The impact of immunosuppressive therapy on secondary infections and antimicrobial use in COVID-19 inpatients: a retrospective cohort study

Our study at a large London tertiary hospital during the first wave of the UK epidemic found that 84.7% of patients with severe COVID-19 received antimicrobial therapy, despite only 13.3% developing a microbiologically confirmed bacterial or fungal infection. The majority of infections were hospital-acquired, secondary infections with Gram-negative organisms. In those who received immunosuppressive therapy with steroids and/or tocilizumab, we found no evidence of an increased incidence of infection; nevertheless, this group did receive on average 5.5 more days of antimicrobial therapy. Half of all antimicrobial consumption was of agents on the UK AWaRe Antibiotic ‘Watch’ or ‘Reserve’ list, illustrating the need for more judicious antimicrobial prescribing in hospitalised patients with COVID-19.

The low incidence of co-infection in our study (4.2%) is in keeping with the results of a recent meta-analysis (5.3%), although our rates of secondary infection were lower (9.3% vs. 18.4%) [12]. This may be due to our more stringent definition of infection, which required both positive culture and supporting clinical or radiological evidence as determined by an infection specialist. These incidences should also be interpreted in the context of widespread antibiotic prescribing early in the pandemic, which may have protected against secondary infection. Distinguishing secondary infection from progression of COVID-19 pneumonitis is challenging in both a clinical and research setting. Studies that rely purely on culture results likely overstate the incidence of infection, as enteric pathogens commonly colonise the respiratory tract of ventilated patients [13] and their presence does not necessarily indicate infection in the absence of clinical or radiographic features of pneumonia [14]. By interpreting culture results in the context of contemporaneous clinical and radiological findings, we sought to distinguish true infections from culture of commensal organisms.

Gram-negative respiratory and bloodstream infections were most common, with the most frequent pathogens being Klebsiella spp., Escherichia coli, and Pseudomonas aeruginosa. This is broadly consistent with other studies, although we had notably fewer infections with Staphylococcus aureus [4, 12, 15]. In the context of a rapid expansion of bed capacity, there was an outbreak of Klebsiella pneumoniae on our ICU during the study period, which may account for the high prevalence of this pathogen [16].

The high rates of antimicrobial prescribing in our cohort (84.7%) are consistent with practice elsewhere during the first pandemic wave [5,6,7, 17,18,19]. Doxycycline accounted for the greatest consumption, in line with our local prescribing policy at the time. Despite this, half of all consumption included ‘watch’ and ‘reserve’ antibiotics and prescribing was largely empiric and continued in the absence of confirmed infection. Consumption by antimicrobial class has varied between settings and studies. The UK ISARIC study found that co-amoxiclav was the most frequently prescribed antimicrobial, although consumption metrics were not reported [4]. A large US study found ceftriaxone and azithromycin to be the most commonly used agents [20]. Policies in ten African countries recommended a variety of broad-spectrum antibiotics (azithromycin, ceftriaxone, co-amoxiclav) in initial national COVID-19 guidelines [21]. The upsurge of antimicrobial use in hospitalised patients during the COVID-19 pandemic is likely to have had a detrimental impact on antimicrobial resistance, although quantification of this is challenging [22]. In response to evolving evidence for the low incidence of bacterial co-infection in COVID-19, our local antimicrobial policy now only recommends antibiotics where there is high clinical suspicion of bacterial pneumonia (for example where there is purulent sputum, focal consolidation or neutrophilia prior to steroid administration).

We found a significantly higher consumption of antimicrobials in those who received immunosuppressive therapy, despite finding no increase in the incidence of secondary infection in this group. Concern that immunosuppressive therapies may increase susceptibility to infections might, in part, have driven this prescribing behaviour. The large, pragmatic, multi-platform RECOVERY COVID-19 treatment trial found that immunomodulation with dexamethasone [23] and tocilizumab [24] improved survival in severe COVID-19, transforming subsequent clinical management and outcomes. However, incidence of non-fatal secondary infection was not reported in this trial, nor in another large randomised controlled trial (RCT) of hydrocortisone and tocilizumab [25, 26]. Our study period was prior to the adoption of these therapies as standard of care, allowing a comparison between patients exposed and unexposed to immunosuppressive therapy. We found no evidence of an increased incidence of infection in patients who had received steroids or tocilizumab. Strengths of our study included a stringent and granular approach to the definition of infection and use of time-dependent modelling to account for timing of immunosuppressive therapy relative to infection. Nonetheless, the number of patients exposed to immunomodulatory therapy in our cohort was relatively small and we were unable to account for the impact of antimicrobial use on detection of microbially-proven infection.

Our findings do echo the results of smaller RCTs which collected data on secondary infections and found no increase in incidence in the steroid treatment arms [27,28,29]. Other retrospective, observational studies, however, have conflicting results, with some suggesting no difference [30] and others suggesting higher infection risk in those receiving steroids [31, 32]. The largest of these studies, which used propensity matching from a cohort of over 4,000 ICU patients, found significantly higher ICU-acquired infections (71% versus 52%, p = 0.001) in those who received steroids but not those who received tocilizumab [32]. However, the use of steroids varied significantly between centres, the definition of infection was not standardised and their findings on mortality contradicted those of a major RCT [23]. A WHO meta-analysis of RCTs evaluating IL-6 inhibitors in COVID-19 found no evidence of increased risk of infection (OR 0.99 [95%CI, 0.85–1.16]) [33]. However, another meta-analysis, albeit with less robust inclusion criteria, suggested tocilizumab may increase the risk of fungal infection [34]. Overall, the number of high-quality studies, particularly RCTs, that considered secondary infection as an outcome is small and more robust data are needed.

Our study has several limitations. As a retrospective study, we were reliant on documentation from clinical notes, which may have been incomplete. Our study cohort was single-centre and from the first wave of the pandemic, prior to widespread immunity through vaccination or previous SARS-CoV-2 infection. As use of steroids and tocilizumab was not yet standard-of-care, the indications for these were varied. Of the 17 patients who received tocilizumab, only 4 (24%) received this as part of a randomised trial; the remaining 13 (76%) were part of a compassionate use programme selecting patients with evidence of hyperinflammation [35]. Of those who received high-dose steroids (n = 54), only 15 (28%) received these through participation in the RECOVERY trial. Samples were sent for culture according to the clinical team’s standard practice: only 16% of our cohort had respiratory cultures which represents under-sampling.