Since the start of the COVID-19 pandemic, efforts have been made to show the role that antibiotics associated with antivirals, anti-inflammatories, and other immunomodulatory drugs may play in order to define an effective therapy against COVID-19.
Some authors think that the difficulty in finding antiviral treatments with proven efficacy along with the anxiety and uncertainty that this generates in physicians has likely led to the uncontrolled prescription of antibiotic therapy in patients worldwide . Indeed, emerging data show that more than 90% of COVID-19 patients receive antibacterial drugs [27, 28].
In the Chinese city of Wuhan, where the pandemic started, most patients with COVID-19 seem to have received empiric antibiotic therapy, mostly respiratory fluoroquinolones . The use of antifungal drugs and corticosteroids was more limited. Similar data are described in other studies in China, revealing use of antibiotic therapy in more than half of hospitalized patients [30,31,32,33].
In the United States of America, the strategy for empiric antibiotic therapy has been along these same lines. More prevalent antibiotic use was revealed in ICU patients, where 94.9% (224/236) were on antibiotics . In another series in Detroit, antibiotic use in 69.2% (148 of 214 patients) of patients admitted to the conventional ward was documented; their study population had baseline characteristics that were similar to ours .
Langford et al. have conducted a rapid systematic review that determined that the majority of patients with COVID-19 received antibiotics (71.8%, 95% CI 56.1–87.7). The most common were broad-spectrum antibiotics, with fluoroquinolones and third-generation cephalosporins representing 74% of the antibiotics prescribed .
The work by Beovic et al. consisted of a survey aimed at doctors in Europe. As was the case in Asia and America, the study revealed indiscriminate use of broad-spectrum antibiotic therapy. In particular, the study highlights that Spain is one of the countries with the highest rates of antibiotic use—only 22.7% of patients with COVID-19 in the conventional ward were not routinely prescribed antibiotics—behind only Italy (18.2%) and Turkey (19.6%) .
What causes the indiscriminate use of empiric antibiotic therapy in COVID-19 patients?
Antibiotics are usually prescribed in light of the possibility that these patients may have a bacterial infection associated with the ailment that is either concomitant with the initial viral infection or in relation to an extended hospital stay [38, 39].
It is known that bacteria (especially Streptococcus pneumoniae and Staphylococcus aureus) as well as other viral or fungal co-infections are frequent complications that occur in seasonal influenza outbreaks which contribute to increased morbidity and mortality in these patients [40,41,42]. Previous studies have documented that fatality associated with viral pneumonias may be influenced by multiple factors, one of the most prominent being bacterial co-infection [43, 44]. In fact, most bacterial co-infections linked to a primary viral infection are seen in influenza cases . Several studies from the USA and Australia found that in the 2009 H1N1 flu pandemic, 4–33% of patients hospitalized due to that disease had bacterial pneumonia [45,46,47,48,49].
Co-infection by bacteria and viruses in respiratory infections is not only restricted to influenza. Similar conditions have also been reported in other respiratory viruses such as the parainfluenza virus, respiratory syncytial virus, adenovirus, rhinovirus, human metapneumovirus, and even in pathogens similar to SARS-CoV-2 such as SARS (Severe Acute respiratory syndrome) and MERS (Middle-East respiratory syndrome) [50,51,52,53].
Nevertheless, the current evidence on SARS-CoV-2 indicates that the risk of bacterial co-infection upon admission is minimal, though risk increases progressively during hospitalization and critical patients are at highest risk . In several studies conducted in China and Italy, rates of bacterial infection of < 10% were found [55, 57]. In a meta-analysis by Langford et al., in which a total of 1308 publications were reviewed with 24 studies included in the final statistical analysis, the presence of bacterial infection was assessed in 3338 patients and found in 281 of them (8.4%) .
Although the actual prevalence of bacterial infection in patients with SARS-CoV-2 pneumonia has not been fully demonstrated and further studies are needed, several clinical guidelines advocate for using empiric antibiotic therapy in patients with COVID-19, especially in critically ill patients [58, 59]. Many guidance documents recommend antibiotic treatment for patients with COVID-19 and ‘pneumonia’ .
