Effect of glucocorticoids on the development of COVID‐19‐associated pulmonary aspergillosis: A meta‐analysis of 21 studies and 5174 patients

COVID‐19‐associated pulmonary aspergillosis (CAPA) remains a high mortality mycotic infection throughout the pandemic, and glucocorticoids (GC) may be its root cause. Our aim was to evaluate the effect of systemic GC treatment on the development of CAPA. We systematically searched the PubMed, Google Scholar, Scopus and Embase databases to collect eligible studies published until 31 December 2022. The pooled outcome of CAPA development was calculated as the log odds ratio (LOR) with 95% confidence intervals (CI) using a random effect model. A total of 21 studies with 5174 patients were included. Of these, 20 studies with 4675 patients consisting of 2565 treated with GC but without other immunomodulators (GC group) and 2110 treated without GC or other immunomodulators (controls) were analysed. The pooled LOR of CAPA development was higher for the GC group than for the controls (0.54; 95% CI: 0.22, 0.86; p < .01). In the subgroups, the pooled LOR was higher for high‐dose GC (0.90; 95% CI: 0.17, 1.62: p = .01) and dexamethasone (0.71; 95% CI: 0.35, 1.07; p < .01) but had no significant difference for low‐dose GC (0.41; 95% CI: −0.07, 0.89; p = .09), and non‐dexamethasone GC (0.21; 95% CI: −0.36, 0.79; p = .47), treated patients versus controls. GC treatment increases the risk of CAPA development, and this risk is particularly associated with the use of high‐dose GC or dexamethasone treatment.

[4][5] Among fungal super infections, invasive pulmonary aspergillosis remained a potential cause of morbidity and mortality in intensive care unit (ICU) patients with severe COVID-19 throughout the pandemic years, and the disease was named COVID-19-associated pulmonary aspergillosis (CAPA). 6,7Recent studies have shown a cumulative incidence of CAPA ranging from 3% to 35%, with a higher incidence in severe COVID-19 patients with acute respiratory distress syndrome (ARDS) requiring ICU care. 7,8][9][10][11] Since COVID-19 is still continuing in several parts of the world, including China and India, and additional waves of the pandemic may follow in the future, it is of paramount clinical importance to understand the root cause of the development of CAPA to improve the management and outcome of the disease.Glucocorticoids (GC) have been consistently used worldwide as a standard treatment for COVID-19 since the publication of the preliminary report of the Randomized Evaluation of COVID-19 Therapy (RECOVERY) trial on 22 June 2020, demonstrating the efficacy of dexamethasone (Dexa) in controlling hyperinflammation and reducing mortality in patients. 12,13However, in addition to reducing inflammation, GC also act as potent immunosuppressants inhibiting innate and adaptive cellular immunity responsible for protective immune responses against fungal and other microbial pathogens. 14,15ese properties of GC raise serious concerns about whether systemic GC treatment of COVID-19 could increase the risk of CAPA development.Although there are reports on the association of systemic GC treatment with CAPA, 10,11,16,17 a meta-analysis with detailed systemic evidence on this issue is still lacking.Therefore, to address this knowledge gap, we conducted a meta-analysis of available studies to synthesize real-life evidence of GC treatment on the development of CAPA.

| Study Protocol and Registration
We carried out this study in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) checklist (www.prisma-state ment.org), and the protocol of the study is registered with the National Institute for Health Research International Prospective Register of Systematic Reviews (PROSPERO) (www.crd.york.ac.uk/prospero) under registration number CRD42022341633.

