Introduction

COVID-19 is caused by SARS-CoV-2 and may result in a plethora of organ-related short- and long-term sequalae.1-5 In severe cases, a prognosis is serious and often unpredictable.6 A notable proportion of critically ill patients with COVID-19 requires prompt hospitalization and respiratory support.7 The clinical course of COVID-19 depends not only on the sole burden of SARS-CoV-2 infection, but also on the risk of secondary infections, including their etiology, severity, susceptibility to chemotherapeutic agents, and many host-related variables.

As many as 1 in 4 patients with COVID-19 presents a concomitant or fungal infection.8 The patients who develop any coinfection or superinfection during COVID-19 have a higher risk of death,9,10 especially those in the intensive care unit (ICU) setting.

In recent months, much attention has been given to devastating consequences of fungal infections among COVID-19 patients. Press reports on strange, potentially mass-spreading and deadly cases have electrified the public opinion11 and shed light on a possible area of new scientific research.

Unfortunately, many available reviews reported data retrieved using unspecific, unsystematic, and incomprehensive strategies. Their conclusions are at times contradictory. This is predominantly seen for fungal secondary infections. Therefore, we decided to collect and analyze all the available data regarding this issue in a comprehensive and systematic manner, according to the PRISMA checklist (Supplementary material, Table S1).12 The PICO (patient / population, intervention, comparison and outcomes) criteria are presented in Table 1.

Table 1. The PICO criteria used for this review

PICO

Description

Participants

Critically ill adult patients with COVID-19

Interventions (exposure)

Fungal infection

Comparisons

Patients without fungal infection (when applicable)

Outcomes

Mortality

Methods

Eligibility criteria

We included studies that focused on critically ill adults diagnosed with COVID-19 (SARS-CoV-2 infection) and sufferring from subsequent fungal infection, defined by investigators as the coinfection, supra- or superinfection, concomitant infection, secondary infection, or mixed infection. All fungal infections except for superficial / skin infections (ie, respiratory system and / or other organs and systems including urinary tract and invasive candidiasis) were investigated. All definitions (diagnostic criteria) of fungal infections used by the authors of primary (included) studies were accepted for the purpose of the review. Mortality was the outcome of interest. We failed to assess secondary outcomes (ie, other than mortality, eg, ICU and hospital length of stay, requirement and length of organ support) due to insufficient reporting of necessary data in available studies. We included papers of which full reports were published before the day of the search (ie, January 8, 2022). We included papers published in English only. Case reports, systematic reviews, and meta-analyses were excluded.

Information sources and search string

The search was conducted within MEDLINE, Web of Science, clinicaltrials.gov, Embase, and Cochrane Library on January 8, 2022, using the following search string:

(((COVID-19 OR SARS-CoV-2 OR coronavirus) AND (mortality OR death OR die* OR survival OR survivor* OR fatality OR prognosis OR outcome*)) AND (candid* OR fung* OR non-candid* OR aspergill*)) NOT (neonat* OR pediatric OR children).

The remaining search strings are presented in Supplementary material, Section S1.

Study selection and data collection process

After importing all the papers from the initial search using the search string, 3 independent investigators (MPP, ZP, and KG) assessed the studies by analyzing their titles and abstracts (via Mendeley). The study was processed further if all 3 adjudicators agreed to include the paper for review. If only 2 reviewers agreed to proceed with the manuscript, the decision was made independently by another investigator (ŁJK).

Data items and quality assessment

The following data were collected: authors, year of publication, type of a study, patients’ characteristics (number of patients, sex ratio, median or mean age, body mass index, organ failure severity, need for and mode of mechanical ventilation), day from admission to the ICU up to the day when a fungal infection was diagnosed, occurrence of any bacterial infections, diagnostic method of the fungal infection confirmation, antifungal treatment, etiology of the fungal infection, any immunomodulatory treatment, and outcomes (mortality as the main outcome and others, if given). Descriptive statistics of quantitative and qualitative data were recorded in an electronic database. Then, we calculated the median and interquartile range (IQR) for available numerical items.

The Newcastle-Ottawa scale (NOS) was implemented to assess the quality of the analytical studies (Supplementary material, Reference S1). The total NOS score of each study was converted to the Agency for Healthcare Research and Quality standards (Supplementary material, Reference S2). Two independent investigators (ZP and KG) performed the quality assessment. Subsequently, any differences were resolved by a discussion, and the final decision was accepted by ŁJK.

Data analysis

The principal summary measure used for the quantitative analysis was an odds ratio (OR) with a 95% CI for mortality in COVID-19 patients with and without fungal infection. For the purpose of statistical analysis, OpenMeta[analyst] (CEBM Brown, Providence, Rhode Island, United States) software and MedCalc Statistical Software version 18.1 (MedCalc Software Ltd., Ostend, Belgium) were used.

