Background

Candida auris is a recently emerging nosocomial pathogen which was initially described in Japan in 2009 and then reported in over 30 countries worldwide afterwards [1, 2]. C. auris is usually resistant to several drugs, such as fluconazole, voriconazole, amphotericin B. However, resistance rate varies between studies. According to the genome sequences, C. auris isolates were divided into four clades that were separated by tens of thousands of SNPs: Clade I (South Asian), Clade II (East Asian), Clade III (South African), Clade IV (South American) [3]. Besides, a potential Clade V was found in Iran recently [4].

C. auris can infect or colonize in humans, especially the low-immunity patients in the intensive care unit. Infection and colonization of C. auris are associated with varied treatment strategies and clinical outcomes, so they should be differentiated. Blood stream infections (BSI) are the most common infections with serious outcomes. Overall mortality of C. auris and that for patients with BSI may be as high as 59 and 68% respectively [3]. Nevertheless, other studies reported different data.

Due to its transmissibility, multidrug resistance and severe outcomes, C. auris is called “superbug fungus”. Due to the low incidence of C. auris, no large-scale epidemiology studies were reported by now. Therefore, a comprehensive study was needed to summarize the global epidemiology of C. auris. In this present study, we performed a systematic review and meta-analysis to estimate the case count, drug resistance and mortality of C. auris. Moreover, factors that may affect the mortality such as BSI, clade and drug resistant patterns of C. auris were also analyzed.

Methods

Search strategies and study selection

This systematic review and meta-analysis was carried out according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We systematically searched Pubmed, Embase and Cochrane databases from inception until October 6, 2019 with the only keyword “Candida auris”. Additional studies were obtained by screening the references of eligible studies. Besides, we also searched the websites of Centers for Disease Control and Prevention (CDC), European Centers for Disease Control and Prevention (ECDC) and Public Health England (PHE). Three objectives of this study were case count, drug resistance and mortality of C. auris. Since CDC has established breakpoints for fluconazole, amphotericin B, caspofungin, micafungin and anidulafungin in C. auris (fluconazole ≥32, amphotericin B ≥ 2, anidulafungin ≥4, caspofungin ≥2 and micafungin ≥4 deemed to be drug-resistant), only these drugs were analyzed in this present study.

Inclusion and exclusion criteria

All study formats met the following criteria were included in the meta-analysis: 1) Studies that reported the information of case count, drug resistance and mortality of C. auris, with no limit regarding the diagnostic test used for detecting C. auris. 2) Studies that provided the case count of patients with C. auris, number of resistant isolates/total number of C. auris isolates, number of deaths/total number of cases; 3) Studies with sample size larger than 5 for meta-analysis of drug resistance and mortality. While studies met the following criteria were excluded from the analysis: 1) Duplicate studies contained the same patients; 2) For meta-analysis of drug resistance of C. auris, studies of which the drug resistance data can’t be reinterpreted according to the CDC breakpoints.

Data extraction

Title and abstract review of all searched articles was completed by two of the authors (Jingjing Chen and Sufei Tian) to identify relevant studies on the clinical report of C. auris. Then full texts of relevant articles were independently reviewed by two of the authors (Xiaoxu Han and Sufei Tian) to determine eligible studies by research objectives. Data in the articles were collected with a standardized form by two of the authors (Jingjing Chen and Xiaoxu Han) independently. Disagreements were discussed by three authors to reach consensus. The following information was extracted: first author’s name, publication year, country, research time, study design, clade, case count, sample type, mortality, drug resistance patterns, methods of drug resistance methods. Drug resistant data were reinterpreted according to the CDC breakpoints.

Quality of the studies included for mortality and drug resistance analysis were assessed by the Agency for Healthcare Research and Quality (AHRQ) checklist (https://www.ncbi.nlm.nih.gov/books/NBK35156/). This 11-item checklist assesses studies in terms of the source of information, inclusion and exclusion criteria, selection of participants, researcher bias, quality assurance, possible confounding variables, handling of missing data, participant response rates, and completeness of data collection. An item would be scored “1” for “YES” and scored “0” for “NO” or “UNCLEAR”. Article quality was classified as follows: low quality = 0–3; moderate quality = 4–7; high quality = 8–11.

