22%0.78 · 0.30–2.01hosp. time7 (n)3 (n)
Late treatment−41%1.41 · 0.22–9.253/560/4541% higher risk
Tau² = 1.04, I² = 45.7%, p = 0.73
Mannumbet.. (PSM)−6%1.06 · 1.02–1.10hosp. time Griffith−80%1.80 · 1.68–1.93hosp.
Prophylaxis−38%1.38 · 0.82–2.3238% higher risk
Tau² = 0.14, I² = 99.4%, p = 0.23
All studies−31%1.31 · 0.82–2.103/560/4531% higher risk
Tau² = 0.14, I² = 98.4%, p = 0.2600.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 9. Random-effects meta-analysis for hospitalization.
c19early.org
October 2026

Cannabidiol COVID-19 recovery results

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−33%1.33 · 0.86–2.08recov. time49 (n)42 (n)
Late treatment−33%1.33 · 0.86–2.0849 (n)42 (n)33% higher risk
Tau² = 0.00, I² = 0.0%, p = 0.2
All studies−33%1.33 · 0.86–2.0849 (n)42 (n)33% higher risk
Tau² = 0.00, I² = 0.0%, p = 0.200.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 10. Random-effects meta-analysis for recovery.
c19early.org
October 2026

Cannabidiol COVID-19 case results

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Nguyen50%0.50 · 0.31–0.82cases26/53148/531 Huang19%0.81 · 0.73–0.90cases13,099 (all patients) Merianos−212%3.12 · 1.87–4.97symp. case94/41620/384 Lehrer−24%1.24 · 1.05–1.45casesn/an/a Milad−101%2.01 · 1.37–2.94cases74/55436/541 Torrejón-Guirado−154%2.54 · 1.47–4.12cases22/15667/895
Prophylaxis−39%1.39 · 0.92–2.11216/1,657171/2,35139% higher risk
Tau² = 0.24, I² = 94.0%, p = 0.12
All studies−39%1.39 · 0.92–2.11216/1,657171/2,35139% higher risk
Tau² = 0.24, I² = 94.0%, p = 0.1200.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 11. Random-effects meta-analysis for cases.
c19early.org
October 2026

Cannabidiol COVID-19 peer reviewed studies

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Kuntzman (DB RCT)77%0.23 · 0.01–4.49death0/71/3
Late treatment−24%1.24 · 0.05–33.043/561/4524% higher risk
Tau² = 3.34, I² = 59.6%, p = 0.91
Nguyen50%0.50 · 0.31–0.82cases26/53148/531 Huang−181%2.81 · 1.04–7.58death13,099 (all patients) Merianos−212%3.12 · 1.87–4.97symp. case94/41620/384 Lehrer−24%1.24 · 1.05–1.45casesn/an/a Shover (PSM)2%0.98 · 0.94–1.04death3/69199/1,762 Griffith3%0.97 · 0.82–1.14death Milad−101%2.01 · 1.37–2.94cases74/55436/541 Torrejón-Guirado−154%2.54 · 1.47–4.12cases22/15667/895
Prophylaxis−47%1.47 · 1.03–2.09219/1,726370/4,11347% higher risk
Tau² = 0.19, I² = 87.7%, p = 0.032
All studies−46%1.46 · 1.03–2.07222/1,782371/4,15846% higher risk
Tau² = 0.19, I² = 84.9%, p = 0.03200.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 12. Random-effects meta-analysis for peer reviewed studies. 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. Analysis validating pooled outcomes for COVID-19 can be found below.
c19early.org
October 2026