A study was included in the meta-analysis if it compared mortality in COVID-19 patients with and without fungal infections. Forest plots were drawn to display the estimated results of the included studies. To measure the effect of fungal infections on mortality, the ORs with their 95% CIs were calculated. Due to the observational nature of the included studies, we used random-effect models for the synthesis of results. In order to measure the heterogeneity of the included studies, I2 was calculated.

We prepared a funnel plot of all studies included in the meta-analysis in order to assess the publication bias. We performed subgroup analyses for an Aspergillus infection group, a Candida infection group, and for studies deemed to be of “good” quality (assessed using NOS).

Results

Study selection

By using the search strategy, we identified 5961 articles in total. After removing duplicates (n = 2410), we screened the remaining papers by evaluating their titles and abstracts (n = 3551). By using the PICO criteria and the inclusion and exclusion criteria, we selected 287 papers for the full-text read assessment (Table 1). After excluding the articles for prespecified reasons, the final 38 papers were included in the systematic review (Table 2 and Supplementary material, Table S2). All of the papers were observational (n = 38), of which 13 were prospective13-15,17,24,34,36,40,41,44,46,48,49 and 25 retrospective studies.16,18,19-25-33,35-37-39,42-45,47 Among them, 8 studies were classified as case series.18,21,22,27,37,40,45,47 The 2 remaining studies were both prospective and retrospective in design.48,50 The study selection process is presented in a flowchart (Figure 1).

Table 2. Summary of the included studies (n = 38)

Author and year

Study type

Fungal patients

Outcome

Quality assessmenta

Bartoletti et al13 (2020)

Prospective cohort

Out of 108 ICU patients, 30 (27.8%) were diagnosed with probable CAPA

30-day mortality: 44% vs 19% (CAPA vs non-CAPA)

9/9 GOOD

White et al14 (2020)

Prospective cohort

Out of 135 critically ill patients with prolonged MV, 36 patients had at least 1 positive mycological test (CAPA, 14%, Candida spp., 12.6%)

30-day mortality: 52%; Aspergillus spp.: 52%; Candida spp.: 47%; nonfungal infection group: 31%

7/9 POOR

Gangneux et al15 (2020)

Prospective cohort

Out of 37 ICU patients, 7 (15.5%) were diagnosed with putative / probable invasive aspergillosis

ICU mortality: 29% vs 13% (Aspergillus group vs non-Aspergillus group)

7/9 POOR

Bishburg et al16 (2020)

Retrospective case-control

Out of 89 ICU patients, 8 (8.9%) were diagnosed with candidemia

In-hospital mortality: 38% vs 54% (candidemia group vs non-candidemia group)

6/9 POOR

Segrelles-Calvo et al17 (2020)

Prospective cohort

Out of 215 ICU patients, 49 (22.8%) were diagnosed with invasive fungal infection, 7 (3.6%) of these patients had an infection caused by Aspergillus spp.

Mortality: 86% vs 37% (Aspergillus group vs non-Aspergillus group)

7/9 POOR

Agrifoglio et al18 (2020)

Case series

Out of 139 critically ill patients, 15 (10.8%) had candidemia, 2 (1.4%) showed Aspergillus fumigatus growth from BAL culture

Mortality for candidemia: 40%

N/A

Sarrazyn et al19 (2020)

Retrospective cohort

Out of 131 ICU patients, 4 had putative IPA

Mortality: 100%

7/9 POOR

Fekkar et al20 (2020)

Retrospective cohort

Out of 145 ICU patients screened for fungal superinfections, 7 (4.9%) were diagnosed with probable (2.8%) and putative (2.1%) invasive mold infection

30-day mortality: 44% vs 27% (invasive mold infection group vs noninvasive mold infection group)

9/9 GOOD

Nasir et al21 (2020)

Case series

Out of 147 ICU patients, in 9 (6.1%) Aspergillus was isolated from tracheal aspirates and 5 of these patients were diagnosed with CAPA (3.4%)

Mortality: 60%

N/A

Alanio et al22 (2020)

Case series

Out of 27 ICU patients, probable IPAs were diagnosed in 1 (4%) and putative IPAs were diagnosed in 8 patients (30%)

Mortality: 44%

N/A

Reizine et al23 (2021)

Retrospective cohort

Out of 49 severe COVID-19 patients, 10 (20.4%) were diagnosed with CAPA (40% probable, 60% possible)

90-day mortality: 70%

7/9 POOR

Macauley et al24 (2021)

Retrospective cohort

Out of 236 ICU patients, 12 (5.1%) were diagnosed with candidemia

Mortality: 75%

6/9 POOR

Silva et al25 (2021)

Retrospective cohort

Out of 212 severe COVID-19 patients, 55 (25.9%) had a fungal culture growth (Candida spp. in 54 cases and Aspergillus spp. in 3 cases)

Mortality: C. non-albicans, 91%; C. albicans, 76%; Aspergillus spp., 68%

8/9 GOOD

Kayaaslan et al26 (2021)

