Abstract Background: Identification of therapies to prevent severe COVID-19 remains a priority. We sought to determine whether hydroxychloroquine treatment for outpatients with SARS-CoV-2 infection could prevent hospitalization, mechanical ventilation, or death. Methods: This randomized controlled trial was conducted in Alberta during the first wave of the COVID-19 pandemic without direct contact with participants. Community-dwelling individuals with confirmed SARS-CoV-2 infection (by reverse transcription-polymerase chain reaction [RT-PCR] viral ribonucleic acid test) within the previous 4 days, and symptom onset within the previous 12 days, were randomly assigned to oral hydroxychloroquine or matching placebo for 5 days. Enrolment began on Apr. 15, 2020. The primary outcome was the composite of hospitalization, invasive mechanical ventilation, or death within 30 days. Secondary outcomes included symptom duration and disposition at 30 days. Safety outcomes, such as serious adverse events and mortality, were also ascertained. Outcomes were determined by telephone follow-up and administrative data. Results: Among 4919 individuals with a positive RT-PCR test, 148 (10.2% of a planned 1446 patients) were randomly assigned, 111 to hydroxychloroquine and 37 to placebo. Of the 148 participants, 24 (16.2%) did not start the study drug. Four participants in the hydroxychloroquine group met the primary outcome (4 hospitalizations, 0 mechanical ventilation, 4 survived to 30 days) and none in the placebo group. Hydroxychloroquine did not reduce symptom duration (hazard ratio 0.77, 95% confidence interval 0.49–1.21). Recruitment was paused on May 22, 2020, when a since-retracted publication raised concerns about the safety of hydroxychloroquine for hospitalized patients with COVID-19. Although we had not identified concerns in a safety review, enrolment was slower than expected among those eligible for the study, and cases within the community were decreasing. Recruitment goals were deemed to be unattainable and the trial was not resumed, resulting in a study underpowered to assess the effect of treatment with hydroxychloroquine and safety. Interpretation: There was no evidence that hydroxychloroquine reduced symptom duration or prevented severe outcomes among outpatients with proven COVID-19, but the early termination of our study meant that it was underpowered. Trial registration: ClinicalTrials.gov, no. NCT04329611
Eighteen years ago, the severe acute respiratory syndrome (SARS) experience highlighted limited knowledge of early treatments for novel pandemic respiratory viruses. With the emergence of SARS-CoV-2, early experience in Wuhan, the Lombardy region of Italy, and New York City demonstrated the need to suppress severe disease to prevent health system collapse. Hydroxychloroquine, derived from the centuries-old antimalarial medicine quinine, has broad antiviral effects and immunomodulatory properties, making it an attractive candidate to be repurposed for SARS-CoV-2 infection. The precise mechanisms of immunomodulation are uncertain, but the net result is inhibition of macrophage production of proinflammatory cytokines tumor necrosis factor (TNF)–α and interleukin (IL)–6. Hydroxychloroquine was explored as a putative agent for SARS in 2003, but that epidemic was contained before it could be adequately tested. In vitro effects on SARS-CoV-2 and enthusiasm from preliminary clinical investigations in COVID-19 resulted in its rapid, widespread, off-label use worldwide.–
We began a randomized placebo-controlled trial, leveraging the entire Alberta public health system infrastructure, to assess whether early hydroxychloroquine treatment in outpatients with SARS-CoV-2 infection would prevent progression to a severe disease requiring hospitalization or mechanical ventilation, or resulting in death.
Methods Study design and setting This investigator-initiated, randomized, double-blind, placebo-controlled trial was conducted in Alberta, with enrolment beginning on Apr. 15, 2020. Alberta has a population of 4.4 million, of whom about two-thirds live in urban settings. The protocol is available in Appendices 1, 2, and 3, available at www.cmajopen.ca/content/9/2/E693/suppl/DC1. An independent data and safety monitoring committee provided study oversight. The trial was designed to determine whether early hydroxychloroquine treatment in community-dwelling individuals infected with SARS-CoV-2 prevented progression to severe disease. The publicly funded health system in Alberta is singularly responsible for testing, reporting, and providing health services to all residents, permitting all individuals with confirmed SARS-CoV-2 infection to be identified. Alberta Health Services (AHS) staff obtained permission to share contact information with researchers after results of reverse transcription-polymerase chain reaction (RT-PCR) tests were disclosed to infected individuals. Research coordinators then telephoned individuals who consented to be contacted and discussed the study, conducted screening, obtained informed consent, and randomly assigned eligible participants by telephone. To limit risk to study personnel and enable province-wide participation, all study interactions were conducted by telephone (including obtaining informed consent) or email. The screening was supported by access to the participants’ provincial electronic health record, discussion with a study physician (as needed), and a telephone language translation service used during the calls. Calls, including the consent discussion, were recorded for quality assurance. Participants Adults with SARS-CoV-2 infection confirmed by RT-PCR from a nasopharyngeal or pharyngeal swab within the previous 4 days, with symptom onset within the previous 12 days and with at least 1 risk factor for severe disease (Appendix 1, Table S1) were eligible. Those who were hospitalized, pregnant or breastfeeding, unable to swallow pills or unable to comply with the medical regimen, or had used hydroxychloroquine, chloroquine, lumefantrine, mefloquine, or quinine within the previous 30 days were excluded. Those at higher risk for arrhythmia secondary to hydroxychloroquine, including those concurrently using a drug that prolonged the corrected QT interval (QTc) and those with a modified Tisdale Risk Score of 7 or greater (Appendix 1, Table S2), were