Randomized Double-blind Placebo-controlled Proof-of-concept Trial of Resveratrol for Outpatient Treatment of Mild Coronavirus Disease (COVID-19)

Resveratrol is a polyphenol that has been well studied and has demonstrated anti-viral and anti-inflammatory properties that might mitigate the effects of COVID-19. Outpatients (N=105) were recruited from central Ohio in late 2020. Participants were randomly assigned to receive placebo or resveratrol. Both groups received a single dose of Vitamin D3 which was used as an adjunct. The primary outcome measure was hospitalization within 21 days of symptom onset; secondary measures were ER visits, incidence of pneumonia and pulmonary embolism. Five patients chose not to participate after randomization. Twenty-one day outcome was determined of all one hundred participants (mean [SD] age 55.6 [8.8] years; 61% female) (or their surrogates). There were no clinically significant adverse events attributed to resveratrol. Outpatients in this phase 2 study treated with resveratrol had a lower incidence compared to placebo of: hospitalization (2% vs. 6%, RR 0.33, 95% CI 0.04-3.10), COVID-related ER visits (8% vs. 14%, RR 0.57, 95% CI 0.18-1.83), and pneumonia (8% vs. 16%, RR 0.5, 95% CI 0.16-1.55). One patient (2%) in each group developed pulmonary embolism (RR 1.00, 95% CI: 0.06-15.55). This underpowered study was limited by small sample size and low incidence of primary adverse events. A larger trial could determine efficacy. TRIAL REGISTRATIONS: ClinicalTrials.govNCT04400890 26/05/2020; FDA IND #150033 05/05/2020


Study Rationale
Resveratrol (RV) is a polyphenolic phytoalexin produced by certain plants in response to injury or infection. RV has been associated with a variety of positive health effects in areas of in ammation, cardiovascular diseases, cognitive disease, cancer, diabetes, and infectious disease (including viral diseases) 1,2 . RV is readily available commercially as a dietary supplement produced from plant extracts or by genetically engineered yeast 3 . COVID-19 is the disease caused by a novel coronavirus (SARS-CoV-2) that can result in life threatening complications, including lung injury. Outpatient treatment options for COVID are limited. Multiple lines of preclinical data suggest that RV could be effective against coronavirus disease 2019 ( Figure 1). Background SARS-CoV-2 is characterized by surface spike proteins that bind to the Angiotensin-Converting Enzyme 2 (ACE2) of the respiratory tract. After entry into the cell, a variety of processes occur, including the down regulation of ACE2, subsequent destruction of the pneumocyte, the release of in ammatory mediators, and the subsequent release of cytokines (IL1, IL6, TNF-α) and reactive oxygen species 24,25 . A "cytokine storm" results in further damage to the alveoli and the development of Acute Respiratory Distress Syndrome (ARDS) 25 . Resveratrol's multimodal antiviral, anti-in ammatory, and antioxidant properties as well as its ability to upregulate ACE2 receptors could be helpful in reducing the clinical effects of COVID.

