Azithromycin versus standard care in patients with mild-to-moderate COVID-19 (ATOMIC2): an open-label, randomised trial
[ad_1]
Introduction
Early in 2020 it was highlighted by in silico and in vitro
screens as a potential candidate therapy to be repurposed for treatment of COVID-19. Macrolides, particularly azithromycin, have previously been used to treat other viral infections, including one in three severe cases of MERS-CoV,
although randomised controlled trial data for its use in any coronavirus disease were absent.
Azithromycin is inexpensive, safe, and widely available, and—stimulated by a small, non-randomised clinical report
—its use in the context of COVID-19 has become widespread in clinical practice and clinical trials.
and against SARS-CoV-2, being shown to reduce viral replication alone
or in combination with hydroxychloroquine.
Azithromycin was also associated with reduced viral load of non-SARS-CoV-2 alphacoronaviruses and betacoronaviruses in children receiving azithromycin during a mass distribution programme.
Although antivirals probably have little efficacy in severe disease after viraemia has peaked,
,
,
azithromycin does have anti-inflammatory properties, including dose-dependent suppression of lymphocyte expression of perforin, and a range of proinflammatory cytokines, including IL-1β, IL-6, and TNF, IL-8 (CXCL8), IL-18, G-CSF, and GM-CSF
,
and other components of the IL-1β–IL-6-induced acute phase response such as serum amyloid protein A.
,
,
,
However, these trials were all done in late-stage, severe disease with 17–40% mortality. They did not study patients at earlier stages of disease in the community and are not able to make conclusions about the effectiveness of azithromycin outside the hospital setting. The efficacy of therapies in COVID-19 depends on the timing in the course of disease and in the populations being studied. Dexamethasone showed a strong survival benefit with treatment in patients with severe COVID-19 but no benefit, or even potential harm, in those not requiring oxygen therapy.
Conversely, studies of neutralising antibodies showed benefits in early disease
but not in patients admitted to hospital.
The antiviral and anti-inflammatory properties of azithromycin are suited to earlier-stage disease; thus, we studied an ambulatory population to establish whether it averts disease progression.
Evidence before this study
We searched MEDLINE and the Cochrane Central register of Controlled Trials with the terms (“azithromycin”) AND (“COVID” OR “COVID-19”) AND (“clinical trials”), until March 25, 2021, with no language restrictions. We identified 42 studies, among which there were four completed randomised trials of azithromycin (with or without hydroxychloroquine) in patients admitted to hospital with severe COVID-19 disease, and three completed randomised trials of azithromycin in patients with mild COVID-19 in primary care. The four randomised trials in patients admitted to hospital assigned 8988 participants to azithromycin or standard care or hydroxychloroquine and found no evidence of a difference in mortality, duration of hospital stay, or peak disease severity. The three trials in primary care settings randomly assigned participants with early disease to 3 days or 5 days of therapy, and only one assessed azithromycin as standalone therapy. That trial was a large, adaptive platform trial in the UK that randomly assigned 540 participants in primary care to 3 days of treatment with azithromycin and 875 participants to standard care alone and found no meaningful difference in time to first reported recovery or in rates of hospital admission (3% in both groups), and there were no deaths. We did not identify any randomised trials in patients with COVID-19 managed in ambulatory care.
Added value of this study
The ATOMIC2 trial was uniquely designed to assess azithromycin as a standalone therapy in those with mild-to-moderate COVID-19 presenting to emergency care but assessed as appropriate for initial ambulatory management without hospital admission. ATOMIC2 also uniquely assessed high-dose, long-duration treatment to investigate the efficacy of putative anti-inflammatory effects. We found that azithromycin 500 mg daily for 14 days did not reduce the proportion of participants who died or required hospital admission from any cause over the 28 days from randomisation compared with standard care.
Implications of all the available evidence
Our findings, taken together with existing data, suggest there is no evidence that azithromycin reduces hospital admission, respiratory failure, or death compared with standard care, either in early disease in the community, or those admitted to hospital with severe disease, or in those with moderate disease managed on an ambulatory pathway.