In the survey of European doctors carried out by Beovic et al., nearly two-thirds of participants reported that they did indeed have local guidelines regarding antibiotic use in patients with COVID-19 , but more often than not, they followed their hospital’s community-acquired pneumonia guidelines . Most professionals opted for coverage of pathogens that cause atypical pneumonia. However, these guidelines appear to be grounded in the experience gained in studies of co-infection in patients with influenza, in which the majority were caused by Streptococcus pneumoniae and Staphylococcus aureus . In light of this, several authors recommend that if antibiotics are considered, a beta-lactam providing coverage for S. pneumoniae ± methicillin-susceptible S. aureus should be the first . In contrast, other researchers, such as the Greek group Karampela et al., recommend fluoroquinolones when starting antibiotic therapy  based on the fact that these quinoline derivatives (the prodrome of chloroquine) appear to have an ability to suppress SARS-CoV-2 replication by exhibiting a stronger capacity for binding to its main protease than chloroquine and antiretrovirals such as nelfinavir [62, 63].
The Spanish group García-Vidal et al. aimed to determine the epidemiology, impact, and outcomes of co-infections in a cohort of 989 consecutive patients hospitalized with COVID-19 . A total of 88 co-infections were documented in 72 patients (7.3%). They recommend using empiric antibiotic therapy only in COVID-19 patients who had a chest x-ray suggestive of associated bacterial pneumonia, those who required admission to the ICU, and those who were previously immunosuppressed.
We conclude that the use of antibiotic therapy has been unreasonable given that nearly 90% of patients admitted to internal medicine departments received them empirically (12,238 of 13,932 patients, 87.8%). The most used antibiotics were beta-lactams (72.0%), macrolides (60.2%), and fluoroquinolones (13.3%), which is in line with the available data from the rest of EU (European Union). This pattern of use can plausibly be attributed to the fact that empiric use of third-generation cephalosporins together with azithromycin was included in most hospital protocols in the first months of the pandemic.
The vast majority of our patients had community acquisition of COVID-19; only 6.6% acquired the infection in a hospital. Also of note is the fact that infection in nursing homes occurred in < 10% of cases. Antibiotic use, and specifically macrolide use, correlated to where the infection was contracted: their use was more common among those with community-acquired infection and less common among those who contracted the disease in nursing homes or the hospital.
For which patient profiles should antibiotic therapy be considered?
There appears to be broad consensus on initiating antibiotic treatment in all severely ill patients who require direct admission to the ICU upon arrival at the hospital [24, 59]. However, most authors coincide in highlighting the difficulty of distinguishing SARS-CoV-2-related pneumonia versus atypical pneumonia or nosocomial ventilator-associated pneumonia in COVID-19 patients based on symptoms alone, given that all present with similar signs and symptoms consisting of fever, dry cough, dyspnea, and bilateral involvement on imaging tests. For this reason, they argue that physicians should avail themselves of analytical results when making a decision on whether or not to use antibiotics [10, 26, 32, 39, 65].
Indeed, this is precisely what is being done on a daily basis at the patient's bedside. In research by Beovic et al., physicians indicated that patients’ clinical presentation was the most significant factor when considering starting antibiotic therapy, followed by elevated inflammatory parameters on laboratory tests and radiological findings of pneumonia. Among the analytical results, the most relevant were elevated procalcitonin levels, the neutrophil count, the degree of leukocytosis, and elevated C-reactive protein (CRP) levels .
In our population, we found that the most critical clinical information used when determining whether to begin empiric antibiotic therapy in COVID-19 patients was symptoms such as the presence of fever, dyspnea, and cough (especially productive) were similar to what was reported in the literature. Other symptoms that are more closely related to viral infections, such as arthralgia; fatigue; anorexia; and gastrointestinal symptoms such as nausea, vomiting, and diarrhea, are also associated with greater use of antibiotics. On the other hand, the presence of anosmia, ageusia, headache, or abdominal pain did not seem to have an influence on antibiotic use. The most relevant data on the physical examination were those that reflected more severe disease: oxygen saturation < 90%, tachypnea, and tachycardia. Furthermore, patients who had crackles and rhonchi were more likely to receive antibiotics, findings that were statistically significant; those with wheezing were also more likely to receive antibiotics, but this finding was not significant.
In regard to patients’ previous treatment, it would be logical to believe that those on immunosuppressive treatments would have received antibiotics at a higher rate, but no differences were observed in antibiotic use according to prior immunosuppressive treatment and as such, these drugs were not found to be critical in decision-making regarding use of antibiotics. Only those taking hydroxychloroquine were observed to have received antibiotics more often. Among the group that received macrolides, antibiotics were used less frequently among those being treated with systemic corticosteroids or biological therapies.