Cannabidiol COVID-19 symptomatic vs. case outcomes

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Huang−181%2.81 · 1.04–7.58death13,099 (all patients) Merianos−212%3.12 · 1.87–4.97symp. case94/41620/384 Shover (PSM)2%0.98 · 0.94–1.04death3/69199/1,762 Mannumbet.. (PSM)56%0.44 · 0.35–0.56death Griffith3%0.97 · 0.82–1.14death Kuntzman (DB RCT)77%0.23 · 0.01–4.49death0/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.05death0/452/45
Symptomatic−15%1.15 · 0.61–2.18100/586222/2,23615% higher risk
Tau² = 0.50, I² = 94.3%, p = 0.68
Nguyen50%0.50 · 0.31–0.82cases26/53148/531 Huang19%0.81 · 0.73–0.90cases13,099 (all patients) Merianos−3%1.03 · 0.96–1.09cases367/416317/384 Lehrer−24%1.24 · 1.05–1.45casesn/an/a Milad−101%2.01 · 1.37–2.94cases74/55436/541 Torrejón-Guirado−154%2.54 · 1.47–4.12cases22/15667/895
Cases−14%1.14 · 0.90–1.45489/1,657468/2,35114% higher risk
Tau² = 0.07, I² = 91.8%, p = 0.27
All studies−14%1.14 · 0.86–1.50589/2,243690/4,58714% higher risk
Tau² = 0.18, I² = 95.3%, p = 0.3600.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 13. Random-effects meta-analysis for non-symptomatic vs. symptomatic results. Effect extraction is pre-specified, using the most serious outcome reported, see the appendix for details. Analysis validating pooled outcomes for COVID-19 can be found below.
Randomized Controlled Trials (RCTs)
Fig. 14 shows a comparison of results for RCTs and observational studies. Fig. 15 and 16 show forest plots for random-effects meta-analysis of all Randomized Controlled Trials and RCT mortality results. RCT results are included in Table 1 and Table 2.
Loading..
Fig. 14. Results for RCTs and observational studies.
RCTs have many potential biases
RCTs help to make study groups more similar and can provide a higher level of evidence, however they are subject to many biases43, and analysis of double-blind RCTs has identified extreme levels of bias44. 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.
Conflicts of interest for COVID-19 RCTs
RCTs are expensive and many RCTs are funded by pharmaceutical companies or other organizations with conflicts of interest, for example governments that previously denied treatment with the study drug. 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. Bekelman et al. and Lundh et al. show that industry-sponsored studies are more likely to be favorable. For COVID-19, some major philanthropic organizations are largely funded by investments with extreme conflicts of interest for and against specific COVID-19 interventions.
RCTs for novel acute diseases requiring rapid treatment
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 226 treatments we have analyzed, 67% 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 may be less reliable
Concato et al. report a paradoxical finding—RCT results had higher variability, and only RCTs were found to sometimes report significant results the opposite of the overall result. The same trend is seen for the most popular (most politicized) COVID-19 treatments—considering all statistically significant results reported in studies, RCTs are slightly more likely to report a result in the opposite direction. In other words, for these COVID-19 treatments and for the topics covered by Concato et al., assuming causality from a single study is more likely to result in an incorrect conclusion for RCTs.
Increased risk of inconsistent results for RCTs suggests higher prevalence of bias, which may arise due to many issues including design bias, conflicts of interest, treatment differences by physicians aware of allocation, attrition bias, ascertainment bias, randomization failures, errors, or fraud.
Using all studies identifies efficacy 8+ months faster (9+ months for low-cost treatments)
Currently, 59 of the treatments we analyze show statistically significant efficacy or harm, defined as ≥10% decreased risk or >0% increased risk from ≥3 studies. Of these, 54% have been confirmed in RCTs, with a mean delay of 7.8 months (62% with 8.7 months delay for low-cost treatments). The remaining treatments either have no RCTs, or the point estimate is consistent.
All studies must be carefully analyzed
Neither observational studies nor RCTs prove causation—any study can be flawed or fraudulent. We need much more, for example a combination of results from many independent teams, detailed understanding of each study, knowledge of conflicts/team reliability, dose-response relationships, delay-response relationships, logical results across outcomes, or details consistent with preclinical expectations.
All studies must be evaluated individually. 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.
c19early.org
October 2026

Cannabidiol COVID-19 Randomized Controlled Trials

StudyImprovementRR · 95% CIOutcomeTreatmentControlRelative Risk
Crippa (DB RCT)−557%6.57 · 0.35–124hosp.3/490/42 Kuntzman (DB RCT)77%0.23 · 0.01–4.49death0/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.05death0/452/45
Late treatment32%0.68 · 0.07–6.473/1013/9032% lower risk
Tau² = 1.65, I² = 41.9%, p = 0.75
All studies32%0.68 · 0.07–6.473/1013/9032% lower risk
Tau² = 1.65, I² = 41.9%, p = 0.7500.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 15. Random-effects meta-analysis for all Randomized Controlled Trials. This plot shows pooled effects, see the specific outcome analyses for individual outcomes. Analysis validating pooled outcomes for COVID-19 can be found below. Effect extraction is pre-specified, using the most serious outcome reported. For details see the appendix.
c19early.org
October 2026