ACE2 Upregulation
In addition to ACE2 being a binding site for coronavirus (CoV), it is also associated with protective effects in SARS induced lung injury 26,27 . ACE2 may attenuate vascular permeability, in ammatory cell in ltration, pulmonary edema, hyaline membrane formation, and prevent acute lung injury 28 . Resveratrol has been shown to upregulate ACE2 29 . A de ciency of ACE2 caused by SARS is associated with lung injury 28 . The upregulation of ACE2 by resveratrol might provide protective effects in COVID-19 28, 30-32 . Anti-viral Effects RV has demonstrated antiviral effects in a variety of animal and human disease 2 . Speci c to CoV, in vitro studies demonstrate that RV inhibits MERS-CoV infection by decreasing nucleocapsid protein resulting in reduced viral production and increased cell survival 33 . Starting at the rst steps in the infection in silico modeling suggests that RV would interfere with the binding of CoV spike protein to the ACE2 receptor ( Figure 1 Anti-in ammatory effects COVID-19 is associated with the potential for excessive in ammation. Coronavirus has been shown to activate Toll-Like Receptor 4, increase pro-in ammatory cytokines IL-1, IL-6, CCL5 (RANTES) and TNF-α leading to an unbalanced in ammatory response and damaging in ammation [34][35][36][37] . In contrast, RV has been shown to reduce in ammation via a variety of mechanisms ( Figure 1) [11][12][13]38 . RV has been demonstrated to inhibit TLR4 activation, decreasing the release of in ammatory cytokines in the macrophages of patients with COPD, and inhibit the proin ammatory transcription factor NF-κB 14,19,39 .
RV has also demonstrated inhibition of pro-in ammatory Th17 helper T-cells (Figure 1) 20 . Inhibition of NF-κB has been shown to increase survival in a mouse model of SARS-COV1 40 .
The anti-in ammatory effects of RV might be bene cial in mitigating the cytokine storm that is associated with ARDS and high mortality of COVID-19 25 . A mouse model of cytokine storm showed a 100% mortality reduced to 0% in RV treated mice with minimal lung injury in the treated group 41 . RV has demonstrated protective effects in LPS induced lung injury, a mouse model of ARDS 42,43 . The proposed mechanism is RV's inhibition of NLRP3 in ammasomes 42 . Inhibition of NLRP3 in ammasomes in another proposed therapeutic target in COVID-19 44 .

Antioxidant Effects
Depletion of the endogenous antioxidant glutathione has been attributed to poor outcomes and death in patient with COVID-19 ( Figure 1) 21 . The use of antioxidants has been proposed in the treatment of COVID-19 45 . RV's antioxidant properties as well as its ability to induce glutathione synthesis might provide additional outcome bene ts 22 .

Animal Models of Viral Infections
As the above discussion regarding RV's effects are largely based on in vitro models of disease, there is always a concern regarding whether in vitro models will translate into in vivo e cacy. Multiple animal studies have shown that RV does improve outcomes in animal models of viral infections. A porcine model of pseudorabies virus, a respiratory illness, shows that piglets inoculated with the virus had no mortality compared to a 40% mortality in the untreated group. Speci cally, that study showed alveolar destruction in the untreated group vs mild lung injury in the treated group. The proposed mechanism is inhibition of IκB kinase by RV 46 . It is notable that a drug prediction analysis of SARS-CoV-2 suggests that IκB kinase inhibition is a potential target for COVID-19 47 . Similarly, a murine model of H1N1 in uenza showed a 60% survival rate in RV treated mice compared to 20% in placebo treated mice 48 . In Respiratory Syncytial Virus (RSV) infected mice, RV treated mice showed signi cantly less lung damage compared to untreated mice 49 .

Vitamin D3
Vitamin D3 was included in the treatment protocol as an adjunct to RV based upon prior research showing that it has synergistic anti-in ammatory effects, inhibiting IL-6 and TNF-α 11 . Both treatment arms received a single 100,000 IU dose of D3 to quickly assure adequate serum concentrations of D3, as well as to potentially remove vitamin D3 de ciency as a confounding variable, noting that multiple publications raised concerns that vitamin D3 de ciency might be associated with worse outcomes in COVID-19 50,51 .