Retrospective cohort

Out of 5542 ICU patients, 105 (1.9%) were diagnosed with candidemia

28-day mortality: 88%

8/9 GOOD

Rabagliati et al27 (2021)

Case series

Out of 146 ICU patients, 16 (11%) were diagnosed with invasive mold infection

Overall mortality: 31%

N/A

Miao et al28 (2021)

Retrospective cohort

Out of 94 severe and critically-ill patients, 17 (18.1%) were diagnosed with coinfection, of which 10 (10.6%) were fungal coinfections

Overall mortality: 30%; Aspergillus spp., 100%; Cryptococcus spp, 100%; Candida spp, 14%

7/9 POOR

Garcia-Vidal et al29 (2021)

Retrospective cohort

Out of 146 ICU patients, 7 (4.8%) had fungal hospital-acquired superinfections

Mortality: Aspergillus fumigatus group, 33%; Candida albicans group, 50%

7/9 POOR

Velez Pintado et al30 (2021)

Retrospective cohort

Out of 83 ICU patients, 16 (19.2%) were diagnosed with CAPA during hospitalization, of which 87.5% met the criteria for a probable case (GM+) and 12.5% met the criteria for a proven case (culture+)

In-hospital mortality: 31% vs 13% (CAPA group vs non-CAPA group)

6/9 POOR

Permpalung et al31 (2021)

Retrospective cohort

Out of 396 ICU patients, 20 (5.1%) were diagnosed with probable CAPA and 19 (4.8%) with putative CAPA

Overall mortality: 56% vs 40% (CAPA group vs non-CAPA group)

7/9 POOR

Nucci et al32 (2021)

Retrospective cohort

Out of 41 ICU patients with candidemia, 9 were diagnosed with candidemia during COVID-19

30-day mortality: 68%

5/9 POOR

Arastehfar et al33 (2021)

Retrospective cohort

Out of 1988 critically ill patients, 7 (0.03%) had candidemia

Mortality: 87%

4/9 POOR

Lahmer et al34 (2021)

Prospective cohort

Out of 32 ICU patients, 11 (34%) were diagnosed with putative CAPA

Mortality: 36% vs 10% (CAPA group vs non-CAPA group)

8/9 GOOD

Delliere et al35 (2021)

Retrospective cohort

Out of 108 ICU patients, 21 (19.4%) developed probable IPA

Mortality: 71% vs 37% (IPA group vs non-IPA group)

6/9 POOR

Roman-Montes et al36 (2021)

Retrospective cohort

Out of 129 ICU patients, 14 (9.7%) were diagnosed with CAPA

30-day mortality: 57% vs 49% (CAPA group vs non-CAPA group)

7/9 POOR

Szabo et al37 (2021)

Case series

Out of 90 critically ill patients, 20 (22.2%) met the criteria for invasive fungal infection (7 cases of candidemia and 16 cases of putative / probable IPA)

In-hospital mortality: 80%

N/A

Ghazanfari et al38 (2021)

Prospective cross-sectional

Out of 105 ICU patients, 40 (38.1%) were identified as probable CAPMI. Twenty-nine (27.6%) had a positive BAL culture in which CAPA was diagnosed in 22 patients (75.9%) and 6 patients (24.1%) were diagnosed with fusariosis.

In-hospital mortality: 92.5% vs 100% (probable CAPMI group vs non-CAPMI group)

8/10 POOR

Prattes et al39 (2021)

Prospective cohort

Out of 592 critically ill patients, 109 (18.4%) were diagnosed with CAPA (histologically proven, [n = 11]; probable, [n = 80]; possible, [n = 18])

90-day mortality: 71% vs 43% (CAPA vs non-CAPA group)

9/9 GOOD

Paramythiotou et al40 (2021)

Case series

Out of 179 ICU patients, 6 (3.3%) were diagnosed with CAPA (4 probable and 2 possible)

Mortality: 66.7%

N/A

Wasylyshyn et al41 (2021)

Retrospective cohort

Out of 256 ICU patients, 3 (1%) were diagnosed with CAPA

12-week CAPA group mortality: 33.3%; 30-day non-CAPA mortality: 24%

6/9 POOR

Iqbal et al42

(2021)

Prospective cohort

Out of 307 ICU patients, 61 (19.9%) were diagnosed with probable CAPA

Mortality: 91.8% vs 53.3% (CAPA group vs non-CAPA group)

6/9 POOR

Rouzé et al43 (2022)

Retrospective cohort

Out of 568 ICU patients, 14 (2.5%) were diagnosed with putative IPA and 17 (3%) with probable IPA

28-day mortality: 35.7%

9/9 GOOD

Al Mutair et al44 (2021)

Retrospective cohort

Out of 1740 ICU patients, 37 (2.1%) were diagnosed with candidiasis

Mortality: 67.6% vs 41.1% (candidiasis group vs non-candidiasis group)

9/9 GOOD

Coşkun et al45 (2021)