Concerning the influence of analytical parameters on the decision to start antibiotic therapy, the results are clear: the elevation of inflammatory parameters such as CRP, procalcitonin, ferritin, LDH (lactate deshidrogenase), and D-dimer have proven to be the most relevant factors in the decision to begin antibiotic treatment, as indicated in previous works. Leukocytosis, interpreted as a sign of risk of bacterial infection, was related to greater use of antibiotics whereas lymphopenia, more often linked with viral symptoms, was inversely related to the use of antibiotics.
Rapid characterization of co-infection is essential in order to properly guide antibiotic management and could help to save lives during the pandemic . Huttner et al. recommended that in cases in which antibiotics are to be started, microbiological samples such as a urinary antigen test for Legionella and blood cultures, should be obtained beforehand in order to diagnose the co-infection . Mirzaei et al. also advocated for a proper diagnosis, noting the importance of a broad-spectrum molecular diagnostic panel for rapid detection of the most common respiratory pathogens .
We believe that actively searching for possible bacterial co-infection and early diagnosis are aspects of caring for COVID-19 patients that must be improved. A urinary antigen test for Legionella and S. pneumoniae was performed in less than half of patients and though there was a very small rate of positive tests (1.5%), mortality was found to be higher among those who did test positive. Antibiotic therapy was used less frequently in patients who did not have a urinary antigen test, but this is likely due to little suspicion of initial bacterial co-infection that resulted in these patients not being prescribed antibiotics. Unfortunately, we do not have information on blood or sputum cultures; this is a possible area of future research.
Comparisons to other studies
Other retrospective case series similar to ours found. A work by Argenziano et al. analyzed the first 1000 patients hospitalized for COVID-19 in the New York City region . The mean age was 63.0 years and predominantly male (57.5%). There were high rates of baseline comorbidities, the most common of which were hypertension and diabetes mellitus. The most common symptoms on admission were dry cough (73.2%), fever (72.8%), and dyspnea (63.1%). They also report that patients with marked elevation of inflammatory parameters (CRP, ESR -erythrocyte sedimentation rate-, D-dimer, ferritin, and LDH) were those who most frequently required transfer to the ICU. In this series, 21.1% of patients across all levels of care died (14% when only considering patients in conventional wards).
Suleyman et al., in a series of 463 cases in Detroit, studied a population with a mean age of 57.5 years that was predominantly female (55.9%) and African American (72.1%) . Virtually all patients (94%) had at least one comorbidity, the most common of which were hypertension (63.7%), chronic kidney disease (39.3%), and diabetes (38.4%). They had similar symptoms upon admission as those in our study: cough (74.9%), fever (68.0%) and dyspnea (60.9%). A higher death rate (20%) was observed in this work compared to previous studies, with male gender and age (over 60 years) shown to be the most relevant risk factors.
In Liang et al.’s work on a cohort of 1590 cases in China, a younger mean age was observed: 48.9 years. Nine hundred and four (57.3%) patients were male and 399 (25.1%) had comorbidities, including hypertension (16.9%), diabetes (8.2%), and cardiovascular disease (3.7%). Fever (88.0%), dry cough (70.2%), fatigue (42.8%), productive cough (36.0%) and shortness of breath (20.8%) were the most common symptoms . The overall rates of severe cases and fatality was 16.0% and 3.2%, respectively.
We found higher death rates in our patient sample compared to previous research. The overall fatality rate was 20.7% (2840 of 13,736 patients). A striking finding was the higher death rate among those who received any antibiotic (OR 1.39, 95% CI 1.20–1.61) except macrolides, in which there was a higher survival rate (OR 0.70, 95% CI 0.64–0.76; p < 0.001). Even considering that use of antibiotic therapy was lower in patients who a priori had a higher risk of dying, namely older or more frail patients, the relationship between antibiotic therapy and fatality persisted even after controlling for these confounding favors on the logistic regression (OR 1.52, 95% CI 1.29–1.80).
In terms of the clinical progress of patients in whom antibiotics were used, improvement was observed in most inflammatory parameters, though there was radiological worsening, with an increase in the proportion of patients with consolidation or interstitial infiltrates. Moreover, antibiotics did not diminish the risk of developing bacterial co-infections among hospitalized patients, as bacterial pneumonia was found in 1481 patients (10.8%) and it was more frequent in those who received antibiotics.