Study Design
Overview This study was a phase 2, double-blind, randomized, placebo-controlled trial to evaluate the safety and explore the e cacy of RV plus vitamin D3 based on the hypothesis that RV with the adjunct vitamin D3 can reduce hospitalization and morbidity in patients with COVID-19. The study was approved by the U.S. Food and Drug Administration as an investigational new drug trial (FDA IND #150033 05/05/2020; ClinicalTrials.gov NCT04400890 26/05/2020), and the intuitional review board of Mount Carmel Health Systems in Columbus, Ohio, USA. All patients were provided informed consent and screening remotely via phone interview, educated via online animated presentation, and e-consented via REDCap electronic data capture tools hosted at the Ohio State University Medical Center and incorporated questions from the REDCap Shared Library 52,53 .
Patients were recruited primarily from the Mount Carmel Health System testing centers by way of "cold calls" to patients 45 and older who tested positive for COVID-19. A few patients were recruited in response to research advertisements in the central Ohio area (social media, radio, and yard sign advertising), as well as physician referrals. Due to pandemic related safety concerns, the patients remained in quarantine within their home with all trial contact via phone, email, and web (REDCap), with contactless delivery of study packets via courier/mail. Packets were delivered within 7 days from the onset of symptoms, typically < 24 hours after consent signed. Due to reports of patients self-medicating with investigational drugs (e.g., hydroxychloroquine) in the setting of COVID-19, the speci c nature of the trial substance was concealed from subjects until after the study was complete. Patients were informed that they were receiving a "commercially available dietary supplements", but the use of RV was not disclosed. The use of Vitamin D3 was open-label for both groups.
Patients were provided with a study packet containing identically prepared capsules containing a 15-day supply of either resveratrol or placebo, a one-time dose of vitamin D3, a thermometer, a pulse oximeter, and an instruction booklet with dosing log.
Data was collected via REDCap surveys on days 1-15, 21, 30, and day 60 with adverse symptoms assessed using selected PRO-CTCAE questions 54 . All patients were given daily online reminders of when to seek medical care based upon CDC recommendations. Primary and secondary outcome measures (including hospitalization, ER visits, history of chest imaging, and pneumonia) were assessed by phone interviews after 21 days from randomization. All radiology reports were reviewed by the principal investigator.

Sample Size Determination
The maximum total number of randomized subjects was capped at 200 by FDA request. Power analyses were conducted for the primary outcome measure (hospitalization) assuming multiple placebo arm hospitalization rates and effect sizes, as well as for secondary outcome measures.
At the time the protocol was developed, the rate of hospitalization among con rmed cases of COVID-19 ranged between 21% in the 45-54 age bracket, up to 31% for patient's >85 55 . A planned sample size of 190 subjects with complete observations yielded 80% power at the 5% two-sided signi cance level to detect a difference in the primary endpoint (hospitalization) rate of 10% in the resveratrol arm versus 25% in the placebo arm.
An interim analyses was completed by an independent data and safety monitoring board. The analayis used the Hwang-Shih-DeCani alpha spending function with parameter gamma = -4 (O'Brien-Fleming-like) for the upper (superiority) bound under the null hypothesis with total one-sided Type I error 2.5%, and for the lower (safety or futility) bound under the alternative hypothesis with total Type II error 20% (80% power). Under the assumption of a binding futility bound and a placebo arm hospitalization rate of 25%, the probability of declaring futility at the interim analysis is 3% if the resveratrol arm hospitalization rate is 10% (alternative hypothesis), 55% if the resveratrol arm hospitalization rate is 25% (null hypothesis), and 75% if the resveratrol arm hospitalization rate is 30%. The R package gsDesign was used to determine stopping boundaries.

Participants
Due to low risk of hospitalization (the primary outcome measure), patients younger than 45 were excluded 55 . Patients were eligible for enrollment if they tested positive for SARS-CoV-2 and had symptoms for less than 7 days by the expected delivery date of study packet. Exclusion criteria included cognitive impairment that would prevent the patient from cooperating with study procedures; asymptomatic patients; known history of cirrhosis, hepatic impairment, or Hepatitis C; known of history of renal impairment as measured by an eGFR of < 60 mL/min; patients receiving chemotherapy or who are on chronic immunosuppressants; allergy to grapes or rice; co-morbidities with a high likelihood of hospitalization within 30 days; currently pregnant; hospitalization; patients taking immunosuppressants and drug interactions in medications with a narrow therapeutic index. Patients on "statins" and PDE-5 inhibitors were instructed to withhold while on the study treatment.
It is notable that the renal disfunction exclusion was an FDA requirement. Prior research has explored possible bene ts of RV for patients with chronic kidney disease 56 . Furthermore, increased plasma levels of RV that might be attained in the setting of kidney disease might be bene cial.