Case series

Out of 627 ICU patients, 32 (5.1%) were diagnosed with fungal infection

Mortality: 78.1%

N/A

Ergün et al46 (2021)

Prospective case-control

Out of 219 ICU patients, 39 (17.8%) developed proven or probable CAPA

30-day mortality: 53.8% vs 24.1% (CAPA group vs non-CAPA group)

9/9 GOOD

Sivasubramanian et al47 (2021)

Case-series

Out of 970 ICU patients, 2 (0.2%) were diagnosed with proven CAPA, 9 (0.9%) with probable CAPA, and 37 (3.8%) with possible CAPA

Mortality: 83%

N/A

Gangneux et al48 (2021)

Prospective and retrospective cohort

Out of 509 ICU patients, 128 (25.1%) were identified as proven, probable or possible IFI (76 cases of CAPA, 32 cases of candidemia, 6 cases of mucormycosis, and 1 case of invasive fusariosis)

In-hospital mortality: 55% vs 30% (IFI group vs non-IFI group); 61.8% vs 32.1% (CAPA group vs non-CAPA group); 56.3% vs 30% (Candidemia group vs non-IFI group)

9/9 GOOD

Xu et al49 (2021)

Retrospective cohort

Out of 335 ICU patients, 78 (23.3%) were diagnosed with CAPA

ICU mortality: 52.6% vs 28.4% (CAPA group vs non-CAPA group); 180-day mortality: 65.4% vs 33.5% (CAPA group vs non-CAPA group)

9/9 GOOD

Janssen et al50 (2021)

Prospective and retrospective cohort

Out of 823 ICU patients, 63 (7.6%) were diagnosed with CAPA

ICU mortality: 49.2% vs 29.8% (CAPA vs non-CAPA group)

9/9 GOOD

a Quality assessment performed using the Newcastle-Ottawa Scale

Abbreviations: BAL, broncho-alveolar lavage; CAPA, COVID-19–associated pulmonary aspergillosis; CAPMI, COVID-19–associated pulmonary mold infection; ICU, intensive care unit; IFI, invasive fungal infection; IPA, invasive pulmonary aspergillosis; MV, mechanical ventilation; N/A, not available

Figure 1. Study selection process

Quality assessment

By implementing NOS, we assessed 30 analytical studies (the remaining 8 studies were descriptive in design). Overall, we identified 12 studies of “good” quality13,20,25,26,34,39,43,44,46,48-50 and 18 studies of “poor” quality.14-17,19,23,24,28-33,35,36,38,41,42 The majority of studies lacked the “Comparability” criterion, as the aforementioned 26 studies did not control for any confounding factors (Supplementary material, Table S3).

Study characteristics

Basic study characteristics are shown in Table 2, while detailed data are enclosed in Supplementary material, Table S2.

The total number of patients included in the systematic review was 17 695 and varied from 27 to 5542 participants in individual studies.13-50 The number of patients who developed any fungal infection over the course of COVID-19 disease was 1182 and accounted for 6.7% of all patients. The mean or median age of the patients with fungal infection in 19 studies was above 65 years,15,16,19,21,23,26,27,29,31,33,34,37-41,45-47 and in the remaining 15 studies it was below 65 years.13,14, 17,18,20,22,24,30,32,35,36,42,43,48,49 Information regarding age was unavailable in 4 studies.23,28,44,50 The percentage of men with fungal infection varied from 43% to 89%, with a median (IQR) of 64.5% (54%–75%). Organ failure scores were assessed in 16 studies, of which the SOFA score was the most common (n = 9), and it varied from 3 to 12 points (median, 6). The number of studies that reported the frequency of mechanical ventilation of patients with fungal infections (either invasive or noninvasive) was 27, while the frequency of mechanical ventilation varied from 37% to 100% (median [IQR], 100% [87%–100%]).13,14,17-22,24,26,27,29-34,37-43,47-49 The mean or median ICU length of stay was reported in 18 studies and varied from 7 to 41 days (median [IQR], 21 [13–29] days).13-17,20,21,23,26,31,33,34,37-39,42,43,46

Population characteristics of patients with fungal infections versus controls

Information regarding the characteristics of both groups, that is, exposures (fungal infection) and controls, was provided in 17 studies.13-17,20,30,31,34-36,38,39,41,46,48,49