We did a randomised, open-label clinical trial to establish whether azithromycin is effective in preventing hospital admission or death in adult patients with clinically diagnosed COVID-19 infection being managed on an ambulatory care pathway.
Methods
Study design
and all applicable laws and regulations including, but not limited to, the principles as stated in the International Council for Harmonisation Guideline for Good Clinical Practice, the standards set out by the Research Governance Framework, the Medicines for Human Use (Clinical Trials) Regulations 2004, and the ethical principles that have their origin in the Declaration of Helsinki. Safety data were reviewed and monitored by an independent data safety monitoring committee. The trial protocol was reviewed and approved by the UK Medicines and Healthcare products Regulatory Agency and an independent ethical committee (London—Brent Research Ethics Committee, reference number 20/HRA/2105).
Participants
All patients provided electronic informed consent before randomisation.
Randomisation and masking
Patients were randomly assigned (1:1) to either azithromycin plus standard care or standard care alone using a web-based automated service, with a minimisation algorithm to ensure balanced allocation across treatment groups, stratified by centre, sex, and presence of hypertension and diabetes. To ensure the unpredictability of treatment allocation, the first 30 participants were randomly assigned by simple randomisation and the minimisation algorithm included a probabilistic element (participants had an 80% chance of being allocated to the treatment, which minimised imbalance between the groups). Patients, investigators, and health-care providers were not masked to study drug assignment.
Procedures
(a widely-used consensus set of clinical outcomes arising from a Delphi survey), and a nine-level severity score of respiratory illness (0–8, where 0 indicates “Ambulatory. No limitation of activities” and 8 indicates “Death”). Participants also had a 12-lead electrocardiogram. Optional study samples were taken at baseline and on one further occasion if the participant was admitted: oropharyngeal swab for SARS-CoV-2 PCR and nasal and blood samples for RNA transcriptomic analysis. Bloods and chest x-rays were done if clinically required.
Patients in the azithromycin group received 500 mg azithromycin once daily orally plus standard care for 14 days and those in the control group received standard care according to local guidelines. Use of corticosteroids, other immunomodulators, antibiotics, and antivirals was permitted after randomisation, but the protocol excluded concomitant use of quinolone or macrolides antibiotics at enrolment or during follow-up. Subsequent assessments were conducted by telephone at days 14 and 28, and radiology results and clinical notes were assessed daily during hospital admission if this occurred.
Outcomes
The primary outcome was the proportion of participants with hospital admission or death from any cause within 28 days from randomisation.
Statistical analysis and protocol changes
that the primary endpoint should be responsive to the eligible patient population and the definition of the endpoint should be fine-tuned for the pivotal phase, based on the pilot phase of the trial. This change was approved by the research ethics committee and the Medicines and Healthcare products Regulatory Agency on Feb 4, 2021, and implemented before final analyses were done. Two other protocol amendments were made while the trial was ongoing to broaden inclusion criteria to include and ensure the safety of participants taking selective serotonin reuptake inhibitors (appendix p 144).
an interim analysis was planned to establish a definitive sample size. Following the revision to the primary outcome and based on masked data from the pilot phase, the definitive sample size was determined. Assuming a 15% rate of all-cause hospital admission or death in the standard care group, we estimated that a minimum of 276 participants providing primary endpoint data would provide 80% power and 5% (two-sided) significance to detect a difference from 15% to 5% in the azithromycin group, a relative reduction of 66%. To allow for 5% loss to follow-up we therefore aimed to recruit a minimum of 291 participants. The full statistical analysis plan is in the appendix (pp 40–86).
For the primary outcome, the difference in proportions between the treatment groups was assessed using a χ2 test and a 5% (two-sided) significance level. Adjusted analysis was done using logistic regression with progression as the binary outcome, adjusting for the following stratification factors: centre, hypertension, diabetes, and sex. A supporting analysis was also done to further adjust for the following important prognostic variables: age 65 years and older, presence of chronic lung disease, and treatment for cancer. Time-to-event analysis was also done to explore whether the active treatment delays progression. The success of the trial was based on the adjusted analysis. Both relative and absolute differences in proportions are reported together with 95% CIs. Other binary outcomes were assessed using similar methods. Peak severity of illness was considered a categorical variable and assessed using ordinal logistic regression analysis. The change in severity scale score from baseline was summarised on a continuous scale using means, SDs, medians, IQRs, and ranges. We explored consistency of results for the following prespecified stratification factors: hypertension, diabetes, sex, and age using treatment by variable interaction tests and forest plots. Analyses were done using Stata IC, version 15.1.
Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.
Results
Figure 1Trial profile
*These participants withdrew completely and asked for all their data collected to date to be removed; therefore, they have not been included in any summaries or analyses.
Table 1Baseline characteristics and concomitant treatments of the intention-to-treat population
Data are n (%) or mean (SD). COS=core outcome set.
Table 2Comparison of primary and secondary binary outcomes, and time-to-event in the ITT populations
Comparisons for the primary outcome were performed using a logistic regression model adjusted for stratification factors centre, hypertension, diabetes, and sex (adjusted), or fully adjusted for centre, hypertension, diabetes, sex, age 65 years and older, presence of chronic lung disease, and treatment for cancer. Comparisons for the secondary outcomes were performed using Fisher’s exact test because of small numbers of events. ITT=intention to treat.
Figure 2Time to hospital admission, length of stay, and time to death in the 32 participants admitted to hospital during the study
One participant in the azithromycin group was known to have been admitted to hospital but exact dates were not available so details of their stay could not be included in the plot.
Figure 3Kaplan-Meier plot of time to hospital admission in the intention-to-treat population (A) and in the intention-to-treat positive population (B)
Figure 4Severity scores at days 0, 14, and 28 in 254 participants with complete data from the intention-to-treat population
Table 3Comparison of peak severity scores in the intention-to-treat-population
odds ratio for peak severity scores was 0·91 (95% CI 0·57–1·46), p=0·69.
Discussion
In this trial of people with clinically diagnosed mild-to-moderate COVID-19 managed without hospital admission, adding azithromycin to standard care treatment did not reduce the risk of subsequent hospital admission or death, or of time to hospital admission.
Likewise, COALITION II randomly assigned 447 patients admitted to hospital to azithromycin 500 mg daily for 10 days or standard care, and again found no difference in day 15 ordinal score (OR 1·36 [0·94–1·97], p=0·11).
RECOVERY randomly assigned 7763 of its participants to azithromycin 500 mg for 10 days or standard care and found no difference in 28-day mortality (rate ratio 0·97 [95% CI 0·87–1·07], p=0·50), length of stay, or invasive mechanical ventilation and death.
However, none of these trials assessed the potential for efficacy in early, milder disease.
or 5 days
,
of therapy. Azithromycin was assessed as standalone therapy only in the PRINCIPLE trial;
a large, adaptive platform trial in the UK, which randomly assigned 540 participants to 3 days treatment with azithromycin 500 mg daily versus 875 participants to standard care. This study found no difference in time to first reported recovery (HR 1·08 [95% Bayesian credibility interval [BCI] 0·95 to 1·23]), and, although only 3% of participants were admitted to hospital, there was no significant difference between groups (absolute benefit in 0·3% [95% BCI −1·7 to 2·2]). The remaining two trials also used short courses of azithromycin 500 mg for 1 day followed by 250 mg for 4 days and taken in conjunction with hydroxychloroquine. Q-PROTECT recruited healthy, SARS-CoV-2-positive men in a quarantine site in Qatar and found no difference in time to virological cure (p=0·82),
with low rates of hospital admission in all groups (2·4%). A study in the USA assessed progression to lower respiratory tract infection, hospital admission, or death and time to viral clearance in SARS-CoV-2-positive outpatients, but was stopped early for futility because of a low rate of clinical outcomes in this population, and found no difference in the co-primary outcome of time to virological clearance (HR 1·25 [95% CI 0·75 to 2·07], p=0·39).
No studies have assessed azithromycin in patients presenting to hospital with substantial symptoms, but early enough in the disease process to be managed in ambulant care, and neither have previous studies assessed high-dose, long-duration azithromycin therapy in early disease.