Randomization
The random allocation list was blocked and strati ed by a third-party group. Randomization used balanced blocks of size 2 or 4, selected randomly for each block. Randomization schedules were generated and rejected until the randomization schedule was balanced at 100, 200, and 210 subjects to align with the planned interim and nal analyses, and in case of a 10-subject overrun. During the trial, only the third-party group and Data Safety Monitoring Board (DSMB) had access to the randomization list. The study personnel created identical-looking packets with identical-appearing study agents containing a 15-day dosing regimen according to the random allocation list. Study personnel were blinded to the contents of the distributed packets, with bottles only differentiated by a tamper-resistant serial number label applied by the third-party group which corresponded to the randomization list.
Blinding A disinterested third party (Capital University, Department of Mathematics, Columbus, Ohio) was hired to assign tamper resistant serial number stickers as either RV or placebo based upon the output of randomization script from the R statistical software 57 . The third party, using a two-person team to provided validation, assigned serial numbers to appropriate manufacturer sealed RV or placebo bottles. The prepared bottles were returned to the research team such that the bottle could only be differentiated by the serial numbers. The randomization table of the serial number labels was kept only by the third party and the Data Safety Monitoring Board until the completion of the study.

Intervention
Patients received identically appearing bottles containing 60 identically appearing capsules of either >98% pure trans-resveratrol (from Japanese Knotweed Root, Polygonum cuspidatum extract) (500mg per capsule) or placebo (brown rice our) (both prepared and bottled by Vita-Age, Vancouver, BC) with instructions to take 2 capsules 4 times per day for at least 7 days, and up to 15 days if COVID symptoms persisted. Dosing was determined based upon publish IC50 of resveratrol against MERS-COV and previously published pharmacokinetic literature of resveratrol plus its metabolites. (See the study protocol PDF at www.clinicaltrials.gov/ct2/show/NCT04400890 for detailed dose justi cation and products certi cates of analysis.)

Participant Monitoring and Follow-up
Starting on day 1, and continuing daily for 15 days, subjects were contacted via automated e-mail. Messages includes a reminder to take their study medication as scheduled and complete the daily surveys. Subjects were asked to complete a short questionnaire covering: 1) symptoms they had that day that could be related to COVID-19 (e.g., fever, cough, dyspnea), their frequency and severity; 2) report any other related or non-related medical events; 3) any medications they have taken to relieve symptoms, or other new medications they have not previously reported to study personnel; and 4) any visits they have made to healthcare providers, outpatient centers or hospitals, and details regarding those visits. Subjects received reminders when to seek care if they experience symptoms that are worsening or that are concerning to them.
All subjects provided a surrogate/secondary contact (spouse/family member/friend) in order to determine the subject's status if the subject could not be reached. All patients or their secondary contact were interviewed for follow up after 21 days from randomization (no participants were lost to follow up for their post-21 day follow up brief interview).

Endpoints
Hospitalizations were determined based on query of subject or the subject's secondary contact, and the patient's medical records. Additional outcomes include assessing number of days with fever, and to assess symptoms, including dyspnea and fatigue. Questionnaires to assess symptoms and adverse events were based on the PRO-CTCAE (Supplemental Tables S-2b, and S-3b) 54 .

Statistical Analysis
Data management Anonymized data were extracted from REDCap and processed into a dataset with one row per participant.
Self-reported symptom and adverse event data were retained for every patient contact over the 21 days following randomization. Data were analyzed using Stata version 17 58 .

