Introduction

Increasing incidence of candidemia and in particular of nosocomial bloodstream infections has been observed over the recent years [1,2,3]. Candidemia is a life-threatening condition associated with a high attributable mortality [4, 5]. Since the outcome depends on early and targeted treatment (“hit hard and early”), prompt diagnosis is essential [4, 6].

Cultivation from blood samples remains the reference method for the detection of candidemia [7, 8]. Compared to bacterial pathogens, blood culture (BC) for Candida spp. typically suffer from a lower sensitivity (approx. 50%) and longer time to positivity (TTP) [9]. Serologic tests for invasive fungal diseases allow for a short turnaround time. Hence, testing for biomarkers from blood specimens is recommended when candidemia is suspected [7]. Cell wall constituents are common target structures for antigen detection assays. The widely used Platelia Candida serology (Bio-Rad Laboratories, Marnes-la-Coquette, France) is based on two tests that employ cell wall mannan glycoproteins [10]: The Candida antigen Plus ELISA is based on the monoclonal anti-mannan antibody EBCA-1 to detect the mannan glycoproteins in serum, and the Candida antibody Plus ELISA detects human antibodies that bind to C. albicans cell wall mannans. The combination of these two tests for the detection of Candida antigen (Ag) and antibodies (Ab) has been extensively studied and is the basis for recommendations in current guidelines [7, 8]. While mannans are specific for Candida, β-1,3-d-glucan (BDG) is a general constituent of the cell wall of ascomycete fungi. BDG can be detected in the blood of patients suffering from infections such as invasive candidiasis, invasive aspergillosis, or P. jirovecii pneumonia. Although several BDG assays are commercially available, only the FDA-approved and CE-marked Fungitell BDG assay (Associates of Cape Cod, East Falmouth, MA) is commonly used in European and North American laboratories. Consequently, most current studies and guidelines rely on data obtained with the Fungitell BDG assay [7, 8]. However, the usage of this test is confined by methodological and economic challenges [11].

Studies analyzing the performance of serologic tests other than the Platelia and the Fungitell systems are scarce. No published data are available for the Hemkit Candida IHA (Ravo Diagnostika) and only two studies with small patient cohorts (21 and 14 cases of invasive candidiasis, respectively) evaluated the Serion ELISA antigen Candida kit (Institut Virion\Serion) [12, 13]. For several years, the Wako BDG assay (FUJIFILM Wako Pure Chemical Corporation) was only available in Japan. Very recently, this test has been CE marked and is now available in the Western Hemisphere. To better characterize these three serologic tests, we retrospectively evaluated and compared the Hemkit Candida Ab test, the Serion Ag ELISA, and the Wako BDG assay in a clinical cohort comprising 120 episodes of candidemia.

Methods

Study population

This study took place at the Max von Pettenkofer-Institute for Hygiene and Medical Microbiology that hosts the central microbiology laboratory for the University Hospital of Ludwig-Maximilians-University (LMU) Munich, a 2000 bed university medical center in Munich, Germany. Between 2010 and 2017, we identified 120 cases of culture-proven candidemia with corresponding serum samples, which were obtained up to 6 days before sampling of the positive BC. The majority of sera dates from the same day as BC (hereafter referred to as “day 0”) (Table 3). We included four patients with two episodes of candidemia representing 7% of all cases. These cases were considered to represent a new episode, since they were characterized by either a temporal distance of ≥ 3 weeks (n = 2) or by different species causing the infection (n = 2). For the Candida culture-negative cohort (n = 44), we included all patients from the period of 2015 to 2017 meeting the following criteria: (1) culture-confirmed bacteremia caused by Staphylococcus aureus or Escherichia coli; (2) availability of corresponding sera sampled up to 6 days before positive BC. The subgroup of critically ill patients was defined by stay in an intensive care unit. The subgroup of transplant recipients included known recipients of solid organ or bone marrow transplants.

Cultivation and serology

Candida was cultured using the BD BacTec blood culture system (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). For the detection of anti-Candida-Ab an indirect hemagglutination assay (IHA) was performed at a cutoff titer of 1:320 (Hemkit Candida IHA, Ravo Diagnostika, Freiburg, Germany). Candida Ag was analysed using the Serion ELISA antigen Candida-kit at a cutoff of 2.6 U/ml (Institut Virion\Serion, Würzburg, Germany). BDG measurement was conducted using the Wako BDG assay at a cutoff concentration of 11 pg/ml (FUJIFILM Wako Chemicals Europe, Neuss, Germany). All assays were performed according to the manufacturers’ instructions.