  • Age (18 studies): For the fungal group, 8 studies reported the mean or median age of above 65 years,15,16,31,34,38,39,41,46 whereas for the remaining 10 studies, the patients were younger than 65 years.13,14,17,20,30,35,36,39,48,49 As for the controls, the mean or median age in 2 studies was above 65 years,16,38 whereas in the remaining 15, it was below 65 years.13-15,17,20,30,31,34-36,39,42,46,48,49
  • Gender (18 studies): For the fungal group, the percentage of men varied from 47.5% to 86% (median [IQR], 70% [56%–79%]). As for the controls, men accounted for 50% to 83% (median [IQR], 70% [60%–78%]).13-16,20,30,31,34-36,38,39,41,46,48,49
  • Body mass index (BMI) (6 studies): For the fungal group, the mean or median BMI varied from 25 to 34 (median [IQR], 28 [26–31]). As for the controls, the mean or median BMI varied from 27 to 30 (median [IQR], 28 [27–29]).13,15,16,31,35,48
  • SOFA score (7 studies): For the fungal group, the mean or median SOFA score varied from 3 to 12 points (median [IQR], 8 [5–10] points). As for the controls, the mean or median SOFA score varied from 3 to 10 points (median [IQR], 7 [4.5–9.5] points).13,15,16,34,35,48,49
  • The ICU length of stay (LOS) (13 studies): For the fungal group, the mean or median ICU LOS varied from 9 to 41 days (median [IQR], 20 [16.5–28.5] days). As for the controls, the mean or median ICU LOS varied from 10 to 25 days (median [IQR], 19 [12–21] days).

Aspergillus spp. group

The number of studies that focused on Aspergillus spp. was 31 (of which 20 focused only on Aspergillus spp.13,15,17,19,21-23,30,31,34-36,40-43,46,47,49,50 and 11 focused on various fungal infections).14,18,20,25,27-29,37,38,45,48 All Aspergillus spp. cases concerned invasive pulmonary aspergillosis (IPA) associated with COVID-19. The total number of patients in these studies was 8059, of which COVID-19–associated pulmonary aspergillosis (CAPA) was diagnosed in 787 (9.8%) patients. Among prospective studies (n = 10), CAPA was diagnosed in 17.7% (median) papers,13-15,17,22,34,38,39,42,46 whereas in retrospective studies (n = 22), this Figure was 7.5% (median).18-21,23,24,25-31,35-37,40,41,43,45,47,49 In prospective studies, the median time from ICU admission to the CAPA diagnosis was 7 days, whereas in retrospective studies, this amounted to 9 days. Information about antibacterial treatment or bacterial culture growth was provided in 14 studies.14,21,25,27,29,30,34,35,37,38,43,45,47,49 In 22 studies, information regarding immunomodulatory treatment was provided, and it included glucocorticosteroids implemented in 7%–100% of patients (median [IQR], 70% [57%–94%]), and tocilizumab in 10%–100% of patients (median [IQR], 30% [17%–72%]).13,14,17,18,20-23,27-31,35-37,39,41,43,46,47,48

Regarding the diagnostic methods for aspergillosis, the following was observed: 5 studies reported compliance with influenza-associated pulmonary aspergillosis (IAPA) criteria;13,14,21,30,38 19 studies with the COVID-19–associated pulmonary aspergillosis European Confederation of Medical Mycology (ECMM) and the International Society for Human and Animal Mycology (ISHAM) criteria;14,23,27,35,38-42,46-48,50 5 studies with the European Organization for Research and Treatment of Cancer / Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) criteria,20,22,35,37,49 and 5 studies with the AspICU criteria14,15,34,36,37 (7 studies implemented >1 criteria).14,22,34,35,37,38,43One study used Blot and Verweij criteria.43 The remaining studies (n = 7) did not provide specific information regarding the abovementioned diagnostic criteria.17-19,21,25,28,31

Full species identification was implemented in 19 studies.13,14,17,18,21-23,27,29,34,37,38,40-43,45,47,49 Among these, A. fumigatus was the most common cause of aspergillosis (range, 5%–100%; median, 50%; IQR 20%–78%). A. flavus and A. niger were diagnosed less frequently (median, 21% and 11%, respectively).

Candida spp. group

The number of studies that focused on Candida spp. was 14 (of which 6 focused only on Candida spp.16,24,26,32,33,44 and the other 8 focused on various fungal infections)14,18,25,28,29,37,45,48. The total number of patients in these studies was 11 678, of which Candida spp. infection was diagnosed in 351 (3%). In a prospective study,16 COVID-19–associated candidiasis (CAC) was diagnosed in 12.6% of the patients, whereas in retrospective studies (n = 12), this Figure was 2.5% (median).16,18,24-26,28,29,32,33,37,44,45,48 The number of days from ICU admission to Candida spp. infection diagnosis was reported in 7 studies, ranging from 9 to 26 days (median [IQR], 17 [9–19.5] days). In the prospective study the median day from ICU admission to CAC was 9 days, whereas in retrospective studies, this amounted to 17 days. Information about antibacterial treatment or bacterial culture growth was provided in 6 studies.14,24-26,29,45 In 7 studies, information regarding immunomodulatory treatment was provided, namely: glucocorticosteroids were implemented in 12%–100% of patients (median [IQR], 50% [45%–100%]),14,16,24,26,28,37,48 while tocilizumab was implemented in 30% of patients.16,37,48

Regarding the diagnostic methods for Candida spp. infection, in all the studies (n = 14) blood samples were collected for Candida spp. cultures.14,16,18,24-26,28,29,32,33,37,44,45,48

Full species identification was performed in 11 studies.14,16,18,24-26,29,32,33,37,45 Among these, C. albicans was the most common cause of Candida spp. infection (range, 15%–71%; median [IQR], 56.% [32%–60%]). C. parapsilosis, C. tropicalis, and C. glabrata were diagnosed less frequently (median, 25%, 23%, and 15%, respectively).