Cannabidiol COVID-19 RCT mortality results

StudyImprovementRR · 95% CITreatmentControlRelative Risk
Kuntzman (DB RCT)77%0.23 · 0.01–4.490/71/3 McNamara (DB RCT)80%0.20 · 0.01–4.050/452/45
Late treatment78%0.22 · 0.03–1.780/523/4878% lower risk
Tau² = 0.00, I² = 0.0%, p = 0.15
All studies78%0.22 · 0.03–1.780/523/4878% lower risk
Tau² = 0.00, I² = 0.0%, p = 0.1500.511.52+
Rotate screen for more detailsIncrease width for more details
← Cannabidiol
reduces riskCannabidiol
increases risk →
Fig. 16. Random-effects meta-analysis for RCT mortality results.
Heterogeneity
Heterogeneity in COVID-19 studies arises from many factors including:
Treatment delay
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 hours49,50. Baloxavir marboxil 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 marboxil for influenza show that early treatment is more effective.
Treatment delayResult
Post-exposure prophylaxis86% fewer cases51
<24 hours-33 hours symptoms52
24-48 hours-13 hours symptoms52
Inpatients-2.5 hours to improvement53
Fig. 17 shows a mixed-effects meta-regression for efficacy as a function of treatment delay in COVID-19 studies from 226 treatments, showing that efficacy declines rapidly with treatment delay. Early treatment is critical for COVID-19.
Loading..
Fig. 17. Early treatment is more effective. Meta-regression showing efficacy as a function of treatment delay in COVID-19 studies from 226 treatments.
Patient demographics
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, for example as in López-Medina et al.
SARS-CoV-2 variants
Efficacy may depend critically on the distribution of SARS-CoV-2 variants encountered by patients. Risk varies significantly across variants55, for example the Gamma variant shows significantly different characteristics56-59. 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 variants60,61.
Treatment regimen
Effectiveness may depend strongly on the dosage and treatment regimen.
Medication quality
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.
Other treatments
The use of other treatments may significantly affect outcomes, including supplements, other medications, or other interventions such as prone positioning. Treatments may be synergistic64-88, therefore efficacy may depend strongly on combined treatments.
Effect measured
Across all studies there is a strong association between different outcomes, for example improved recovery is strongly associated with lower mortality. However, efficacy may differ depending on the effect measured, for example a treatment may be more effective against secondary complications and have minimal effect on viral clearance.
Meta-analysis
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. 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.
Pooled Effects
Combining studies is required
For COVID-19, delay in clinical results translates into additional death and morbidity, as well as additional economic and societal damage. Combining the results of studies reporting different outcomes is required. There may be no mortality in a trial with low-risk patients, however a reduction in severity or improved viral clearance may translate into lower mortality in a high-risk population. Different studies may report lower severity, improved recovery, and lower mortality, and the significance may be very high when combining the results. "The studies reported different outcomes" is not a good reason for disregarding results. Pooling the results of studies reporting different outcomes allows us to use more of the available information. Logically we should, and do, use additional information when evaluating treatments—for example dose-response and treatment delay-response relationships provide additional evidence of efficacy that is considered when reviewing the evidence for a treatment.
Specific outcome and pooled analyses
We present both specific outcome and pooled analyses. In order to combine the results of studies reporting different outcomes we use the most serious outcome reported in each study, based on the thesis that improvement in the most serious outcome provides comparable measures of efficacy for a treatment. A critical advantage of this approach is simplicity and transparency. There are many other ways to combine evidence for different outcomes, along with additional evidence such as dose-response relationships, however these increase complexity.
Ethical and practical issues limit high-risk trials
Trials with high-risk patients may be restricted due to ethics for treatments that are known or expected to be effective, and they increase difficulty for recruiting. Using less severe outcomes as a proxy for more serious outcomes allows faster and safer collection of evidence.
Validating pooled outcome analysis for COVID-19
For many COVID-19 treatments, a reduction in mortality logically follows from a reduction in hospitalization, which follows from a reduction in symptomatic cases, which follows from a reduction in PCR positivity. We can directly test this for COVID-19.
Analysis of the the association between different outcomes across studies from all 226 treatments we cover confirms the validity of pooled outcome analysis for COVID-19. Fig. 18 shows that lower hospitalization is very strongly associated with lower mortality (p < 0.0000000001). Similarly, Fig. 19 shows that improved recovery is very strongly associated with lower mortality (p < 0.0000000001). Considering the extremes, Singh et al. show an association between viral clearance and hospitalization or death, with p = 0.003 after excluding one large outlier from a mutagenic treatment, and based on 44 RCTs including 52,384 patients. Fig. 20 shows that improved viral clearance is strongly associated with fewer serious outcomes. The association is very similar to Singh et al., with higher confidence due to the larger number of studies. As with Singh et al., the confidence increases when excluding the outlier treatment, from p = 0.0000000074 to p = 0.00000000015.
Loading..
Fig. 18. Lower hospitalization is associated with lower mortality, supporting pooled outcome analysis.
Loading..
Fig. 19. Improved recovery is associated with lower mortality, supporting pooled outcome analysis.
Loading..
Fig. 18. Improved viral clearance is associated with fewer serious outcomes, supporting pooled outcome analysis.
Pooled outcomes identify efficacy 5 months faster (7 months for RCTs)
Currently, 59 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 these have been confirmed with one or more specific outcomes, with a mean delay of 4.6 months. When restricting to RCTs only, 53% of treatments showing statistically significant efficacy/harm with pooled effects have been confirmed with one or more specific outcomes, with a mean delay of 7.5 months. Fig. 21 shows when treatments were found effective during the pandemic. Pooled outcomes often resulted in earlier detection of efficacy.
Loading..
Loading..
Fig. 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.
Limitations
Pooled analysis could hide efficacy, for example a treatment that is beneficial for late stage patients but has no effect on viral clearance may show no efficacy if most studies only examine viral clearance. In practice, it is rare for a non-antiviral treatment to report viral clearance and to not report clinical outcomes; and in practice other sources of heterogeneity such as differences in treatment delay are more likely to hide efficacy.
Summary
Analysis validates the use of pooled effects and shows significantly faster detection of efficacy on average. However, as with all meta-analyses, it is important to review the different studies included. We also present individual outcome analyses, which may be more informative for specific use cases.
Discussion
Publication bias
Publishing is often biased towards positive results. Trials with patented drugs may have a financial conflict of interest that results in positive studies being more likely to be published, or bias towards more positive results. For example with molnupiravir, trials with negative results remain unpublished to date (CTRI/2021/05/033864 and CTRI/2021/08/0354242). For cannabidiol, there is currently not enough data to evaluate publication bias with high confidence.
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. Fig. 22 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 < 0.0590-97. 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.
Log Risk Ratio Standard Error 1.406 1.055 0.703 0.352 0 -3 -2 -1 0 1 2 A: Simulated perfect trials p > 0.05 Log Risk Ratio Standard Error 1.433 1.074 0.716 0.358 0 -4 -3 -2 -1 0 1 2 B: Simulated perfect trials with varying treatment delay p < 0.0001
Fig. 22. Example funnel plot analysis for simulated perfect trials.
Limitations
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 for 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 with conflicts of interest 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 alone64-88. 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 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.
Reviews
Multiple reviews cover cannabidiol for COVID-19, presenting additional background on mechanisms and related results, including98,99.
Perspective
Results compared with other treatments
SARS-CoV-2 infection and replication involves a complex interplay of 500+ host and viral proteins and other factors31-38, providing many therapeutic targets. Over 12,000 compounds have been predicted to reduce COVID-19 risk39, either by directly minimizing infection or replication, by supporting immune system function, or by minimizing secondary complications. Fig. 23 shows an overview of the results for cannabidiol in the context of multiple COVID-19 treatments, and Fig. 24 shows a plot of efficacy vs. cost for COVID-19 treatments.
Loading..
Fig. 23. Scatter plot showing results within the context of multiple COVID-19 treatments. Diamonds shows the results of random-effects meta-analysis. 0.5% of 12,000+ proposed treatments show efficacy100.
Loading..
Loading..
Fig. 24. Efficacy vs. cost for COVID-19 treatments.
Conclusion
Meta-analysis using the most serious outcome reported shows 23% [-22‑95%] higher risk, without reaching statistical significance.
 