Statistical analysis

The pooled estimate and corresponding 95% confidence interval (CI) were calculated with STATA11.0 software. Statistical heterogeneity was evaluated with Q statistic (p < 0.10 indicating statistically significance) and quantified using the I2 index. Due to the heterogeneity among studies, all pooled estimates were performed with random-effects model. Furthermore, we did subgroup analyses for mortality stratified by continents, publication year / research year, clade of C. auris, sample type (BSI and non-BSI) and drug resistance rate (higher than overall estimate and lower than overall estimate). Moreover, meta-regression was performed to assess risk factors associated with mortality, with variables such as bloodstream infection, clade, fluconazole resistance, amphotericin B resistance, continent, and publication year included into the analysis. Sensitivity analysis was also performed by omission of studies. Begg’s and Egger’s tests were used to assess publication bias, with p < 0.05 deemed as statistically significant.

Results

Search and identification of eligible studies

As shown in Figure S1, a total of 577 citations were obtained according to the designed search strategy as described in methods. Among them, 97 eligible articles on the clinical report of C. auris were selected for further evaluation and 67 studies were included in the meta-analysis. Finally, 57, 21 and 19 studies were enrolled in the analysis for case count, drug resistance and mortality of C. auris respectively [1, 4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67].

The publication year of eligible studies ranged from 2009 to 2019. Most studies were observational studies except for two studies which were case-control studies [14, 26]. Detailed characteristics of the eligible articles were summarized in Table S1. The mean quality score of the studies included in the meta-analysis for mortality and drug resistance patterns was 6.2 (range: 4–9), with only one high quality study (Table S2). The main problems of the included articles were lack of information on quality assurance, possible confounding variables, handling of missing data, and completeness of data collection.

Case count and clade of C. auris

A total of 4733 cases of C. auris were reported in 33 countries (aligning in descending order: South Africa, United States of America, India, Spain, United Kingdom, South Korea, Colombia, Pakistan, Kenya, Kuwait, China, Russia, Venezuela, Japan, Panama, Israel, Oman, Germany, Brazil, Saudi Arabia, Singapore, France, Australia, Malaysia, Netherlands, Belgium, Norway, Switzerland, United Arab Emirates, Canada, Iran, Greece and Italy) from six continents. The earliest report was in 2009 in Japan, and the earliest isolate of C. auris traced back to 1996 in South Korea as showed by several screening experiments [16, 68]. Moreover, an epidemic curve which depicted the case count of C. auris by detection year was drawn with studies that contained the detailed information. Notably, this was based on publication data rather than surveillance data. It showed that most cases were detected between 2013 and 2019, peaking in 2016 and decreasing thereafter.

Different clades of C. auris were reported to emerge simultaneously from different continents. Four clades of C. auris have unique geographical characteristics. Clade I was mainly reported in India, Pakistan, Kuwait, Russia, United States, United Kingdom, Germany, Malaysia, Netherlands, Italy, etc.; And Clade II were mainly in Japan and South Korea. Clade III was mainly found in South Africa, United States, United Kingdom and China, whereas Clade IV mainly distributed in Colombia and Venezuela. Clade I and III were the most prevalent clades which have more reported cases and wider geographical distribution. The case count and clade of C. auris stratified by country were shown in Fig. 1.

Fig. 1
figure 1

Global reported cases of C. auris by country (adapted from Robinson projection map). The reported case count of patients with C. auris and clade(s) in different countries were represented in descending order. An epidemic curve showing case count of C. auris by year was also portrayed based on publication data

Full size image

Blood stream infection of C. auris

Infection and colonization of C. auris should be differentiated due to varied clinical significance. However, it is difficult to perform due to unavailable data in the original studies. Therefore, rate of BSI which is the most common and serious infection is analyzed instead. Studies enrolled only the candidaemia patients of C. auris were excluded. As shown in Fig. 2a, the frequency of BSI of C. auris varied between studies [25, 36, 37], with a pooled rate of blood stream infection of 32% (95% CI: 21–42%). However, heterogeneity (p = 0.00, I2 = 98.7%) was observed between studies. Subgroup analysis showed that Clade I and Clade IV of C. auris has a high percentage of BSI compared to Clade II and Clade III (Fig. 2b). It is worth mentioning that Clade II has a low rate of BSI rate with ear discharge as the main specimen type, which is different from the other clades of C. auris [9, 17, 69].

