ABSTRACT
We conducted a case-control study using the Fungitell assay, the novel Platelia Candida Antigen (Ag) Plus and Candida Antibody (Ab) Plus assays, and the Cand-Tec latex agglutination test to evaluate the usefulness of (1→3)-β-d-glucan (BDG), mannan antigen with/without anti-mannan antibody, and Cand-Tec Candida antigen measurement for the diagnosis of candidemia. A total of 56 patients fulfilled the inclusion criteria and were enrolled. One hundred patients with bacteremia and 100 patients with sterile blood cultures served as negative controls. In the candidemia group, median (1→3)-β-d-glucan, mannan antigen, and anti-mannan antibody levels were 427 pg/ml, 190 pg/ml, and 18.6 antibody units (AU)/ml, respectively. All three parameters were significantly elevated in patients with candidemia. The sensitivity and specificity were, respectively, 87.5% and 85.5% for (1→3)-β-d-glucan, 58.9% and 97.5% for mannan antigen, 62.5% and 65.0% for anti-mannan antibody, 89.3% and 63.0% for mannan antigen plus anti-mannan antibody, 89.3% and 85.0% for BDG plus mannan antigen, and 13.0% and 93.9% for Cand-Tec Candida antigen. The low mannan antigen sensitivity was in part caused by Candida parapsilosis and Candida guilliermondii fungemias, which were not detected by the Platelia Candida Ag Plus assay. When the cutoff was lowered from 125 pg/ml to 50 pg/ml, mannan antigen sensitivity increased to 69.6% without severely affecting the specificity (93.5%). Contrary to recently published data, superficial candidiasis was not associated with elevated mannan antigen levels, not even after the cutoff was lowered. Combining procalcitonin (PCT) with (1→3)-β-d-glucan to increase specificity provided a limited advantage because the benefit of the combination did not outweigh the loss of sensitivity. Our results demonstrate that the Cand-Tec Candida antigen and the mannan antigen plus anti-mannan antibody measurements have unacceptably low sensitivity or specificity. Of the four tests compared, (1→3)-β-d-glucan and mannan antigen are the superior biomarkers, depending on whether a sensitivity-driven or specificity-driven approach is used.
INTRODUCTION
Candida species account for approximately 10% of bloodstream infections (BSI) in intensive care units (ICUs) and are associated with a crude in-hospital mortality rate of 30% (1, 2). Because a delay in the initiation of antimycotic therapy is associated with increased mortality, timely diagnosis is of utmost importance (3). While blood cultures are still regarded as the gold standard for diagnosis of candidemia, it takes about 2 days to obtain a positive result, and sensitivity can be as low as 50% (4, 5). With these limitations of culture-based Candida detection, it is quite evident that faster and more sensitive techniques are required.
For this purpose a number of clinical prediction rules (6–8) and non-culture-based methods are available. The latter include the detection of Candida DNA and circulating fungal antigens in serum. While nucleic acid amplification techniques are still lacking standardization, commercial tests are available for the measurement of (1→3)-β-d-glucan (BDG), mannan antigen (Ag), and Cand-Tec Candida antigen (CA). All of these tests have been evaluated for their performance in the diagnosis of invasive candidiasis with sensitivities and specificities, respectively, of 77% and 85% for BDG, 58% and 93% for mannan Ag, and 64% and 58% for CA (9). By combining mannan Ag with anti-mannan antibody (Ab) measurement, the sensitivity and specificity can be increased to 83% and 86%, respectively (10). Recently, there have been changes concerning these biomarkers. It has been reported that BDG levels are elevated in bacteremia, questioning the validity of this marker for the diagnosis of invasive fungal disease (11, 12). However, studies including relevant numbers of bacteremic patients are lacking. Furthermore, the most widely used assay for mannan Ag detection, the Platelia Candida Ag assay, was recently refined (now known as the Platelia Candida Ag Plus assay). So far, only one study has examined this novel test format (13).
We therefore conducted a case-control study to compare the diagnostic performance of serum BDG (Fungitell), mannan Ag with/without mannan Ab (Platelia Candida Ag/Ab Plus), and CA (Cand-Tec) detection for the diagnosis of candidemia.
(Part of this research was presented at the 18th Congress of the International Society for Human and Animal Mycology, Berlin, Germany, 11 to 15 June 2012.)
MATERIALS AND METHODS
Candidemia patients.All patients presenting at the University Medical Centre Freiburg, Germany, between January 2001 and May 2012 were enrolled. Inclusion criteria were a culture-confirmed candidemia and an archived serum sample from day 0 until day 2 after blood culture sampling. An exclusion criterion was treatment with intravenous immunoglobulins (IVIG) or albumin in the 7 days prior to serum sampling. Patients receiving IVIG and albumin were excluded because we have found in the past that, without exception, even a single administration of these substances causes significantly elevated BDG levels that usually normalize within 1 week (14).