Statistical analysis and ethics approval

Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA). The McNemar’s exact test was used for comparison of sensitivities and specificities. Statistical significance was assumed based on an α-level of 0.05. This retrospective study was reviewed and approved by the ethics committee of our university hospital (Ethikkommission der Medizinischen Fakultät der LMU München), and a waiver of informed consent was granted. Sample processing and data analysis were performed anonymously.

Results

Estimation and comparison of test sensitivities

The Ab assay yielded positive results in 48 of 120 cases (40%) (Table 1). The measurements ranged from < 1:80 to > 1:20,480. Using the Ag ELISA, 36 cases (30%) were tested positive. According to the manufacturer’s instructions, the Ag ELISA can yield results between defined positivity and negativity. This indeterminate result was reported in five cases (4%) and subsequently analyzed as being negative. Only 19 cases (16%) were tested positive by both the Ab and the Ag assays. Current guidelines recommend the combined use of the Ab and Ag assays [7, 8]. In this study, this combination increased the sensitivity to 54% (65 of 120 cases) (Table 1). The BDG assay demonstrated superior sensitivity compared to the other tests, identifying 80 of 120 (67%) candidemia cases (p < 0.001 for BDG versus Ag and for BDG versus Ab, p = 0.04 for BDG versus Ag/Ab). Height and range of measurement results of the different tests are depicted in Fig. 1.

Table 1 Comparison of sensitivities of Ag/Ab- and BDG-based serology
Full size table
Fig. 1
figure 1

Box plot (with median and interquartile ranges) of results of a Ag (in U/ml), b anti-Candida-Ab (in titres of 1: dilution factor) and c BDG (in pg/ml) assays. Dotted lines indicate the cutoff (11 pg/ml). Results beneath the limit of detection are plotted (not to scale) in the shaded area

Full size image

Very recently, Friedrich et al. proposed a cutoff of 3.8 pg/ml for the Wako BDG assay [14]. Applying this cutoff in the present study increased the sensitivity to 81%, but decreased the specificity from 93 to 86%. To evaluate the potential for improved diagnostic accuracy, a receiver operating characteristic (ROC) curve analysis was performed (Fig. 2). The highest Youden’s index was found to be 0.71 for a BDG cutoff of 3.95, which is very similar to the cutoff proposed by Friedrich et al. However, according to the manufacturer, the assay’s technical limit of detection is 6 pg/ml. By reducing the cutoff from 11 to 7 pg/ml, the sensitivity of BDG testing increased to 73% without any loss of specificity. A sensitivity of 53% and a specificity of 100% were found when a cutoff of 0.57 U/ml was used for the Ag ELISA (Youden’s index of 0.57). The combined Ag/Ab testing with the optimized ELISA cutoff identified 83 of 120 cases of candidemia (69%).

Fig. 2
figure 2

Receiver operating characteristic (ROC) analysis of aCandida Ag and b BDG for discrimination between candidemia and control patients. Results for the area under the curve (AUC) are specified including the confidence interval (in brackets)

Full size image

Analysis of host and pathogen factors that influence test performance

While sensitivities were not different in the subgroup of critically ill patients, we found a significant decrease when Ag/Ab serology was applied in transplant recipients (40% versus 62% in non-transplanted patients, p < 0.05). Interestingly, this cannot be solely attributed to a deficiency in Ab production, but also to a reduced performance of the Ag assay (Table 1). For the individual Ab and Ag detection systems, the respective sensitivities dropped from 40 to 26% and from 30 to 16% in the cohort of transplant recipients. In contrast, the performance of the BDG-based serology was not affected in this subgroup with a sensitivity of 67%. Surprisingly, sensitivities of all tests were lower in the subgroup of female patients ≤ 65 years. Particularly, the Ag ELISA performed poorly, identifying only 9% of cases of candidemia (versus 41% in the residual cases, p < 0.002).

In 58 of the 120 cases, only one BC was obtained (Table 2). A set of two BCs sampled on day 0 was available in 38 cases and a set of three BCs sampled on day 0 in 24 cases (Table 2). Only in the minority of these cases with multiple BCs, all BCs yielded growth of Candida (29%). The most frequently isolated species was C. albicans (48%), followed by C. glabrata (23%), C. parapsilosis (9%), C. tropicalis (8%), C. krusei (6%) and C. guilliermondii (3%). Furthermore, single episodes of candidemia were caused by C. fabianii, C. kefyr, C. lusitaniae and C. pelliculosa. Ag/Ab testing performed significantly better in blood stream infections caused by C. albicans than in cases caused by other species (67% versus 42%, p < 0.01). Strikingly, Ag/Ab-based serology failed to detect any candidemia caused by C. parapsilosis. The different species did not significantly influence the performance of the BDG assay. Particularly in C. non-albicans blood stream infections, BDG measurement identified significantly more infections than Ag/Ab testing (69% vs. 42%, p < 0.01).