Contact.
Contact us on X at @CovidAnalysis.
Funding.
We have received no funding or compensation in any form, and do not accept donations. This is entirely volunteer work.
Conflicts of interest.
We have no conflicts of interest. We have no affiliation with any pharmaceutical companies, supplement companies, governments, political parties, or advocacy organizations.
AI.
We use AI models (Gemini, Grok, Claude, and ChatGPT) tasked with functioning as additional peer-reviewers to check for errors, suggest improvements, and review spelling and grammar. Any corrections are manually verified. Our preference for em dashes is independent of AI.
Updates.
Our COVID-19 meta-analyses involve the extraction of over 228,000 datapoints from thousands of papers for 226 treatments. We thank the thousands of scientists, physicians, and other contributors that have provided updates, suggestions, feedback, and corrections. These are all welcome and can be submitted at https://c19early.org/cbdmeta.html.
Dedication.
This work is dedicated to top evidence-based physicians that worked tirelessly to analyze evidence and greatly reduce mortality and morbidity during the pandemic. In alphabetical order: Dr. Thomas J. Borody, Dr. Mary Talley Bowden, Dr. Flavio Cadegiani, Dr. Shankara Chetty, Dr. Ryan Cole, Dr. George Fareed, Dr. Sabine Hazan, Dr. Pierre Kory, Dr. Tess Lawrie, Dr. Robert Malone, Dr. Paul Marik, Dr. Peter McCullough, Dr. Didier Raoult, Dr. Harvey Risch, Dr. Jackie Stone, Dr. Brian Tyson, Dr. Joseph Varon, and Dr. Vladimir Zelenko.
Public domain.
This is a public domain work distributed in accordance with the Creative Commons CC0 1.0 Universal license, which dedicates the work to the public domain by waiving all rights worldwide under copyright law. You can distribute, remix, adapt, and build upon this work in any medium or format, including for commercial purposes, without asking permission. Referenced material and third-party images retain any original copyrights or restrictions. See: https://creativecommons.org/publicdomain/zero/1.0/.
Study Notes
Late treatmentRCT · NCT04467918 · 91 patients · Brazil · Jul – Oct 2020
Cannabidiol for COVID-19
Crippa et al. Cannabis and Cannabinoid Research · Oct 2021
Outcome
Improvement
Relative Risk · 95% CI
Hospitalization
-557%6.57 · 0.35 to >9
Recovery time
-33%1.33 · 0.86–2.08
00.511.52+
← Cannabidiol
reduces riskCannabidiol
increases risk →
Higher hospitalization and slower recovery
Not statistically significant · p = 0.25 (hospitalization) and p = 0.21 (slower recovery)
RCT 105 patients recruited in an ER in Brazil, 49 treated with CBD, showing no significant differences with treatment. 300mg CBD for 14 days.