| Search strategy and selection criteria
We performed a comprehensive advanced literature search of four electronic databases, including PubMed/MEDLINE, Google Scholar, Scopus and Embase, to identify all studies reporting CAPA and GC from inception to 31 December 2022.The search was independently carried out by two authors (NKT and ZH) using the following Boolean combinations of MeSH terms: (("Coronavirus disease-2019" OR "COVID-19" OR "Coronavirus-2" OR "SARS-CoV-2") AND ("Aspergillus Infection" OR "Aspergillosis")) AND ("Steroids" OR "Glucocorticoids" OR "Corticosteroids" OR "Prednisone" OR "Prednisolone" OR "Methylprednisone" OR "Methylprednisolone" OR "Dexamethasone") with appropriate search limits such as full text, article type, English language and search period.In addition, we manually searched the reference list, similar studies and citations of included studies as other or extra sources to identify any additional studies missed in the electronic search.
9][20][21][22][23] We excluded (i) studies published in non-English language or a nonpeer-reviewed journal; (ii) reviews, case reports, opinion articles, meta-analyses, editorials and conference abstracts; (iii) studies reporting CAPA in nonadult (<18 years of age) patients; and (iv) studies with no clear comparative data of GC therapy or CAPA development as an outcome measure.
To select eligible studies, we first removed duplicate records from the total identified studies using EndNote (version 20) and manual checks.The remaining studies with no duplicates were subjected to two-level systems to select eligible studies.In the first level, two authors (NKT and ZH) independently screened the identified studies by titles and abstracts to select relevant full-text articles.In the second level, the same authors independently reviewed the full texts of the relevant studies in light of eligibility criteria to select the eligible studies for inclusion in the meta-analysis.Any disagreement between the two author reviewers was resolved by mutual discussion to reach a consensus.Any unresolved disagreement between the two authors was arbitrated by a third author (AK).
The quality of studies was assessed using the Newcastle-Ottawa scale (NOS) to detect the adequacy of the selection of exposed and nonexposed cohorts, comparability of groups, and adequacy of outcome assessment. 24,25The quality of the studies was independently analysed by two authors (ZH and NKT), and any discrepancy was resolved by mutual discussion or discussion with the third author (AK).

| Data extraction and analysis
Data extraction from the included studies was performed by two authors (NKT and ZH) using a predesigned MS Office Excel worksheet.
Any disparity between the extracted data of the two authors was re-examined, and a consensus was achieved following the discussion.In addition, the extracted data were independently reviewed by an additional author (AK).The following data from the text and supplementary information of each included study were collected and recorded.First author name, publication year, country, study design, setting, sample size, GC treatment, number of patients in the GC-treated and control group, and number of CAPA in the GC and control groups.
Several of the included studies used various other immunomodulators (IM), such as tocilizumab, sarilumab, anakinra and anti-IL1 antibody, along with GC.To evaluate the effect of GC alone on the risk of CAPA development, we excluded IM-treated CAPA and noCAPA patients from such studies.Thus, patients who received standard of care (SOC) and GC but no other IM before CAPA diagnosis were categorized as the GC group, while those who received only SOC but not GC and IM were categorized as the control group.
Based on the dose and type of GC treatment, the included studies were also categorized into low-dose GC and high-dose GC and Dexa and non-Dexa GC subgroups for analysis against their respective control counterparts.For GC dosing, hydrocortisone potency was taken as the standard, and the intervention was classified as high or low dose according to whether the dose was greater or less than or equal to 400 mg of hydrocortisone per day. 26If any cumulative dose due to prolonged GC treatment was higher than the total cumulative dose regimen for 10 days (i.e.>400 mg of HC), it was considered a high dose (i.e.cumulative high dose).The conversion of any mg/kg dose into total daily dose was performed taking the weight of the patient 75 kg as the standard.We also subcategorized the included studies using the ECMM/ISHAM definition for the diagnosis of CAPA.In addition, the LOR is more stable than other effect-size measures and thus produces more reliable results.A LOR ratio greater than zero with a positive 95% CI is considered to be significantly higher, while that having a negative 95% CI is not statistically significant.
Publication bias or asymmetry of the included studies was assessed qualitatively using funnel plots and quantitatively using Egger's regression test.Heterogeneity in the studies was determined using
Among these 21 included studies, 13 studies were from Europe, 6 were from Asia and 2 were from North America.They comprised  S2).The studies of the GC versus control group had no publication bias, as shown by the visual symmetry of the funnel plot and values of Egger's regression test (β1 = .87;p = .10)(Figure S1).However, the studies had heterogeneity (I 2 value: 45.46, p = .01,p = .01).
Out of 21 included studies, 11 used GC as well as other IM treatments.We excluded IM-treated CAPA and noCAPA cases from these 11 studies to obtain patients treated with or without GC alone.
(Table S3).After exclusion of IM-treated cases, one study had no GC group and was excluded.The remaining 10 studies after IM exclusion and 10 studies using only GC but no IM treatment were combined together for subsequent analyses.These 20 studies included in the analysis had a total of 4675 patients consisting of 2565 GC-treated patients and 2110 controls (Figure 2).For subanalyses, we categorized these studies into four subgroups based on GC dose and type used in the treatment of the patients.Any study having no defined GC dose or type was excluded from these subgroups (Table S4).
Out of 20 analysed studies, 16 used the ECMM/ISHAM definition for the diagnosis of CAPA (Table S5).The poled LOR for CAPA development was also significantly higher for the GC group versus controls (0.49; 95% CI: 0.15, 0.83; p < .01) in these studies (Figure S2).