Table 2 Characteristics of blood cultures sampled at day 0
Full size table

The present study includes sera that were collected up to 6 days before the corresponding BC. The specimens were divided into three subgroups depending on the date of sampling: sera obtained simultaneously with the positive BC (= day 0; 43% of all cases), samples obtained up to 3 days before positive BC (43%) and samples even further predating the positive BC (14%) (Table 3). Notably, we observed no significant differences in sensitivity in these different groups. With respect to this finding, we aimed to evaluate the ability of serologic testing to provide earlier evidence for candidemia compared to BC. We identified 50 cases which met the subsequent criteria. (1) Besides the positive BC, additional BC(s) had been collected, which were negative for Candida. (2) The serum was sampled simultaneously with or prior to the negative BC(s). 50% and 58% of these bloodstream infections were tested positive by Ag/Ab and by BDG analysis, respectively.

Table 3 Comparison of Ag/Ab and BDG-based serology with a focus on the course of infection
Full size table

Estimation of test specificities

To establish a control group, we identified 44 patients suffering from bacteremia with either S. aureus (19 cases) or E. coli (25 cases) (Table 4). None of the samples were tested positive by the Ag ELISA, but three specimens were tested positive for anti-Candida-Ab. Three different samples yielded positive results in the BDG assay. Thus, specificity for both Ag/Ab- and BDG-based serology was 93%.

Table 4 Comparison of specificities of Ag/Ab- and BDG-based serology in episodes of bacteremia
Full size table

Discussion

To our knowledge and similar to the recently published study of Friedrich et al., this study is based on the largest cohort of patients who meet the EORTC/MSG criteria for a proven invasive Candida infection [14, 15]. The species distribution is in good agreement with previously published surveillance data [16]. Current recommendations and guidelines primarily rely on investigations of the more widespread Fungitell BDG test and the Platelia system [7, 8]. Reviewing 14 studies, which included between 7 and 105 cases of invasive candidiasis, Mikulska and colleagues report a sensitivity of 83% (95% confidence interval, 79–87%) and a specificity of 86% (95% confidence interval, 82–90%) for the Platelia ELISAs [17]. In this study, the combination of the Serion ELISA and the Ravo IHA demonstrated a specificity of 93% and a sensitivity of only 54%. Sensitivity increased to 67% in cases of candidemia caused by C. albicans, but decreased to 42% in infections caused by other Candida species.

This finding could be explained by the nature of the Ag/Ab assays. The ELISA employs polyclonal antibodies directed against C. albicans cell wall Ag and the IHA detects anti-Candida-Ab via agglutination with sheep erythrocytes coated with C. albicans serotype A cell wall components. However, previous data also suggest a hampered performance of the Platelia kits in the setting of C. non-albicans and particularly in C. parapsilosis infections [18,19,20]. Applying the Hemkit Candida IHA and the Serion antigen ELISA, we also identified a striking loss of performance in this subgroup (sensitivity: 0/11). This is of particular interest since this species accounts for an increasing number of Candida infections [21]. Importantly, BDG measurement was not impaired in the setting of C. non-albicans candidemia.

Very recently, Friedrich et al. reported a low sensitivity of only 43% and a specificity of 98% in the setting of candidemia for the Wako BDG test [14]. In marked contrast, our study demonstrated a much higher sensitivity of 67% and a specificity of 93% in a similar-sized patient cohort. Based on the present data, we cannot explain the discrepant results obtained in the two different studies. To improve the sensitivity of the Wako BDG test, Friedrich and colleagues suggested decreasing the cutoff from 11 to 3.8 pg/ml. This increased the sensitivity to 71% [14]. Similarly, when applying a decreased cutoff of 7 pg/ml in our study, we obtained a sensitivity of 73%. Importantly, in both studies, specificities of the Wako BDG test remained as high as 92% (Friedrich et al.) and 93% (our study), even after applying the modified cutoffs. A high specificity is crucial for screening purposes, such as BDG testing twice a week as recommended in the ESCMID (European Society for Clinical Microbiology and Infectious Diseases) guideline for diagnosis and management of Candida disease [7]. The most striking effect upon optimization of the cutoff was observed for the Ag ELISA. The sensitivity increased from 30% (2.6 U/ml) to 53% (0.57 U/ml) without loss of specificity. Thereby, combined Ag/Ab testing, as recommended by current guidelines, even has a sensitivity of 69% [7, 8].