For discussion see101. Submit Corrections or Updates.
ProphylaxisRetrospective · 72,501 patients · USA · Feb 2020 – Jan 2022
Cannabidiol for COVID-19
Griffith et al. JAMA Network Open · June 2024
Outcome
Improvement
Relative Risk · 95% CI
Mortality
3%0.97 · 0.82–1.14
ICU admission
-27%1.27 · 1.14–1.41
Hospitalization
-80%1.80 · 1.68–1.93
00.511.52+
← Cannabidiol
reduces riskCannabidiol
increases risk →
Higher ICU admission and hospitalization
p < 0.0001 (ICU admission) and p < 0.0001 (hospitalization)
Retrospective 72,501 COVID-19 patients in the USA showing cannabis use associated with higher risk of hospitalization and ICU admission. Submit Corrections or Updates.
UK Biobank retrospective with 13,099 cannabis users, showing a lower risk of COVID-19 infection, however regular users had a significantly higher risk of mortality. Submit Corrections or Updates.
Late treatmentDouble-blind RCT · 10 patients · Israel
Cannabidiol for COVID-19
Kuntzman et al. Israel Medical Association J. · February 2025
Outcome
Improvement
Relative Risk · 95% CI
Mortality
77%0.23 · 0.01–4.49
ICU admission
57%0.43 · 0.04–4.82
Hospitalization time
22%0.78 · 0.30–2.01
00.511.52+
← Cannabidiol
reduces riskCannabidiol
increases risk →
Lower mortality with cannabidiol
Not statistically significant · p = 0.3
Is late treatment with cannabidiol beneficial for COVID-19?
c19early.org
RCT 10 hospitalized COVID-19 patients showing potential benefit with sublingual cannabidiol (CBD) oil. There was only 7 patients in the treatment arm and 3 in the placebo group. CBD-treated patients showed a trends toward lower inflammatory cytokine levels including IL-6, IL-10, TNF-α, and C-reactive protein, though none reached statistical significance. Submit Corrections or Updates.
UK Biobank retrospective showing a higher risk of COVID-19 cases with a history of cannabis use. Submit Corrections or Updates.
ProphylaxisPSM retrospective · 39,285 patients · USA · Jan – Dec 2020
Cannabidiol for COVID-19
Mannumbeth Renjithlal et al. SSRN Electronic J. · January 2023
Outcome
Improvement
Relative Risk · 95% CI
Mortality
56%0.44 · 0.35–0.56
Ventilation
25%0.75 · 0.70–0.81
Hospitalization time
-6%1.06 · 1.02–1.10
00.511.52+
← Cannabidiol
reduces riskCannabidiol
increases risk →
Lower mortality and ventilation
p < 0.0001 (mortality) and p < 0.0001 (ventilation)
Retrospective 1,657,800 COVID-19 hospitalizations in the USA including 13,095 patients with cannabis use disorder, showing lower risk of mortality with cannabis use disorder. The text and Table S2 have conflicting results for mortality: 0.45 [0.36-0.57] versus 0.43 [0.34-0.55]. Submit Corrections or Updates.
Early terminated RCT showing lower mortality with cannabidiol treatment, without statistical significance. Submit Corrections or Updates.
Retrospective 800 e-cigarette users in the USA, showing higher risk of COVID-19 diagnosis and symptoms with cannabis use. Submit Corrections or Updates.
Longitudinal observational cohort study of 1,343 community adults in Canada, showing increased risk of self-reported COVID-19 infection with dried cannabis use. Cannabis users were younger than the control group. Submit Corrections or Updates.
Retrospective 1,212 patients in the USA with a history of seizure-related conditions, showing patients treated with CBD100 had significantly lower incidence of COVID-19 cases compared to a matched control group.

In vitro study showing CBD inhibits SARS-CoV-2 with Vero E6 and Calu-3 cells. Mouse study showing CBD significantly inhibited viral replication in the lung and nasal turbinate.

Authors note that CBD does not inhibit ACE2 expression or the main viral proteases, inhibition occurs after viral entry. Authors stress several limitations for use at this time, including purity, quality, and the formulation of products, and potential lung damage based on administration method.