Other complications occurred more frequently during hospitalization, including acute respiratory distress syndrome, acute cardiac failure, arrhythmias, acute renal failure, shock or sepsis, and increased demand for respiratory support (oxygen via high-flow nasal cannula, non-invasive mechanical ventilation, invasive mechanical ventilation, and prone positioning). A higher percentage of patients in the group that received antibiotics required ICU admission. These findings could possibly be explained by the fact that use of empiric antibiotic therapy was widely generalized; its use was only limited among patients who were very frail (and thus not candidates for invasive measures) or, on the contrary, among patients with very mild symptoms.
The role of macrolides
Macrolides have been proposed as a possible treatment for severe acute respiratory distress syndrome caused by COVID-19 since the first months of the pandemic [21, 23]. These bactericidal antibiotics are widely used in habitual clinical practice against gram positive and atypical bacteria species that are usually associated with respiratory tract infections. The antiviral effects of macrolides have attracted considerable attention. Their ability to modulate the immune response and decrease the production of inflammatory cytokines makes them a very interesting tool for battling respiratory viral infections. The efficacy of macrolides in the treatment of other respiratory viruses such as rhinovirus, respiratory syncytial virus, and influenza has long been established [22, 25]. In addition to the aforementioned respiratory viruses, azithromycin has also been reported to inhibit Zika virus .
In terms of COVID-19, azithromycin was one of the drugs included in the large adaptive RECOVERY trial . Based on preclinical and clinical evidence and some preliminary results in COVID-19 patients, azithromycin could have potential in the fight against this new disease .
In a clinical trial led by Gautret et al. in France, a combination of hydroxychloroquine and azithromycin was shown to be effective against COVID-19 . Treatment efficacy was compared in 36 patients divided into three groups: six patients were treated with hydroxychloroquine combined with azithromycin, 14 with hydroxychloroquine in monotherapy, and 16 with a placebo. The results showed that by the sixth day of treatment, all patients in the HCQ + AZM group had no detectable virus in nasopharyngeal exudate samples compared to 57.1% of the HCQ group and 12.5% of the control group (p < 0.001).
In our study, we found a favorable outcome with the use of macrolides compared to other antibiotics. As we have highlighted, the mortality rate was lower in the macrolides group (unlike with other antibiotics) and indeed, the survival ratio was higher among patients who received them, a finding that was statistically significant (OR 0.70, 95% CI 0.64–0.76). Patients in whom macrolides were used were younger than those who received other antibiotics (68 years vs. 71 years). In order to control for possible confounding variables, a multivariate analysis was conducted that showed that the use of macrolides in our population continued to be linked to a lower mortality rate (OR 0.80, 95% CI 0.73–0.88).
Huttner et al. consider that macrolides and quinolones should be avoided due to the risk of cardiotoxicity . Along these lines, a lower rate of use of azithromycin was observed among patients with previous heart disease in our study.
The risk of a rise in multidrug-resistant germs due to indiscriminate antibiotic use has been described in the literature [70,71,72]. The exact incidence of bacterial superinfections in COVID-19 patients is still not entirely clear and the incidence seems to be much lower than in severe influenza . We agree with many other authors that establishing clear criteria for initiating antibiotic therapy in COVID-19 patients is essential in order to prevent the consequences of inappropriate prescribing [26, 37, 64]. We must be aware that a potential consequence of the COVID-19 pandemic is the long-term propagation of antimicrobial resistance resulting from increased patient exposure to antimicrobials that are often suboptimally or inappropriately used [72, 73]. This rapid growth in antibiotic prescribing can exercise a strong selective pressure on bacterial pathogens to develop resistance, leading to increased incidence of drug-resistant bacterial infections in the years following the COVID-19 pandemic. It has been calculated that ten million people could die from antibiotic-resistant bacterial infections each year by 2050 .
Recently, a group of members of ESCMID’s Study Group for Antimicrobial Stewardship (ESGAP) published a paper warning against non-critical use of antibiotics in COVID-19 patients along with some practical recommendations. Huttner et al. indicate that we should periodically reevaluate the suitability of our prescription and discontinue it as soon as possible when there is low suspicion of bacterial infection. In the event its continued use is warranted, switch to oral therapy early and give short cycles of five days . It is important to educate healthcare providers in antimicrobial stewardship to prevent the consequences of excessive antimicrobial use such as toxicities, selection for opportunistic pathogens such as Clostridioides difficile (coinfection with SARS-CoV-2 results in a worsening of outcomes) and antimicrobial resistance [74, 75].