Bacterial co-infection is unusual in noncritical COVID-19 cases, so we must reduce our antibiotic prescribing

‘Antibiotics don’t work for viral infections.’

We all know this important message to limit unnecessary prescribing for sore throats, coughs and colds to reduce risk of antimicrobial resistance (AMR). But this message has not played out in relation to COVID-19 — rates of antibiotic prescribing against the disease have been high. At the beginning of the pandemic, for example, reports from China suggested prescribing of antibiotics for those hospitalised with severe SARS-CoV-2/COVID-19 infection was almost universal; and as the virus spread, high rates of antibiotic prescribing continued in Europe and the Americas​[1]​. By April 2020, in Scotland, where I practise, a survey of antibiotic use in hospitalised patients with suspected COVID-19 suggested that 29% received an antibiotic prior to admission, while 62% received empirical antibiotic therapy following admission​[2]​.

So, what drives these high rates of prescribing in this viral infection? There may be lessons learned from history. In the last great viral pandemic — the 1918−1919 influenza pandemic, before antibiotics — case fatality was 2−3 per 100 people, not dissimilar to COVID-19; and post-mortem studies showed secondary bacterial pneumonia was likely the major cause of death​[3]​. There is little doubt that if antibiotics had been available in 1918, hundreds of thousands of lives could have been saved.

Bacterial co-infection remains an important consideration in severe seasonal influenza; however, it was not a significant factor with either of the other novel coronavirus outbreaks — the severe acute respiratory syndrome (SARS) epidemic in 2003 or Middle East respiratory syndrome (MERS) first recognised in 2012 — and we must be mindful of this.

In the current pandemic, the severity and progression of the disease in the second and third week of illness may have been interpreted to have resulted from bacterial superinfection (occurring on top of viral infection) with bacteria. This assumption is compounded by ongoing clinical uncertainty, limited diagnostics and, importantly, in the first wave of the pandemic, lack of other proven therapeutic options. Besides that, antibiotics are easily available, and their potential harms may not be considered by our prescribing COVID-weary workforce.

Data on antibiotic prescribing data for COVID-19 during the second wave are not yet available, but, anecdotally, rates have been high. This has attracted little attention in the public consciousness, but the longer-term effects on AMR must not be underestimated. COVID-19 has significantly challenged the robustness of our antimicrobial stewardship programmes, but, when it comes to antibiotics for respiratory infections, we must stay vigilant.

Diagnostic challenges

It can be difficult to differentiate COVID-19 pneumonia from bacterial pneumonia, owing to overlapping clinical features — notably cough, fever, hypoxia and chest X-ray changes. As in bacterial pneumonia, blood concentration of C-reactive protein (CRP) is usually raised in COVID-19 and typically increases significantly with severity, reflecting a viral-mediated inflammatory response.

Although a moderate increase in CRP may be seen in other viral infections (and sometimes significantly high in influenza), a significantly raised CRP is frequently used by clinicians to differentiate between viral and bacterial infection. Before the onset of the pandemic, CRP point-of-care testing in primary care has been a useful antibiotic stewardship tool supporting reductions in antibiotic prescribing in lower respiratory tract infection when CRP is low​[4]​.

Little evidence of bacterial co-infection

There are little or no published data on bacterial infection complicating SARS-CoV-2 in non-hospitalised patients, but many will receive empirical antibiotic therapy​[5]​. In those who are hospitalised (the majority of whom have community onset SARS-CoV-2 infection), evidence to date points to low probability of bacterial co-infection: multiple cohort studies and systematic reviews have estimated bacterial co-infection to be infrequent (<4%) in hospitalised patients​[6]​. In one well-investigated cohort from the Netherlands, bacterial co-infection was observed in only 1% of patients during the first week of hospitalisation​[7]​.

It is reasonable to postulate that lack of proven microbiology may be a consequence of lack of respiratory tract sampling, owing to infection control concerns around obtaining sputum specimens. However, in the vast majority of patients with COVID-19, the cough is either dry or is associated with only white or non-purulent sputum, and so is unlikely to yield significant pathogenic bacteria. Lack of sputum purulence in the context of cough and/or fever is a clear deterrent to an antibiotic prescription whether in primary or secondary care.