excluded. We anticipated that enrolment could be completed between April and September 2020 given the rate of SARS-CoV-2 infection, as we expected up to half of the infected patients would participate. Intervention The hydroxychloroquine dose was 800 mg orally in divided doses on day 1 followed by 200 mg twice daily for 4 days, or an identical matching placebo (12 tablets over 5 days). The study drug was delivered to participants’ homes anywhere in the province by courier. Treatment initiation was confirmed by telephone or email. Outcomes The primary outcome was the development of severe disease defined as the composite of hospitalization, invasive mechanical ventilation, or death within 30 days. Secondary outcomes included 1) days to COVID-19 recovery (symptom duration), defined as the number of days from randomization to symptom resolution; 2) disposition at 30 days, defined as recovered, ongoing symptoms but not hospitalized, hospitalized, or deceased; and the proportion of participants 3) deceased, 4) admitted to ICU and 5) hospitalized, within 30 days. Safety outcomes were the proportion of participants with serious adverse events and the proportion with emesis. Data sources
The primary outcome was obtained from administrative data, including vital statistics, hospital admission dates, intensive care admission, and hospital discharge summaries. The Alberta electronic medical record includes Alberta Netcare, which encompasses all hospitalizations, diagnostic test results, and outpatient pharmacy prescriptions. In addition, the routine administrative data from the Discharge Abstract Database, the provincial vital statistics registry, and the National Ambulatory Care Reporting System were used. Telephone interviews at 7 and 30 days, supported by a review of electronic medical records, determining adherence, adverse events, disposition at 30 days, symptom duration, and care during hospitalizations. Only serious adverse events and the predetermined adverse event of new or worsening emesis — considered because of the potential effect on adherence — were collected. Randomization and blinding Randomization was conducted using a custom-developed online tool to allow for dynamic randomization and allocation concealment. We used a minimal sufficient balance randomization tool to ensure balance on age, sex, risk status (binary variable based on age and other identified risks), days since symptom onset, and provincial health zone (5 categories)Participants were randomly assigned to receive either hydroxychloroquine or placebo in a stochastically governed (not blocked) 2:1 ratio. We chose the 2:1 ratio to encourage participation by allocating a greater chance of receiving the active agent. Masking to allocation sequence was complete because randomization assignment was determined dynamically at randomization. All participants and the research team were blinded except for the research pharmacist and randomization website programmer. Statistical analysis The absolute effect size was estimated based on the Italian experience, assuming that up to 20% of the Alberta population (4.4 million) could acquire SARS-CoV-2 infection (n = 840 000), that 16% of those infected (n = 134 400) could require hospitalization and that 3% of those infected (n = 25 200) could require invasive mechanical ventilation. We estimated the risk of severe disease to be at least twice as high in high-risk populations, so low-risk individuals were excluded. Assuming a 16% rate of the primary outcome, a risk ratio of 0.65, with 2:1 randomization and 85% power, we estimated that 1446 evaluable patients with complete follow-up were required (n = 482 placebo; n = 964 active treatment). Comparisons were conducted under a superiority framework with a 2-sided level of 0.05. Secondary analyses followed a prespecified protected hierarchy; adjustments were not made for multiplicity. Treatment effects were reported with 95% confidence intervals (CIs). The intention-to-treat (ITT) population included all randomized participants. The per-protocol population included participants who were adherent to the treatment, defined as taking 10 or more of 12 tablets. The safety population consisted of participants who took any study drug. We compared the proportion of participants in each treatment group who reached the primary outcome using the Fisher exact test. Although analysis of symptom recovery originally stipulated a semi-competing risks model with a competing risk of death, this was not required as no deaths were observed. Days to symptom recovery were plotted using Kaplan–Meier curves, and a log-rank test was used to test the hypothesis that the recovery-free curves did not differ between treatment and placebo. We estimated hazard ratios for treatment from a Cox proportional hazard regression model. The proportional hazards assumption was assessed graphically and through statistical testing. We tested the proportion of participants with safety outcomes using the Fisher exact test. For assessing symptom duration, participants who recovered before randomization or were asymptomatic were removed, as were those without follow-up at day 7 or day 30 whose symptoms at randomization were unconfirmed. Participants without recovery dates were censored at their last follow-up with a known disposition. When disposition was known only at randomization, participants were considered lost to follow-up for this outcome and censored at day 1. We used the statistical software SAS (SAS Institute Inc.) and R (R Core Team). Study data were collected and managed using REDCap electronic data capture tools hosted and supported by the Women and Children’s Health Research Institute at the University of Alberta.19 Ethics approval The trial was approved by the health research ethics committees of the University of Calgary and the University of Alberta, and all participants provided informed consent.
Results During the study period, 4919 individuals with a positive PCR test for SARS-CoV-2 were identified (Figure 1), and 1207 consented to be contacted. A total of 233 participants were screened, and 148 were randomized, 111 to hydroxychloroquine and 37 to placebo (Figures 1 and. Participant characteristics are provided in Table 1. Telephone translation services were required by 10.6% of those contacted and 8.1% of randomized participants (Appendix 1, Table S3).