Authors recommend clinical trials, but do not mention the existing RCT by Crippa et al. Submit Corrections or Updates.
ProphylaxisRetrospective · 1,831 patients · USA · Feb 2020 – Feb 2021
Cannabidiol for COVID-19
Shover et al. J. Cannabis Research · August 2022
Outcome
Improvement
Relative Risk · 95% CI
Mortality
2%0.98 · 0.94–1.04
Ventilation
5%0.95 · 0.91–0.99
ICU admission
9%0.91 · 0.85–0.99
Oxygen therapy
3%0.97 · 0.92–1.02
00.511.52+
← Cannabidiol
reduces riskCannabidiol
increases risk →
Lower ventilation and ICU admission
p = 0.017 (ventilation) and p = 0.018 (ICU admission)
Retrospective 1,831 hospitalized COVID-19 patients in the USA, showing lower mechanical ventilation and ICU admission, but no significant difference in mortality. Submit Corrections or Updates.
Analysis of 1,051 adolescents aged 14-18 in Spain, showing that those testing positive for COVID-19 were approximately 2.9 times more likely to have used cannabis in the last month.

Limitations include the small number of COVID-19-positive participants (n=89), reliance on self-reported data, and the Nagelkerke R² indicating ~60% of variability was unexplained by the model. Submit Corrections or Updates.
Appendix 1. Methods and Data
Search methods
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 c19early.org, which regularly receives notification of studies upon publication. Search terms are cannabidiol 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 cannabidiol for COVID-19 that report a comparison with a control group are included in the main analysis. Studies with major unexplained data issues, for example major outcome data that is impossible to be correct with no response from the authors, are excluded.
Fig. 25. Mid-recovery results can more accurately reflect efficacy when almost all patients recover. Mateja et al. confirm that intermediate viral load results more accurately reflect hospitalization/death.
Effect extraction
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 reported then they are both used in specific outcome analyses, while mortality is used for pooled analysis. If symptomatic results are reported at multiple times, we use 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 outcomes. 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. An IPD meta-analysis confirms that intermediate viral load reduction is more closely associated with hospitalization/death than later viral load reduction102. 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.
Statistical methods
Forest plots are computed using PythonMeta103 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. When results provide an odds ratio, we compute the relative risk when possible, or convert to a relative risk according to Zhang et al. Reported confidence intervals and p-values are used when available, and adjusted values are used 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 1107. 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 (3.14.7) with scipy (1.18.1), pythonmeta (1.26), numpy (2.5.3), statsmodels (0.15.0), and plotly (6.9.0). Mixed-effects meta-regression results are computed with R (4.4.0) using the metafor (4.6-0) and rms (6.8-0) packages, and using the most serious sufficiently powered outcome. For all statistical tests, a p-value less than 0.05 was considered statistically significant. Grobid 0.8.2 is used to parse PDF documents.
When evaluating potential effect modification across groups, we use an interaction test as described by Altman (C) et al. We compared the log-transformed relative risks using a z-test, deriving the standard error of the difference from the 95% confidence intervals. A two-sided interaction p-value of < 0.05 was considered a statistically significant difference in treatment effect between the groups.
Quality evaluation
Cochrane RoB 2/ROBINS-I are often used to evaluate studies, and have the advantage of providing standardized rules that can be applied with minimal understanding of the domain and study. However, the rules do not account for many real-world issues, often overemphasize or underemphasize others, and studies show low inter-rater reliability115. Certain domains are more applicable for these tools, however the time-sensitive nature of a pandemic, with significant mortality for every day of delay in evidence assessment, and the characteristics of COVID-19 make them inappropriate for this domain. This can be demonstrated with examples where expert RoB 2/ROBINS-I ratings do not match reality for COVID-19. Popp et al. use RoB 2 to classify Reis et al. as low risk of bias, however this is the opposite of reality—the trial not only has very high risk of bias, but has very high actual known bias, refusing to release data despite pledging to, reporting multiple impossible numbers, having blinding and randomization failure, and many other issues117. Axfors et al. use RoB 2 to classify Horby et al. as low risk of bias, however this is the opposite of reality—the very late treatment and excessive dosage used produces results with no relevance to recommended usage. HCQ shows poor results with late treatment and excessive dosage, and the combination shows harmB. Hempenius et al. use ROBINS-I to classify 33 studies for HCQ. The two rated as having the lowest risk of bias113,114 are far from the most informative. Both involve very late treatment, providing no information on recommended usage, and ROBINS-I does a very poor job of accounting for the impact of confounding factorsC.