Fig. 2
figure 2

Forest plot on BSI rate of C. auris (a) and subgroup analysis by clade (b). ES: Effect size

Full size image

Drug resistance patterns

Meta-analyses of drug resistance were performed with data obtained according to the breakpoints for C. auris established by CDC. As shown in Fig. 3, the pooled resistance rate for fluconazole and amphotericin B were 91% (95% CI: 88–95%) and 12% (95% CI: 7–17%) respectively. Yet there was significant heterogeneity between studies. Besides, publication bias was observed for meta-analysis of resistance rate for fluconazole (Figure S2), yet trim and fill method did not get good result.

Fig. 3
figure 3

Forest plot on the drug resistance of C. auris to fluconazole (a) and amphotericin B (b)

Full size image

Meta-analyses for the resistance rate to echinocandins could not be performed as resistance for these drugs are rare in C. auris. Descriptive analysis was performed alternatively with frequencies of resistant isolates divided by total isolates. Therefore, resistance rate to caspofungin, micafungin and anidulafungin in C. auris were 12.1% (n/N = 101/838), 0.8% (n/N = 8/927) and 1.1% (n/N = 9/840) respectively. However, almost all isolates resistant to caspofungin were from India, with resistance rate of 23.6% (n/N = 100/424) for Indian isolates and 0.2% (n/N = 1/414) for non-Indian isolates.

Mortality of C. auris

The overall crude mortality of C. auris ranged from 0 to 78%, with a pooled crude mortality of 39% (95% CI: 32–47%, Fig. 4). While the mortality for BSI of C. auris was 45% (95% CI: 39–51%, Figure S3). Negligible publication bias and significant heterogeneity (p < 0.05; I2 = 72%) was observed. Sensitivity analysis indicated that the pooled estimate was quite stable when excluding any of the studies.

Fig. 4
figure 4

Forest plot on the crude mortality of C. auris

Full size image

Then subgroup analyses were performed to assess factors that may influence the mortality of C. auris, such as continent, publication year, clade of C. auris, BSI, and resistance to fluconazole, amphotericin B (Table 1). For the subgroup analysis by clade, as most studies contained patients infected with C. auris of Clade I, so we stratified the studies as Clade I and non-Clade I, which showed no significant difference. Studies with C. auris of Clade II were not included in this analysis as lack of data which may be due to rare death. Notably, mortality of patients with BSI of C. auris (45, 95% CI: 39–51%) was higher than that in non-BSI patients (21, 95% CI: 8–33%). Besides, mortality of C. auris in Europe (20, 95% CI: 4–37%) was lower than that in Asia (44, 95% CI: 38–51%). However, we did not find associations between mortality and resistance to fluconazole, amphotericin B, clade or publication year.

Table 1 Subgroup analyses of pooled mortality
Full size table

Discussion

C. auris is a globally spreading yeast with more than 4733 cases reported by now, covering at least 33 countries from six continents. It showed 91% resistance to fluconazole, 12% resistance to amphotericin B, 12% resistance to caspofungin and were highly sensitive to micafungin and anidulafungin. The pooled crude mortality of C. auris was 39%, while the mortality of BSI was 45%. Subgroup analyses showed that cases of BSI and from Europe were factors that affected the mortality. This study is helpful for the surveillance and clinical management of C. auris.

Although a simple meta-analysis of C. auris was performed previously [70], we comprehensively described the epidemic situation and mortality of C. auris. Referring to the epidemic situation of C. auris, over 4733 cases from 33 countries were reported. However, the actual number of cases was underreported in this study. There may be publication bias and bias based on type of surveillance conducted. First of all, there are countries with C. auris cases but not published in literature, such as Thailand, Chile and Bangladesh, Austria and Costa Rica [52, 71]. Secondly, there is bias based on type of surveillance conducted, as screening for C. auris may not be adequate in some countries. For instance, although many cases were reported in South Africa and Kenya, the other underdeveloped countries in Africa did not report cases of C. auris. Moreover, many patients colonized with C. auris which are difficult to identify may be overlooked [9]. This indicates that more intensive surveillance is needed to better understand its epidemic situation. An epidemic curve was drawn using studies with detection time, which showed a peak in 2016 and a fall thereafter. Whether this was a true reduction in case count or a delay in case report needs further follow-up.