Primary analysis (including missing data)
The primary analysis is a comparison of the proportion of persons in the two groups who were hospitalized within 21 days of symptom onset. The comparison was evaluated using Fisher's exact test, considering the difference to be statistically signi cant if the two-sided p-value is smaller than 0.05. The analysis uses the intent-to-treat method where all participants are analyzed as part of their randomization group, regardless of whether and when they withdrew from the study and regardless of whether or how well they complied with the study protocol. Missing outcome data were subject to tipping point sensitivity analysis to understand what distribution of missingness, if any, would change the conclusion reached using complete case analysis 59 .

Secondary analyses (including missing data)
Secondary outcomes were analyzed using Fisher's exact test, also, with no adjustment for multiple comparisons. Those outcomes were also subject to tipping point analysis of missing outcome data.

Sub-group analyses
The primary and secondary endpoints were analyzed among planned sizable sub-groups using Fisher's exact test with no adjustment for multiple comparisons. Participants were asked about symptoms a) at enrollment (current symptoms), b) in a daily diary during 15 days of treatment (symptoms today), and c) on days 1, 8, 15, and 21 of the study (over the past 7 days). For questions about presence of a symptom, prevalence was compared using Fisher's exact test. For questions about severity, frequency, or interference with activities of daily living (ADL), the proportion who answered 1+ and the proportion who answered 3+ were compared using Fisher's exact test. Responses at enrollment or on day 1 of the study were used to characterize differences between study groups at baseline. Responses on days 2-21 were used to characterize differences in effects of placebo vs. resveratrol.

Study participants
Between September 13, 2020 and December 11, 2020, 1,694 patients were telephoned within 24 hours of testing positive for COVID-19 to be recruited into the clinical trial ( Figure 2). One-hundred-ve were enrolled and randomized (Table 1). Five withdrew after receiving treatment packets (four withdrew before starting treatment and one withdrew after one treatment day citing "too many pills" as reason for withdrawal).
There was no indication of systematic biases in randomization: 4 / 122 = 3% of hypothesis tests comparing baseline symptoms between randomized groups were statistically signi cant at the 5% level without adjustment for multiplicity, and none were statistically signi cant following a Bonferroni correction. (Supplemental Tables S-1,S-2a, and S-3a). Table 1 Characteristics of participants. Primary endpoint -hospitalization within 21 days One patient (2%) in the RV group and 3 (6%) in the placebo group were hospitalized within 21 days of symptom onset (risk ratio (RR) = 0.33; 95% con dence interval (CI) = 0.04-3.10; Risk difference = -4.0%; 95% CI: (-11.6% -3.6%); Fisher's exact test p-value = 0.617; see Table 2). Tipping point analysis of missing outcome data indicate that no possible combination of outcomes among the ve patients whose data are missing would have yielded a p-value below 0.  All outcomes evaluated over the 21 days that followed patient randomization to study group.
Outcomes observed for N=50 patients per group. NA = not applicable; CI = con dence interval; ICU = intensive care unit; ER = emergency room P-value from Fisher's exact test.

Secondary endpoints
Among secondary endpoints, there were fewer events in the RV group than the placebo group for incidence of pneumonia and for emergency room visits due to COVID (Table 2). Neither difference was statistically signi cant. There was one pulmonary embolism in each group, so those incidence rates were equal across study groups. There were no events and therefore no differences between study groups, for death, invasive ventilation, or ICU admission. If outcomes had been observed for the ve patients who withdrew from the study, no secondary endpoint could have had a statistically signi cant difference between study groups, even if the ve outcomes had been as favorable as possible for RV (Supplemental Table S-4).

Notable events
One patient in the placebo group was diagnosed with pancreatitis that was attributed to COVID-19 by the patient's emergency department physician.

Adverse events
There were no serious adverse events reported. There were no signi cant differences in the proportion of patients from each study group reporting symptoms in a daily diary (Supplemental Table S-2b). When asked to think back over the previous seven days, the placebo group reported more severe dry mouth and more frequent general pain than the RV group, and the latter reported more frequent diarrhea (87.2% vs.