A total of 79,840 blood cultures were examined. Candida species were detected in 754 of 10,987 positive blood cultures (6.9%) from 307 patients. Sixty-eight patients fulfilled the inclusion criteria. Twelve patients were excluded because they had received IVIG or albumin. Finally, 56 patients were enrolled in the study. The underlying diseases were hematologic malignancy/hematopoietic stem cell transplantation (HSCT; n = 9), solid tumor (n = 12), gastrointestinal disease (n = 6), major surgery (n = 4), abdominal surgery (n = 16), and miscellaneous disease (n = 9). The Candida species isolated were Candida albicans (n = 32), Candida glabrata (n = 11), Candida tropicalis (n = 6), Candida parapsilosis (n = 4), and Candida guilliermondii (n = 3).
Controls.Sera from two control groups were examined. Control group 1 consisted of 100 patients with a negative blood culture, and control group 2 consisted of 100 patients with bacteremia. In both groups an archived serum sample from the day of blood culture sampling was required for inclusion. Control patients receiving IVIG or albumin were excluded. Bacteremia was caused by the following pathogens: Bacillus species (n = 3), Bacteroides species (n = 3), Citrobacter species (n = 2), Enterobacter cloacae (n = 5), Enterococcus faecalis (n = 5), Enterococcus faecium (n = 5), Escherichia coli (n = 5), Haemophilus influenzae (n = 2), Klebsiella oxytoca (n = 3), Klebsiella pneumoniae (n = 3), Listeria monocytogenes (n = 2), Pseudomonas aeruginosa (n = 10), Serratia marcescens (n = 3), Staphylococcus aureus (n = 10), Staphylococcus epidermidis (n = 5), Staphylococcus haemolyticus (n = 3), Staphylococcus hominis (n = 5), Streptococcus agalactiae (n = 3), Streptococcus anginosus group (n = 3), Streptococcus bovis (n = 3), Streptococcus dysgalactiae (n = 2), Streptococcus mitis (n = 3), Streptococcus mutans (n = 2), Streptococcus oralis (n = 2), Streptococcus pneumoniae (n = 5), and Streptococcus pyogenes (n = 3). Patients with bacteremia caused by coagulase-negative staphylococci were included only if at least two different blood culture sets were positive.
Serum and data collection.Patient demographics, clinical characteristics, laboratory results, risk factors for candidemia, and microbiological results were collected (Table 1). Serum samples were originally taken for various microbiological analyses and were frozen at −80°C. The use of these sera was approved by the local ethics committee (application number 293/11).
Baseline characteristics of the study populations and risk factors for candidemia
Serum antigen measurement.All tests were performed according to the manufacturer's recommendations at our own institution. The persons who tested the sera were not blinded. The Fungitell assay (Associates of Cape Cod, MA) was used for BDG measurement. The recommended cutoff is ≥80 pg/ml. Each serum was tested in duplicate. Samples with BDG levels above 500 pg/ml were diluted and retested. BDG levels below 31 pg/ml (lower validation limit) were calculated by extrapolation.
The Platelia Candida Ag Plus and the Platelia Candida Ab Plus assays (Bio-Rad, France) were used for mannan Ag and mannan Ab measurement, respectively. The recommended cutoff is ≥125 pg/ml for the antigen and ≥10 antibody units (AU)/ml for the antibody. Samples with mannan Ag levels above 500 pg/ml were diluted and retested.
The Cand-Tec latex agglutination test (Ramco Laboratories, TX) was used for CA measurement. The recommended cutoff titer is ≥1:4. Each CA test was checked by two readers, and if the results were incongruent, they were ratified by a third user in a blinded fashion. CA-positive sera were not titrated to the endpoint, and therefore the results are only qualitative.
All assays used bear the European CE marking. The Fungitell assay in addition is FDA approved.
Statistical methods.Statistical analysis was performed using SPSS, version 19, and MedCalc, version 12. For the comparison of variables, Fisher's exact test, a Mann-Whitney U test, and a Kruskal-Wallis test were used. Differences were considered significant at a P value of <0.05. The optimal cutoffs were determined by receiver operating characteristic (ROC) analysis (maximum Youden index).