As for the clade of C. auris, Clade I and Clade III are the geographically prevalent clade, whereas Clade II and Clade VI showed local epidemic. Besides, we found that Clade I and Clade VI of C. auris exhibits high BSI rate in comparison with the other clades, which was deemed as severe disease with high mortality. Whether this difference was due to specific genetic features deserves further exploration. Furthermore, as there are genes for mating and meiosis in C. auris, sexual recombination can occur with frequent travelling of people with C. auris. Consequently, the genome of C. auris may become more complicated.

In addition, antifungal resistance patterns were also analyzed. Resistance rate of C. auris to fluconazole, amphotericin B, caspofungin, micafungin and anidulafungin were 91, 12, 12.1, 0.8 and 1.1% respectively. It was surprising that Indian isolates showed a resistance rate of 23.6% for caspofungin, which deserves the attention of the clinicians but also needs further validation. Like the other species in the Metschnikowiaceae family (such as C. haemulonii), antifungal resistance is common in C. auris, limiting the treatment options. Acquired resistance through treatment is another concern which deserves clinicians’ attention and further study [5]. Mutations in ERG11 (Y132F, K143R and F126L) and FKS1 (S639F) play an important role in the drug resistance of fluconazole and echinocandins, which should be detected to guide clinical treatment [72]. Drug resistance to amphotericin B may be inducible and transient, nonetheless the mechanisms are not well understood yet. Moreover, genomic insights and analyses of gene expression showed that genes associated with oligopeptide and ABC transporters, iron transporters, glycophosphatidylinositol-anchored proteins, etcmay be involved in drug resistance of C. auris [73, 74].

The pooled crude mortality of C. auris infection was 39%, with an overall mortality of BSI of 45%. Previous meta-analysis indicated that the mortality of candidemia in Europe was 38% [75]. Moreover, the mortality of C. auris was also compared with other drug-resistant organisms, which spread in similar ways in healthcare centers. A meta-analysis showed that the mortality of patients infected with multidrug-resistant Pseudomonas aeruginosa was 44.6% [76]. Besides, the overall mortality of BSIs of vancomycin resistant Staphylococcus aureus and carbapenem-resistant Klebsiella pneumoniae were 26.8 and 54.3% respectively [77, 78]. This indicates that the mortality of C. auris candidemia was a little higher than the other candidemia and similar to that of some drug-resistant bacterial BSIs.

There was heterogeneity between studies, so we investigated factors that may affect the mortality of C. auris infection, such as clade, BSI, drug resistance, continent and publication year. Results showed that the mortality of BSI of C. auris was higher than that of non-BSI. Besides, the mortality reported in Europe was lower than that in Asia. This indicated that types of infection and continent were factors for significant heterogeneity. In addition, mortality at any time rather than 30-day mortality, clades of C. auris, study designs may be the causes of heterogeneity. Reasons explaining for lower mortality in Europe may be as follows: (1) high percentage of non-BSI [10, 19, 26]; (2) better healthcare systems in developed countries with more intensive surveillance and rational treatment.

Although this study was a comprehensive analysis, some limitations should be noted. Firstly, there was an underestimation in case count of C. auris due to publication bias and bias based on type of surveillance conducted. What’s more, most studies included were observational studies, crude mortality rather than attributable mortality was analyzed. Furthermore, significant heterogeneity was observed between studies, well-designed case-control studies should be carried out to estimate the resistance patterns and mortality of C. auris accurately.

C. auris is an emerging pathogen covering over 33 countries, which may have a decrease in case count after 2016. It showed high resistance to fluconazole, moderate resistance to amphotericin B, and high sensitivity to echinocandins. The crude mortality for BSI of C. auris was 45% which was similar to some drug-resistant bacteria previously reported. In summary, C. auris displayed similar characteristics to some drug resistance organisms. C. auris may not be so scary, yet it should not be underestimated, intensive prevention and control should be taken.