Other fungal infections

The number of studies that provided information regarding fungi other than Candida spp. and Aspergillus spp. was 6.20,27,28,33,38,45 The frequency of those pathogens was low, namely 7 cases of Fusarium spp. infection,20,38,48 6 cases of Mucorales spp. infection,20,48 2 cases of Rhizopus spp.,27 1 case of Scedosporium spp.,27 1 case of Cryptococcus spp.,28 1 case of Trichosporon mucoides,45 and 1 case of Rhodotorula mucilaginosa.33

Mortality: descriptive statistics

The overall mortality in critically-ill COVID-19 patients with fungal infections across papers retrieved for a systematic review (n = 38) varied from 29% to 100% (median [IQR], 56% [40%–77%]).13-50 The mortality among controls (ie, patients without fungal infection) (n = 15) varied from 9.5% to 100% (median [IQR], 30.6% [24%–42%]).13-17,20,30,31,34-36,38,39,41,42,44,46,48-50

Mortality in patients with suspected CAPA varied from 29% to 100% (median [IQR], 54% [36%–70%]). In prospective studies, the median mortality was 52%,13-15,17,22,34,39,42,46 whereas in retrospective studies it reached 57%.18-21,23,25,27-31,35,36,40,41,43,47-50 In analytical and descriptive studies, the pooled mortality was 64.5% (95% CI, 59.9%–69.3%),13-50 and in descriptive studies only it was 68% (95% CI, 55.7%–82.7%).18,21,22,27,37,40,45,47

Mortality in patients with Candida spp. infection varied from 38% to 88% (median [IQR], 68% [47%–87%]). In 1 prospective study, mortality was 47%,14 whereas in retrospective studies it reached 71.5%.16,18,24-26,29,32,33,44

Fungal infections and mortality: meta-analysis

A total of 19 studies were deemed eligible for the quantitative assessment. Critically ill COVID-19 patients who acquired a fungal infection were 2.98 times more likely to die than the controls (OR, 2.987; 95% CI, 2.369–3.767; P <⁠0.001).13-17,20,30,31,34-36,38,39,42,43,46,48-50 The heterogeneity of the included studies was moderate (I2 = 26.63%) (Figure 2).

Figure 2. Forest plot of studies with various fungal infections and the risk of mortality

Abbreviations: Ctrl, total number of patients without fungal infection; Ev, number of deaths; OR, odds ratio; Trt, total number of patients with fungal infection

For the Aspergillus group,14,15,17,20,30,31,34,35,39,42,46,48-50 fungal infection significantly increased mortality (OR, 3.279; 95% CI, 2.692–3.994; P <⁠0.001; I2 = 0%) (Figure 3). Among the studies that investigated Candida infections,13,16,44,48 there was also an effect of infection on mortality (OR, 2.254; 95% CI, 1.322–3.843; I2 = 26.14%) (Figure 4).

Figure 3. Forest plot of studies with Aspergillus spp. infection and the risk of mortality

Abbreviations: see Figure 2

Figure 4. Forest plot of studies with Candida spp. infection and the risk of mortality

Abbreviations: see Figure 2

In a subgroup analysis of studies of “good” quality (n = 9, mostly studies with Aspergillus spp. infection),13,20,34,39,44,46,48-50 the effect was also significant (OR, 3.048; 95% CI, 2.486–3.738; P <⁠0.001; I2 = 0%) (Figure 5). Studies of “poor” quality presented greater heterogeneity but the effect still remained significant (OR, 2.439; 95% CI, 1.324–4.494; P <⁠0.001; I2 = 57%) (Figure 6).

Figure 5. Forest plot of “good” quality studies and the risk of mortality

Abbreviations: see Figure 2

Figure 6. Forest plot of “bad” quality studies and the risk of mortality

Abbreviations: see Figure 2

Publication bias

The publication bias was acceptable: although the funnel plot was symmetrical, there were 3 studies outside the region within which 95% of studies were expected to lie in the absence of bias (Figure 7).13-17,20,30,31,34,35,44,46,48-50

Figure 7. Funnel plot representing a potential publication bias across the included studies

Discussion

Summary of evidence

In this systematic review, we investigated the prevalence and mortality of fungal infections in critically ill COVID-19 patients in a comprehensive manner. Our study covered all types of fungal infections previously investigated in research papers, covering not only infections but also coinfections, super- and suprainfections, secondary infections, and those superimposed on COVID-19.