Our study investigated this intermediate population with early disease, but at high risk of deterioration, in whom 11% required subsequent hospital admission. Therefore, our population represents those with the optimal chance of demonstrating clinical benefit in early disease. We did not observe a significant difference in our primary outcome. Given the small absolute event rates for the primary outcome in our study, a smaller but clinically relevant effect cannot be entirely ruled out, but would be unlikely to change clinical practice. Nonetheless, this finding, taken together with clear negative results across the disease course from early, low-risk patients, to patients admitted to hospital with severe disease, provides strong confirmation that azithromycin is not effective in treating COVID-19.
Another strength of our study is that in contrast to other studies, the high dose (500 mg daily) and long duration (14 days) of azithromycin was used to ensure that we adequately assessed potential antiviral, antibacterial, and anti-inflammatory benefits. COVID-19 is considered to have a distinct early viraemic phase and a late inflammatory phase in some individuals, and therefore assessment of antiviral activity needs to be early in the disease course before onset of severe disease.
Conversely it was not known what doses might be required to produce an adequate anti-inflammatory effect so it was necessary to give a high dose of long duration to ensure the anti-inflammatory effect was tested throughout the late stage of innate or acute phase inflammatory cytokine dysregulation.
An additional strength is that we were also able to exclude a significant benefit from azithromycin’s antibiotic effects, which was not possible in studies of patients in hospital where co-prescribing of β-lactam and other antibiotics was common.
Our data show that secondary bacterial infection is not a major driver of hospital admission in this population.
we used a clinical diagnosis for inclusion, rather than requiring PCR confirmation, and PCR data were not available for all participants—particularly at the early stages of the pandemic in the UK where low testing capacity was directed to patients who needed admission to hospital. While it is likely some participants who did not ultimately have COVID-19 might have been enrolled, this decision reflects the situation in many urgent care settings globally where PCR confirmation is not immediately available, and enhances the generalisability of our findings. Nonetheless, SARS-CoV-2 was detected in 66% of those with successful PCR assays in our study, which is much higher than the 31% PCR-positive rate observed in PRINCIPLE,
and study results were similar in the overall ITT group and the predefined PCR-positive subgroup analysis. An ITT analysis was selected as the most appropriate approach to establish whether this intervention affects clinical outcomes. While throat swab PCR assays have high specificity, they have low sensitivity in routine clinical practice, and consequently where there is a high pre-test probability for COVID-19, as was the situation in those enrolled, a negative PCR result has a low negative predictive value and most negative results will be false negatives.
A third limitation is the relatively young mean age of the study population (45·9 years), which reduces the proportion who are likely to have severe disease. Nonetheless, the primary outcome occurred in more than 10% of participants, and globally, individuals of similar age could have been receiving azithromycin therapy in many countries. Other limitations are incomplete compliance to the long treatment regimen in some individuals and absence of data on microbiology or long-term outcomes beyond 28 days.
Given positive data from in silico and in vitro
screens and data showing suppression of innate inflammatory cytokines in vitro, and clinical trial data in non-SARS-CoV-2 alphacoronaviruses and betacoronaviruses,
why might these have not translated into clinical efficacy? Other antiviral molecules have had little clinical effect in COVID-19 compared with immunosuppressive therapies, except in very early disease. In common with influenza, antivirals are probably only efficacious in the early viraemic disease stage and are ineffective in severe disease, which is more closely linked to differences in host immune factors. In contrast to influenza A pandemics, the antibacterial effects of azithromycin are unlikely to translate into significant clinical benefit in a disease where secondary bacterial pneumonia is rare.
While many studies have shown azithromycin suppression of innate cytokines known to be key mediators of severe disease, including IL-1β, IL-6, CXCL-8, TNF, and GM-CSF, some of these data might be confounded by antibacterial effects in the original studies, and it might be that the suppression achieved by azithromycin is simply insufficient to overcome the overwhelming cytokine production triggered by this virus in susceptible individuals.