Our quality evaluation focuses on known issues and bias, and the potential impact on outcomes, rather than just the risk of bias. The estimated potential impact of each confounding factor, and the direction of the impact is considered. For example, consider a study that shows significantly lower risk, the value of the study varies significantly if confounding points to an underestimate or an overestimate of efficacy. In one case, the real effect may be null, while the other case provides stronger evidence of efficacy (which may be greater than the study shows). Analysis focusing on the risk of bias, while simpler, may penalize studies for theoretical or technical issues that have no or minimal impact on outcomes. Analysis also depends on the outcome, for example certain issues are less relevant for objective outcomes such as mortality. Inaccurate penalization, and inaccurate high-quality evaluation in the face of known major issues affecting outcomes, increases in significance during a pandemic when immediate recognition of new evidence is critical, and when considering all global studies, as required during a pandemic. Investigators in other countries may have different customs for design, analysis, and reporting, and different English language skills, however they may not be less diligent or have greater bias. Investigators in lower-pharmaceutical-profit countries may have lower bias towards profitable interventions.
Treatment time
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 effective49,50.
Living analysis
This is a living analysis and is updated regularly. Submit updates or corrections with the form below. We received no funding, this research is done in our spare time. We have no affiliation with any pharmaceutical companies, supplement companies, governments, political parties, or advocacy organizations.
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://c19early.org/cbdmeta.html.
Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. For pooled analyses, the first (most serious) outcome is used, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.
Crippa, 10/7/2021, Double Blind Randomized Controlled Trial, placebo-controlled, Brazil, peer-reviewed, 32 authors, study period 7 July, 2020 - 16 October, 2020, trial NCT04467918 (history).
Submit Corrections or Updates.
 risk of hospitalization, 557.1% higher, RR 6.57, p = 0.25, treatment 3 of 49 (6.1%), control 0 of 42 (0.0%), continuity correction due to zero event (with reciprocal of the contrasting arm).
recovery time, 33.3% higher, relative time 1.33, p = 0.20, treatment 49, control 42.
Kuntzman, 2/28/2025, Double Blind Randomized Controlled Trial, Israel, peer-reviewed, 3 authors.
Submit Corrections or Updates.
 risk of death, 76.9% lower, RR 0.23, p = 0.30, treatment 0 of 7 (0.0%), control 1 of 3 (33.3%), NNT 3.0, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
risk of ICU admission, 57.1% lower, RR 0.43, p = 1.00, treatment 1 of 7 (14.3%), control 1 of 3 (33.3%), NNT 5.2.
hospitalization time, 22.4% lower, relative time 0.78, p = 0.61, treatment mean 5.9 (±5.13) n=7, control mean 7.6 (±3.0) n=3.
McNamara, 7/25/2025, Double Blind Randomized Controlled Trial, placebo-controlled, USA, preprint, 1 author, trial NCT04615949 (history).
Submit Corrections or Updates.
 risk of death, 80.0% lower, RR 0.20, p = 0.49, treatment 0 of 45 (0.0%), control 2 of 45 (4.4%), NNT 22, relative risk is not 0 because of continuity correction due to zero events (with reciprocal of the contrasting arm).
Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. For pooled analyses, the first (most serious) outcome is used, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.
Griffith, 6/21/2024, retrospective, USA, peer-reviewed, mean age 48.9, 12 authors, study period 1 February, 2020 - 31 January, 2022.
Submit Corrections or Updates.
 risk of death, 3.0% lower, OR 0.97, p = 0.73, treatment 7,060, control 65,441, adjusted per study, multivariable, RR approximated with OR.
risk of ICU admission, 27.0% higher, OR 1.27, p < 0.001, treatment 7,060, control 65,441, adjusted per study, multivariable, RR approximated with OR.
risk of hospitalization, 80.0% higher, OR 1.80, p < 0.001, treatment 7,060, control 65,441, adjusted per study, multivariable, RR approximated with OR.
Huang, 3/8/2022, retrospective, United Kingdom, peer-reviewed, 3 authors.
Submit Corrections or Updates.
 risk of death, 181.0% higher, HR 2.81, p = 0.04, regular users, Cox proportional hazards.
risk of case, 19.0% lower, OR 0.81, p < 0.001, adjusted per study, multivariable, RR approximated with OR.
Lehrer, 6/22/2022, retrospective, United Kingdom, peer-reviewed, mean age 57.0, 3 authors, study period 16 March, 2020 - 26 April, 2020.
Submit Corrections or Updates.
 risk of case, 23.8% higher, OR 1.24, p = 0.009, RR approximated with OR.
Mannumbeth Renjithlal, 1/28/2023, retrospective, propensity score matching, USA, preprint, 7 authors, study period 1 January, 2020 - 31 December, 2020.
Submit Corrections or Updates.
 risk of death, 55.6% lower, RR 0.44, p < 0.001, treatment 380 of 13,095 (2.9%), control 1,430 of 26,190 (5.5%), NNT 39, odds ratio converted to relative risk.