Aspergillosis (A. fumigatus principally) was the most frequently described fungal infection (approx. 1 in 10 patients), followed by candidiasis (C. albicans mainly) (approx. 1 in 33 patients). Other types of fungi were extremely rare. The infections were more frequently reported in prospective observations and all were nosocomial in their origin. Mortality was very high in both CAPA and CAC cases, and was reported higher in retrospective papers (ie, those with lower recognizability of infections).

A subsequent meta-analysis revealed that fungal infections in critically ill COVID-19 patients increased mortality by approximately 3 times, the finding that was confirmed in studies of “good” quality. Importantly, the majority of studies focused solely on CAPA, therefore, the results are primarily shaped by Aspergillus spp. infections that limit the evidence regarding the overall effect of fungal infection on mortality. Indeed, CAC also had a significant impact on mortality but OR was lower (OR, 2.3).

Data comparison with previous studies is limited. To the best of our knowledge, this is the first such a comprehensive review of research regarding fungal infections in severe COVID-19.

In their excellent review, Arastehfar et al33 (Supplementary material, Reference S3) estimated the prevalence of CAC (both superficial and invasive) to vary from 0.7% to 23.5%, depending on the study population and geographical region, along with other factors. C. albicans was the most prevalent organism (44.1%), followed by C. auris (23.2%), and others. The mortality rate of invasive Candida infections was 46%, and varied depending on the species and the antifungals used to treat the infections (ie, 42% for C. albicans but 100% for C. glabrata).

Apostolopoulou et al (Supplementary material, Reference S4) described the prevalence of CAPA to vary from 3.8% to 35% of all ICU patients with COVID-19. The vast majority of patients (91.8%) had no pre-existing immunocompromising or immunomodulating conditions. A. fumigatus was the most prevalent fungus (82.3%), followed by A. flavus (11.8%), and others. The case-fatality rate was 54.1%, which was close to a value from a recent systematic review (ie, 54%; 95% CI, 45%–62%) [Supplementary material, Reference S5] In the latter analysis, the patients with CAPA had almost 3-fold higher mortality than controls (OR, 2.83; 95% CI, 1.80–4.46; I2 = 0%).

Lansbury et al9 in their systematic review reported only 4 fungal pathogens from 3 studies, that is, C. albicans was isolated from the respiratory tract in 5 patients and the urinary tract of a sixth, A. flavius (2 patients) and A. fumigatus (1 patient) caused CAPA, and C. glabrata (1 patient) was a cause of CAC. Their analysis focused more on bacterial coinfections. In short, COVID-19 patients with any coinfection were more likely to die than controls (OR, 5.82; 95% CI, 3.4–9.9; n = 733, 4 studies). Unfortunately, no subgroup analysis for fungal infections was performed.

Musuuza et al8 reported the prevalence of fungal coinfections at 4% (95% CI, 2%–7%) and superinfections at 8% (4%–13%) across 35 retrieved studies. Both co- and superinfections were more prevalent in ICU patients. Aspergillus spp. was the most frequently reported microorganism among coinfections, while Candida spp. was the most frequent cause of fungal superinfections. In their study, the patients who developed any viral, bacterial, or fungal infection had 3-fold higher odds of dying (OR, 3.31; 95% CI, 1.82–5.99; for coinfection: OR, 2.84, and for superinfection: OR, 3.54), with no differences related to etiology. The authors reported neither detailed characteristics, nor causative subanalyses of the patients with fungal infections.

Finally, fungal coinfections in COVID-19 patients from the ICU were also investigated by Peng et al (Supplementary material, Reference S6). In their subanalysis, including 3 papers covering critically ill patients only (2 of which were case series) and 5 papers with mixed populations, the proportion of fungal coinfections was 0.12 (95% CI, –0.01 to 0.24; P = 0.065; I2 = 93%), while the proportion of mortality in fungal coinfections and COVID-19 was 0.36 (95% CI, 0.09–0.63; P = 0.008; I2 = 96.5%). The overall mortality was only 10.9%.

Till now, no reliable conclusions are available regarding the impact of mucormycosis on mortality in COVID-19. In a recent systematic review, Soni et al (Supplementary material, Reference S7) demonstrated that “black fungus” coinfection may complicate SARS-CoV-2 infection but the quality of data included in that paper was low. In fact, the main goal of that study was to discuss the interplay between the fungal and SARS-CoV-2 infection in order to identify potential biomarkers. The authors advocate that it is necessary to motivate researchers to design follow-up plans for black fungus infection in COVID-19.