In a Danish cohort analysis of 10·6 million prescriptions, azithromycin prescribing has been associated cardiovascular death (rate ratio 2·85 [95% CI 1·13–7·24]) compared with no antibiotics, but this is probably because of the underlying indication because, when compared with penicillin V, there was no increased risk once adjusted for propensity scores (rate ratio 0·93 [0·56–1·55]).
This analysis was performed in a population with a low baseline risk of cardiovascular death, and it should be noted that we excluded patients with a prolonged QT interval at baseline electrocardiogram. Nonetheless, there are considerable population risks of unwarranted prescribing of azithromycin, which is a highly valuable antimicrobial and yet has a particularly high propensity for inducing antimicrobial resistance, both to macrolides and to other drug classes, including β-lactam antibiotics.
In conclusion, our findings in mild-to-moderate COVID-19 managed in ambulatory care, taken together with trials in early disease in primary care and from trials in patients admitted to hospital with severe disease, suggest that azithromycin does not reduce hospital admissions, respiratory failure, or death compared with standard care, and should not be used in the treatment of COVID-19.
TSCH, VSB, JB, SJD, MJ, JM, DL, IDP, and DR contributed to conceptualisation and design of the protocol. SJD performed the power calculation. TSCH, LC, VSB, SM, JLC, JM, MJ, DL, PM, RG, TB, GJ, FC, DC, SE, JU, LC, and DC contributed to acquisition of study data. Data were analysed by AW, SJD, and RK. AW, RK, SJD, and LC have accessed and verified the data. TSCH drafted this submission, which was approved by all authors. All authors had full access to all the data in the study and final responsibility for the decision to submit for publication.
Declaration of interests
TSCH has received grants from Pfizer, University of Oxford, the Wellcome Trust, The Guardians of the Beit Fellowship, and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC) during the conduct of the study; and personal fees from Astra Zeneca, TEVA, and Peer Voice, outside of the submitted work. MJ has received grants from the University of Oxford and NIHR Oxford Biomedical Research Centre. DR has undertaken paid consultancy for GlaxoSmithKline outside of the submitted work. IDP reports personal fees from AstraZeneca, Boehringer Ingelheim, Aerocrine, Almirall, Novartis, GlaxoSmithKline, Genentech, Regeneron, Teva, Chiesi, Sanofi, Circassia, and Knopp; and grants from NIHR, outside of the submitted work. JU has received honoraria for preparation of educational materials and has served on an advisory board for Gilead Sciences and ViiV Healthcare, outside of the submitted work. All other authors declare no competing interests.
Acknowledgments
We are grateful to all the participants who volunteered and to the clinical and research teams at all the participating centres. We are grateful to James Chalmers, Chris Rogers, and Peter Howarth (data safety monitoring committee), to Anne Francis, Rebecca Brown, Elizabeth Hamilton, Samuel Hinks, and Samuel Mills (design, operational, and statistical support), Mona Bafadhel (protocol design), Lucy Eldridge and Patrick Julier (software programming), and to Paul Little, Mike Bradburn, Peter McQuitty, Najib Rahman, and Dominick Shaw (trial steering committee), and to Lou Chan, Ben Bloom, Thomas Knight, Richard Procter, Claire Millins, David Connell, and Kay Adeboye for site participant recruitment. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. The authors vouch for the integrity and completeness of the data and the fidelity of the trial to the protocol. This study has been conducted as part of the portfolio of trials in the registered UKCRC Oxford Clinical Trials Research Unit at the University of Oxford. It has followed their standard operating procedures ensuring compliance with the principles of Good Clinical Practice and the Declaration of Helsinki and any applicable regulatory requirements. This research is funded by the NIHR Oxford BRC, by the University of Oxford, and by an independent research grant from Pfizer. TSCH is supported by a fellowship from the Wellcome Trust (211050/Z/18/z). This research is supported by the NIHR Applied Research Collaboration West Midlands through funding to DL. The funders played no role in the study design or decision to submit for publication. This article is dedicated to the memory of Peter McQuitty.
[ad_2]
Source link
:%20an%20open-label,%20randomised%20trial){kind=link}