| DISCUSS ION
The present meta-analysis aimed to explore the real-life effect of GC treatment on the risk of developing CAPA.The main findings of this meta-analysis include that systemic GC treatment of COVID-19 patients was associated with an increased risk of CAPA development, and this effect of GC was independent of treatment with other IM.
The risk of CAPA was significantly associated with the use of highdose GC or dexamethasone as the GC regimen, while there was no significant association with low-dose GC or non-Dexa GC.To the best of the authors' knowledge, this is the first meta-analysis in the literature to date reporting the effect of GC treatment on the risk of the development of CAPA.
Our data show that COVID-19 patients are at risk of developing CAPA, even in the absence of treatment with GC.SARS-CoV-2 infection targets the respiratory system and causes significant immune dysregulation involving hyperinflammation (cytokine storm) and depletion of CD4+ and CD8+ T cells and NK cells. 5,14,157][48] These pathological features of COVID-19 patients concur with our results of CAPA in patients with no GC therapy treatment.A recent metaanalysis and several research studies reporting CAPA in immunocompetent patients with no underlying risk factors suggest that severe COVID-19 is by itself a risk factor for the development of CAPA. 49,50nce the publication of the preliminary report of the RECOVERY trial in June 2020, that low-dose dexamethasone treatment (6 mg/day for up to 10 days) dampens cytokine storms and reduces 28-day mortality in severe COVID-19 patients with hypoxaemia. 12,13As a result, GC treatment has been widely used worldwide as a first-line therapy for COVID-19 to reduce hyperinflammation (cytokine storm) and improve the outcome of the disease.However, the RECOVERY trial demonstrated the impact of dexamethasone treatment on COVID-19 overall mortality only but did not study its effect on the risk of developing secondary infections, including CAPA.Many clinical trials after the RECOVERY trial also reported the impact of GC on COVID-19 mortality but with conflicting results, and there are limited data concerning the effect of GC on the risk of CAPA. 51,52In addition to immunomodulation, 53 GC are also potent immunosuppressive agents, and they quantitatively and qualitatively affect cells of the innate and adaptive immune systems, making patients prone to developing secondary infections. 14,15In patients with COVID-19, who are inherently immunocompromised and have defective antifungal immunity, 48 GC treatment may further impair antifungal immunity and thereby increase susceptibility to developing CAPA.
A previous meta-analysis on the outcome of CAPA reported that long-term GC treatment of COVID-19 is a risk factor for CAPA. 75][56] All these adverse effects further support that GC therapy may be associated with an increased risk of CAPA development.Approximately half of our included studies had used other IM, including tocilizumab (anti-IL6 receptor monoclonal antibody), sarilumab (IL6 receptor antagonist), anti-IL1 monoclonal antibody and anakinra (IL1-receptor antagonist), along with GC in the treatment of patients, and these IM may also be a risk factor for the development of CAPA. 57After exclusion of other IMtreated CAPA and noCAPA cases, our analysis showed that the association between GC treatment and CAPA is independent of other IM treatments.
The results of our subanalyses showed that the risk of CAPA development was significantly associated with high-dose GC treatment.
A meta-analysis on comparative clinical characteristics and mortality of CAPA has reported that COVID-19 patients receiving long-term GC treatment may be particularly predisposed to CAPA. 7More recently, RECOVERY-II trial has reported an association of non-COVID pneumonia with the use high-dose dexamethasone in the treatment of COVID-19 patients.The trial although not clarified but development of CAPA in these patients may be the sole cause of non-COVID pneumonia. 58The findings of these studies parallel our observations of increased CAPA risk in high-dose GC-treated patients.
Since dexamethasone and non-dexamethasone GC have different pharmacologic characteristics, 59 we categorized the included studies into these two subgroups to evaluate their effect on the risk of CAPA development.Our results showed that dexa- dexamethasone-treated COVID-19 patients. 60An in vitro study reported that dexamethasone inhibits microbial phagocytic activity of neutrophils, while an equivalent dose of methylprednisolone does not have such adverse effects, indicating a reduced risk of infection with the use of this non-dexamethasone GC. 61 The treatment of COVID-19 patients with methylprednisolone compared to dexamethasone has been shown to have a shorter length of ICU stay and fewer episodes of nosocomial sepsis. 62[65][66] This meta-analysis has certain limitations.First, most of the studies included in the meta-analysis were retrospective observational studies and may have confounding factors.Second, there was inevitable heterogeneity across the included studies in this metaanalysis, including variation in the definition of CAPA. 67Importantly, most studies included in our analysis (16/20) used ECMM/ISHAM consensus criteria for defining CAPA, and in a sub-analysis focusing only on these 16 studies with homogenous case definitions, we observed a similar significant association of GC treatment with the risk of CAPA (p < .01).Finally, owing to the very limited number of studies on low and high doses of dexamethasone and non-dexamethasone GC, we could not directly analyse the effect of low versus high doses of these GC types on the risk of CAPA development.However, despite these limitations, the results of this meta-analysis are strengthened by the rigorous study design, high level of acuity of the patient population with well-defined CAPA and appropriately accounted for statistical analysis, including use of LOR, yielding an accurate estimation of the outcome.
In conclusion, our meta-analysis shows that GC therapy independent of other IM treatments increases the risk of CAPA development, and this risk is significantly associated with the use of high-dose GC or dexamethasone treatment.Further clinical evidence from highquality and large-sized studies with additional clinical variables may be needed to generalize and validate the implications of this metaanalysis for clinical practice.