Discussion
Resveratrol is an extensively studied plant phytoalexin that has demonstrated potential bene cial biologic effects in multiple human clinical trials. With respect to COVID-19, multiple publications have suggested its use in humans as a potential treatment. This has been supported by prior research describing resveratrol's poly-mechanistic properties; computerized molecular docking analysis demonstrating resveratrol potential to interfere with coronavirus; as well as multiple in vitro studies demonstrating e cacy against MERS-CoV and SARS-CoV-2.
It should be noted that the much of the resveratrol literature is concerned about poor bioavailability and discounts possible effects of resveratrol metabolites such as the more intravascularly abundant resveratrol-glucuronides 60,61 . This dismissal of resveratrol's metabolites is despite the fact that other drugs have demonstrated increased potency in their metabolized forms (i.e., morphine-6-glucuronide is known to be more potent than morphine) 62 . Molecular docking analysis suggest that resveratrolglucuronide may be more potent against coronavirus since there is a higher binding a nity between resveratrol-glucuronides and coronavirus structures 6 .
Although the primary outcome results are not statistically signi cant, in this phase 2, double-blind, placebo-controlled, randomized clinical trial, resveratrol was associated with a lower incidence of pneumonia, COVID-related ER visits, and hospitalization. The favorable risk ratios could be due to chance, but there are biological reasons to believe that RV would be effective and so the protective effect may be quite real, but not signi cant due to small sample size and low incidence of the outcomes. It is notable that in in uenza, shorter time between the onset of symptoms and the start of antiviral treatment results in improved outcomes such that they CDC primarily recommends starting treatments within 48 hours 63 .
The median time from symptom onset to delivery of treatment packet was 5 days. The magnitude of effect of resveratrol in COVID might be greater if treatment could be started sooner, but due to delays in presentation, test results, and delivery, a 48-hour treatment window was not feasible for this study. There were no serious adverse events attributed to resveratrol in this study, and given resveratrol's long safety history, the data presented here would support a larger clinical trial to determine e cacy, ideally starting treatment shortly after the onset of symptoms.
While the results of this study were largely underpowered due to small sample size, there were a few measures that did reach statistical signi cance. Dry mouth (p=0.046), nausea (p=0.05), and diarrhea (p=0.04) was reported in higher frequency in the RV group. This is certainly consistent with known gastrointestinal side effects of resveratrol.
Resveratrol treated patients had a lower incidence of overall pain (p=0.04). This is consistent with prior preclinical literature demonstrating RV to have analgesic properties as a cyclooxygenase inhibitor (COX I & COX II) 64 . This would also support that orally administer resveratrol is able to achieve system effects despite concerns for limited bioavailability.
This study was a proof-of-concept to primarily determine the safety of using resveratrol in the setting of COVID-19, noting the FDA guidance was to limit this study to no more than 200 participants with a planned interim safety analysis after the rst 100 patients were enrolled. Enrollment in the study was slow initially but did rapidly increase in December as Ohio was starting its third COVID-19 wave.
Enrollment was paused after the 100 th patient so that an interim analysis could be performed. After completion of data collection and an interim analysis by an independent Data Safety Monitoring Board, Ohio's third COVID-19 wave was ending. While the DSMB did approve continuation of the study, a feasibility analysis of daily case rate in the Mount Carmel Health System, and considering the prior rate of enrollment, it was estimated that it would take at least another 6-8 months to enroll another 100 patients. The enrollment rate would further be impacted by the availability of vaccinations and competing treatments (such a monoclonal antibodies). Furthermore, a statistical futility analysis also suggested that 100 more patients would be inadequate to determine e cacy, therefore the study was discontinued after the rst 100 patients.
Additional limitations include limited geographic area, limited racial diversity, and a disproportionate number of heath care providers as subjects in the trial. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
Deidenti ed individual data that supports the results will be shared by written request to the communicating author; provided the requesting investigator has approval from an Institutional Review Con icts of Interest: The authors declare no con ict of interest.