In hospitalised patients with severe COVID-19 pneumonia who progress to requiring mechanical ventilation infection, risks inherent to critical care are amplified by demanding infection control barrier precautions, prolonged intubation (and the challenges of prone ventilation) and the need for haemo-filtration. Microbial investigations are more comprehensive in the critical care setting and are crucial to support targeted prescribing for ventilator-associated pneumonia and catheter-related bloodstream infections.

Little evidence for benefits of antibiotics

The antibiotics azithromycin and doxycycline have been postulated to have anti-inflammatory or antiviral properties, and both have been investigated in COVID-like illness in primary care through the UK PRINCIPLE study.

So far, full data have been published for azithromycin showing no improvement in time to 28-day recovery​[8]​; similar data have been reported regarding doxycycline, although have not yet been fully published. In hospitalised patients with COVID-19 pneumonia, azithromycin was not associated with benefits in either the Brazilian Coalition II randomised controlled study​[9]​, nor in the UK randomised adaptive platform RECOVERY study​[10]​. And on 28 January 2021, the Department of Health and Social Care advised that these antibiotics should not be used to manage confirmed or suspected COVID-19.

When should antibiotics be prescribed?

Infections in primary care

If COVID-19 is suspected in primary care, routine antibiotic use should be strongly discouraged. Antibiotics should be reserved for those where there are specific clinical features indicating another bacterial infection (for example, symptomatic urinary tract infection). Respiratory tract bacterial co-infection is unusual, so antibiotics should be restricted to those with purulent sputum in the context of an infective exacerbation of chronic obstructive pulmonary disease (IECOPD), or if there is strong suspicion of community-acquired bacterial pneumonia (CAP)​[11,12]​. In contrast to pre-pandemic advice, in primary care, CRP should not be used as a guide to initiate antibiotics in the context of COVID-19, although its utility will likely return as COVID-19 infection becomes less prevalent. Amoxicillin or doxycycline are preferred for IECOPD and local guidelines for CAP should be followed. Duration of treatment should be limited to five days (in view of additional benefit after this time, and the impact on AMR of longer courses​[11,12]​).

Patients hospitalised with COVID-19

It is also unlikely that bacterial co-infection will be a factor in the severity of COVID-19 in hospitalised patients, so, again antibiotics are not routinely required; and CRP should not be used as a guide to initiate antibiotics. In those patients who are commenced on antibiotics, baseline investigations including sputum and blood culture should be performed and the need for a previously prescribed antibiotic should be critically reviewed, particularly following a positive SARS-CoV-2 PCR result.

It is important to move away from simply completing the ‘antibiotic course’ and to be proactive in stopping antibiotics which are not required. When CRP has been performed, a lack of reduction at 48 hours most likely reflects SARS-CoV-2 infection rather than ‘antibiotic failure’ and should prompt consideration of discontinuation rather than escalation of antibiotics​[13]​. In those commenced on antibiotics, a procalcitonin (PCT) <0.25 micrograms is also suggestive of non-bacterial infection and should encourage antibiotic discontinuation​[14]​.

Patients requiring mechanical ventilation

The risk of bacterial superinfection (notably ventilator-associated and vascular device-related nosocomial infection) is significant and relates to length of time in critical care. Rapid microbiological diagnostics and empirical therapy based on local epidemiology and antibiograms are essential. In the critical care setting, PCT may assist in guiding discontinuation of empirical therapy, although the significance of a persistently raised PCT in severe COVID-19 is yet to be defined.

Thoughts for the future

Our understanding of COVID-19’s pathophysiology and therapeutics keeps growing. To date, the greatest therapeutic benefits have been seen with drugs that modulate the inflammatory response to the viral infection. Both corticosteroids and IL-6 inhibitors have been shown to reduce COVID-19 mortality; however, they have the potential to increase risk of secondary bacterial (and potentially fungal) infection, and future studies will help to properly define this risk. Meanwhile, vaccination is our best tool for reducing the risk of severe infection and hospitalisation with SARS-CoV-2. Reducing this risk also has the important effect on reducing demand for antibiotics — and we must not underestimate effective vaccine programmes as our cornerstones of antimicrobial stewardship.

R Andrew Seaton, consultant in infectious diseases and antimicrobial management team lead, Queen Elizabeth University Hospital, Glasgow; chair, Scottish Antimicrobial Prescribing Group, Healthcare Improvement Scotland

• This article was amended on 5 May 2021 to correct citations 13 and 14