In our analysis, immunomodulatory treatment (with steroids and tocilizumab) was frequently applied. This finding is rather obvious taking into account current drug recommendations in COVID-19. Pharmacotherapy aimed at controlling the inflammatory cascade triggered by SARS-CoV-2 infection, in which at least 1 immunosuppressive agent is used, poses an additional risk for the development of invasive fungal infection. In addition, the immune responses caused by SARS-CoV-2 can increase the risk of fungal superinfections in several pathophysiological ways (Supplementary material, Reference S3 and S4), although all of them are rather theoretical and require confirmation in clinical studies. Fungal infections are more common in immunocompromised patients (eg, those with low neutrophil counts, on chemotherapy, pharmacological immunosuppression, with AIDS). However, patients with acute respiratory distress syndrome (ARDS) due to influenza were reported to have a higher risk of IPA, even in the absence of predisposing immunocompromising conditions (Supplementary material, Reference S8). The incidence of opportunistic infections is also increased in COVID-19 patients with predisposing factors (eg, older age, diabetes, requiring mechanical ventilation, renal replacement therapy, parenteral nutrition, etc.) (Supplementary material, Reference S3). Aspergillus spores are typically present in the environment and any mechanical ventilation of vulnerable ARDS patients due to SARS-CoV-2 (high alveolar permeability, surfactant damage, microthrombi, ventilator-associated / induced injury, etc) increases the risk of infection, as the spores may easily enter the lungs and the circulation (Supplementary material, Reference S4). Unfortunately, we still do not know the precise mechanisms explaining why aspergillosis occurs in COVID-19 patients so frequently.

Observations from studies concerning other viral infections (eg, influenza) should be interpreted with caution. Candida species are equipped with virulence factors enabling them to invade when opportunities arise and cause various infections, especially when the immune system is impaired (Supplementary material, Reference S9). They may invade through the skin, the respiratory, digestive, or urinary tract. Invasive candidiasis usually develops in patients who require broad spectrum antibiotics and / or parenteral nutrition via central venous access. Up to now, no pathophysiological link between candidiasis and SARS-CoV-2 infection has been found. Indeed, the studies that focused specifically on CAC did not present high prevalence of candidiasis (median, 2.5%), and reported rather similar prevalence of candidiasis to that prior to the COVID-19 pandemic (Supplementary material, Reference S10 and S11).

Early diagnosis of fungal infections is crucial for the implementation of adequate treatment, which is usually long-lasting. Infections must be distinguished from colonization in all cases and reliable diagnostic methods must be used to minimize the risk of mistreatment. Reliable diagnostics of fungal infections is challenging in the clinical setting in COVID-19 patients (Supplementary material, Reference S12). Work overload, understaffing, and other organizational issues (eg, establishment of temporary hospitals) during the pandemic may confound appropriate microbiological testing and increase the risk of inappropriate collection of specimens. On the other hand, the prognosis of critically ill patients infected not only with SARS-CoV-2 but also with other pathogens remains poor, regardless of etiology of infections. Therefore, screening for CAPA or CAC should cover the majority of critically ill patients, regardless of possible susceptibility to fungal infections. Prompt detection of fungi with ELISA- or PCR-based methods in the material from bronchoalveolar lavage accelerates the diagnosis. An alternative method is to identify serum galactomannan which reduces the risk of aerosol generation but has lower sensitivity, often comparable to specimens taken from the upper respiratory tract. While a culture remains the gold standard, even half of the infections could be missed. The diagnosis should be based on international recommendations and the terminology needs to be unified.

Limitations

The main limitations of our study include relatively small populations examined by individual studies, and a descriptive nature of some of the included papers. Secondly, the quality of the retrieved data was rather low and the reliability of many studies could be biased due to their design. This is particularly evident when comparing data from retrospective and prospective studies: the discrepancy in numbers for the prevalence and mortality was clear. The reasons for the differences are that in the retrospective studies the diagnostics did not focus on fungal colonization or infections, whereas in prospective studies the included patients were presumably less severely ill and consequently survived more frequently. Interestingly, data from the meta-analysis could be interpreted without those uncertainties: heterogeneity was rather low and the publication bias was acceptable. However, as the results of our meta-analysis are largely shaped by Aspergillus spp. studies, the overall effect of fungal infection on mortality must be interpreted with caution. Another possible limitation is the way the investigators defined and diagnosed the fungal infections. Even using strict exclusion criteria and ruling out all possible cases of fungal colonization, this diversity remained noticeable. Only by using unified international guidelines would it be possible to set up COVID-19–specific recommendations in terms of prevention, preemptive treatment, or treatment of CAPA or CAC. Indeed, our analysis shows that now more and more studies implement the unified criteria for CAPA diagnosis (ie, ECMM/ISHAM criteria). Moreover, we did not attempt to investigate risk factors of fungal infections in COVID-19. We also failed to register the study protocol before the beginning of scientific investigation but we believe that the rationale for our study is too important to delay publication of the data in the COVID-19 pandemic.

Conclusions

In critically ill patients, COVID-19–associated pulmonary aspergillosis is common and significantly increases mortality. The prevalence of COVID-19–associated candidiasis appears to be lower but is also linked to mortality. Therefore, clinical awareness and screening are needed, followed by personalized antifungal therapy stewardship, which is strongly recommended in order to improve the patients’ prognosis.