STATA 17 .
0 software (StataCorp.2019; Stata Statistical Software: Release 17. College Station, TX: StatCorp LLC) was used for statistical analysis.We calculated the pooled log odds ratio (LOR) with a 95% confidence interval (CI) as the outcome measure using a random effect model.The LOR is a natural logarithm transformation of the odds ratio.It is a new tool for statistical pooling of data for meta-analyses with heterogeneity, as in the present study, to obtain a more accurate comparison of effect size across the studies.

F I G U R E 3
methasone treatment was significantly associated with the development of CAPA, whereas the use of non-dexamethasone regimens did not show a significant association.Although there are no comparative data on the effect of these GC types on the development of CAPA, several studies in literature support the superiority of non-dexamethasone regimens including methylprednisolone/prednisolone over dexamethasone.A meta-analysis of three randomized controlled trials reported significantly fewer adverse events, including hyperglycaemia and secondary infections, in non-dexamethasone GC-treated COVID-19 patients than in Forest plot showing pooled LOR of the CAPA outcome between GC-treated patients and controls.A LOR value of >0 with positive 95% CI was considered to be significant.CAPA, COVID-19-associated pulmonary aspergillosis, CI, confidence interval; GC, glucocorticoids, LOR, log odds ratio.F I G U R E 4 Forest plots showing pooled LOR of CAPA development between GC dose and type-based subgroups.The LOR is (A) low for low-dose-GC treated patients versus controls, (B) high for high-dose-GC treated patients versus controls, (C) high for dexamethasone (Dexa) treated patients versus controls, and (D): low for Non-Dexa-GC treated patients versus controls.A LOR value of >0 with positive 95% CI was considered to be significant.CAPA, COVID-19-associated pulmonary aspergillosis; CI, confidence interval; Dexa, dexamethasone; GC, glucocorticoids, LOR, log odds ratio.
Characteristics of included studies.
Abbreviations: CAPA, COVID-19-associated pulmonary aspergillosis.TA B L E 2 CAPA in GC treatment and control groups of included studies.

Table 1 )
. Details of the GC treatment and number of CAPA in the GC and control groups of the included studies are shown in 15 retrospective observation studies (8 single-centre, 6 multicentre, and 1 multicentre and multinational), one prospective observa-

Included Studies/Patient: 21/5174 Studies/Patients with GC and IM Treatment: 11/3380 Studies/Patients with GC but no IM treatment: 10/1794 Exclusion of IM treated CAPA and noCAPA patients from studies Studies/Patients with GC treatment: 11/2905
Flow Chart showing selection of included studies for meta-analysis.CAPA, COVID-19associated pulmonary aspergillosis, GC, glucocorticoids.