Interestingly, previous meta-analyses evaluating the performance of BDG testing for the diagnosis of invasive fungal infections reported sensitivities of 75–80% for BDG testing. However, the increased sensitivities came at a cost of specificity, which was found to range from 82 to 85% [22,23,24]. The different results reported by the meta-analyses and the present studies by Friedrich et al. and ours are most likely linked to the different test systems evaluated and the applied cutoff values defined by the manufacturers, as discussed above. While most of the earlier work was evaluating the Fungitell assay, the recent work of Friedrich et al., and our work evaluated the Wako BDG test. The results of our study indicate that the BDG assay can be a valuable tool for the clinician to support the diagnosis of candidemia. However, the specificity of 93% limits its use for screening purposes. To improve the positive predictive value, the test could be restricted to patients with a reasonable likelihood for candidemia. Clinical scores such as the Candida score for critically ill patients could help to identify the patients at risk for candidemia [25, 26].

Our data reveal an unexpected influence of host factors on the detection of antigenemia. The sensitivity of the Ag ELISA was nearly reduced to half when the test was applied in transplant recipients. In contrast, the sensitivity of the BDG assay was not affected. However, this finding is surprising, since cells of the immune system are commonly supposed to be a major cause of clearance of fungal Ag from the bloodstream. For instance, testing for the Aspergillus-specific Ag galactomannan is considered to have the highest sensitivity in immunocompromised and immunosuppressed individuals, i.e., in hematologic patients and particularly in bone marrow transplant recipients [27]. Notably, detection rates of both Ag-based tests, i.e., the Wako BDG assay and the Serion ELISA, were demonstrated to be numerically better (but without reaching statistical significance) in male patients than in female patients (71% and 33% vs. 60% and 24%) and in individuals age ≥ 65 years than in younger individuals (73% and 43% vs. 62% and 21%). This difference was significant in the subgroup of women < 65 years, when the sensitivity of the Ag ELISA decreased to only 9%. To our knowledge, the finding of a gender dependency of an Ag-based assay has not been reported before. The nature of this phenomenon remains to be elucidated.

Diagnosis of candidemia via culture is impaired by two major drawbacks: (1) the significant delay due to incubation until positivity and (2) the low sensitivity of BC of about 50% [9]. The phenomenon of poor BC positivity rates was also observed in the present study. In only 29% of all cases with ≥ 2 BC on day 0, Candida was cultivated in all BC samples. Interestingly, in a subgroup of 50 cases, for which sera were available that were sampled prior or simultaneously to at least one BC negative for Candida, 50% and 58% of patients were Ag/Ab and BDG seropositive, respectively. Notably, increasing the time interval between sampling of the serum and the positive BC only marginally affected the sensitivity of serology. Furthermore, the TTP of the BC has to be considered, which typically results in a further delay of diagnosis of additional 2–3 days [28]. Consequently, combining BCs with serology will most likely allow for a more rapid and early diagnosis of candidemia which is pivotal for the outcome [29]. To make use of this temporal advantage, rapid processing of the sample is required. For economic reasons, serology laboratories will typically collect specimens for batch runs, especially if ELISAs such as the Fungitell BDG assay are applied [11]. The Ravo Hemkit hemagglutination assay and the Wako BDG test, which were used in this study, are suitable for analysis of single samples and therefore allow for prompt reporting.

It must be considered that the inclusion criterion of a BC positive for Candida may confound the results and interpretation of this study. For example, it appears more likely to detect fungal Ag in the blood of patients whose vascular fungal load is high enough to yield growth in BC than in blood where the fungal load is too low to yield positive BCs. For similar reasons, it is not possible to estimate the overall sensitivity of serology in patients with Candida blood stream infections. Future studies evaluating the performance of serological biomarkers in culture-negative candidemia are necessary to address this issue.

For all three tests evaluated in this study, the amount of data evaluating their performance is limited. Our results demonstrate that the analysis of serological markers is a valuable tool for the early diagnosis. All tests demonstrated high specificities. The Wako BDG test is characterized by a significantly higher sensitivity compared to the combination of Candida-Ag and anti-Candida-Ab assays. The latter ones were found to suffer from impaired sensitivity depending on host factors and fungal species. Optimizing the cutoff levels for BDG and Ag testing may improve the performance of both tests for the diagnosis of candidemia.