risk of mechanical ventilation, 24.6% lower, RR 0.75, p < 0.001, treatment 925 of 13,095 (7.1%), control 2,455 of 26,190 (9.4%), NNT 43.
hospitalization time, 5.9% higher, relative time 1.06, p < 0.001, treatment mean 7.2 (±12.78) n=13,095, control mean 6.8 (±10.42) n=26,190.
Merianos, 3/31/2022, retrospective, USA, peer-reviewed, survey, 6 authors.
Submit Corrections or Updates.
 risk of symptomatic case, 211.9% higher, RR 3.12, p < 0.001, treatment 94 of 416 (22.6%), control 20 of 384 (5.2%), odds ratio converted to relative risk, COVID-19 symptoms.
risk of symptomatic case, 70.6% higher, RR 1.71, p = 0.008, treatment 77 of 416 (18.5%), control 38 of 384 (9.9%), odds ratio converted to relative risk, COVID-19 diagnosis.
risk of case, 3.4% higher, RR 1.03, p = 0.33, treatment 367 of 416 (88.2%), control 317 of 384 (82.6%), odds ratio converted to relative risk, COVID-19 test.
Milad, 7/26/2024, prospective, Canada, peer-reviewed, 4 authors.
Submit Corrections or Updates.
 risk of case, 100.7% higher, RR 2.01, p < 0.001, treatment 74 of 554 (13.4%), control 36 of 541 (6.7%), 2+ infections.
risk of case, 19.0% higher, RR 1.19, p < 0.001, treatment 374 of 554 (67.5%), control 307 of 541 (56.7%), 1+ infections.
Nguyen, 1/20/2022, retrospective, USA, peer-reviewed, 34 authors.
Submit Corrections or Updates.
 risk of case, 49.6% lower, RR 0.50, p = 0.006, treatment 26 of 531 (4.9%), control 48 of 531 (9.0%), NNT 24, odds ratio converted to relative risk, active CBD100 users.
risk of case, 32.9% lower, RR 0.67, p = 0.009, treatment 75 of 1,212 (6.2%), control 108 of 1,212 (8.9%), NNT 37, odds ratio converted to relative risk, all CBD100 users.
Shover, 8/5/2022, retrospective, USA, peer-reviewed, 7 authors, study period 12 February, 2020 - 27 February, 2021.
Submit Corrections or Updates.
 risk of death, 1.8% lower, RR 0.98, p = 0.56, treatment 3 of 69 (4.3%), control 199 of 1,762 (11.3%), odds ratio converted to relative risk, propensity score matching.
risk of mechanical ventilation, 5.1% lower, RR 0.95, p = 0.02, treatment 3 of 69 (4.3%), control 292 of 1,762 (16.6%), NNT 8.2, odds ratio converted to relative risk, propensity score matching.
risk of ICU admission, 8.6% lower, RR 0.91, p = 0.02, treatment 8 of 69 (11.6%), control 543 of 1,762 (30.8%), NNT 5.2, odds ratio converted to relative risk, propensity score matching.
risk of oxygen therapy, 2.6% lower, RR 0.97, p = 0.27, treatment 35 of 69 (50.7%), control 1,417 of 1,762 (80.4%), NNT 3.4, odds ratio converted to relative risk, propensity score weighting.
Torrejón-Guirado, 5/7/2026, retrospective, Spain, peer-reviewed, 3 authors.
Submit Corrections or Updates.
 risk of case, 153.7% higher, RR 2.54, p = 0.001, treatment 22 of 156 (14.1%), control 67 of 895 (7.5%), adjusted per study, odds ratio converted to relative risk, last month use.
risk of case, 51.5% higher, RR 1.51, p = 0.33, treatment 22 of 164 (13.4%), control 67 of 887 (7.6%), adjusted per study, odds ratio converted to relative risk, last year use.
risk of case, 35.1% higher, RR 1.35, p = 0.21, treatment 39 of 382 (10.2%), control 50 of 669 (7.5%), adjusted per study, odds ratio converted to relative risk, lifetime use.
Footnotes
a.
Viral infection and replication involves attachment, entry, uncoating and release, genome replication and transcription, translation and protein processing, assembly and budding, and release. Each step can be disrupted by therapeutics.
b.
When administered late in infection, HCQ may enhance viral egress by further increasing lysosomal pH beyond the effect of ORF3a's water channel activity, thereby promoting lysosomal exocytosis, inactivating degradative enzymes, and facilitating the release of SARS-CoV-2 particles into the extracellular environment109,110. Research also suggests potential cardioprotective effects at lower doses, but cardiotoxicity with excessive dosage111. Bobrowski et al. also indicate negative effects if HCQ and remdesivir are combined.
c.
Peters (B) et al. is subject to confounding by calendar-time (SOC evolved rapidly early in the pandemic, the linear covariate does not reflect non-linear SOC changes and hospital specific effects), hospital type (non-treatment hospitals were tertiary university centers), confounding by indication (4/7 hospitals initiated treatment on deterioration), immortal-time bias for as-treated (exposure assigned after baseline), significant differences for other experimental treatments, potential overadjustment from collider bias (steroid use and indication bias), limited baseline severity information, differences in hospice referral propensity across hospitals, unadjusted difference in time from onset to admission, difference in PCR positivity, and other factors. Mahévas et al. is subject to confounding by hospital (treatment highly dependent on the hospital, different SOC/ICU transfer practices, not included in PS), immortal time (only partly addressed in sensitivity analysis), co-treatment differences, calendar-time (SOC evolved rapidly early in the pandemic), binary coding for age (age ≥65 despite steep age-risk gradient), residual imbalance (variables dropped from PS), a composite outcome dependent on hospital triage/capacity, and other factors.
References