RESULTS
Diagnostic performance.The main results are shown in Tables 2 and 3. In the candidemia group, median analyte levels were 427 pg/ml (BDG), 190 pg/ml (mannan Ag), and 18.6 AU/ml (mannan Ab). All three parameters were significantly elevated in patients with candidemia (P < 0.001). To determine optimized cutoffs, ROC analyses were performed. The ROC curves are depicted in Fig. 1. As mentioned in Materials and Methods, CA results are only qualitative, and therefore no ROC analysis was performed. The area under the ROC curve (AUC) was largest for BDG (0.925); however, the difference between the AUCs of BDG and mannan Ag (0.898) was statistically not significant (P = 0.361). The AUCs of both BDG and mannan Ag were significantly larger than the AUC of mannan Ab (0.673; P < 0.001). The optimized cutoffs were ≥70 pg/ml for BDG, >50 pg/ml for mannan Ag, and ≥15 AU/ml for mannan Ab.
Main test results stratified according to patient group
Diagnostic performance of the different biomarkers using manufacturer and optimized cutoffs
ROC curves for BDG, mannan Ag, and mannan Ab. BDG shows the largest AUC; however, the difference between the AUC of BDG and that of mannan Ag is statistically not significant (P = 0.361). The AUCs of both BDG and mannan Ag are significantly larger than the AUC of mannan Ab (P < 0.001). According to the manufacturer, the mannan Ab detection should be used only in combination with mannan Ag detection. The manufacturer cutoffs were designed for a broad range of invasive fungal infections and therefore differ from the optimized cutoffs that were determined by us for candidemia alone. The criterion for the optimized cutoffs was the highest Youden index in the ROC analysis (indicated as circles). BDG, (1→3)-β-d-glucan; Ag, antigen; Ab, antibody; AUC, area under the ROC curve; CI, confidence interval; AU, antibody unit.
The sensitivities using the manufacturer's and the optimized cutoffs, respectively, were 87.5% versus 89.3% for BDG, 58.9% versus 69.6% for mannan Ag, 62.5% versus 58.9% for mannan Ab, 89.3% versus 89.3% for mannan Ag plus Ab, and 89.3% versus 92.9% for BDG plus mannan Ag. The overall specificities, again listed for the manufacturer's and the optimized cutoffs, were 85.5% versus 84.0% for BDG, 97.5% versus 93.5% for mannan Ag, 65.0% versus 71.5% for mannan Ab, 63.0% versus 65.5% for mannan Ag plus Ab, and 85.0% versus 81.0% for BDG plus mannan Ag. No optimized cutoffs were determined for CA; using the manufacturer's cutoffs sensitivity was 13.0% and specificity was 93.9% for CA. The specificity of BDG in patients with bacteremia (81%) was lower than in patients with negative blood cultures (90%; P = 0.053). The specificities of the other analytes were independent of the control group.
Combination of BDG with PCT to increase specificity.Procalcitonin (PCT) was measured in 29 patients with candidemia (mean PCT of 0.80 μg/liter), 52 patients with bacteremia (mean PCT of 2.36 μg/liter), and 52 patients with negative blood cultures (mean PCT of 0.27 μg/liter). Patients with bacteremia had significantly higher PCT levels than the combined group consisting of patients with candidemia plus negative blood cultures (P = 0.001). A recommended cutoff for discriminating these two patient populations is a PCT value of ≥2.0 μg/liter (15). The combination of BDG with PCT led to a considerable increase in specificity from 89.4% to 96.2%, accompanied by a loss of sensitivity from 86.7% to 51.7%.
BDG and mannan Ag in microbiological and clinical subgroups.Subgroup analysis of BDG and mannan Ag (Table 4) revealed that both parameters were significantly elevated in control patients following major surgery and abdominal surgery during the 4 weeks prior to serum sampling (P = 0.009 and P = 0.010 for BDG and mannan AG, respectively). While median BDG and mannan Ag levels in patients after major surgery were both below the cutoff, the median BDG level after abdominal surgery was 174 pg/ml and therefore would lead to a positive test interpretation. As mentioned above, the BDG false-positivity rate was higher in patients with bacteremia than in patients with negative blood cultures (19% versus 10%; P = 0.053). While there was no difference in BDG levels between Gram-positive and Gram-negative BSI (P = 0.76), patients with Enterococcus faecalis bacteremia had significantly higher BDG levels than patients with bacteremia of other origins (135 pg/ml versus 15 pg/ml; P = 0.04). Raising the BDG cutoff to 135 pg/ml resulted in an overall sensitivity and specificity of 77.0% and 89.5%, respectively.
BDG and mannan Ag levels in microbiological and clinical subgroups
BDG levels between Candida species showed no significant difference (P = 0.296). In contrast, patients with Candida parapsilosis and Candida guilliermondii fungemia had significantly lower mannan Ag levels than patients with fungemia caused by other Candida species (P = 0.005 and P = 0.046). In our study, Candida parapsilosis and Candida guilliermondii isolates were not detected by mannan Ag measurement. ICU patients and patients with Candida colonization had elevated mannan Ag levels. However, because the levels were below the cutoff, this would have had no clinical consequence.
Comparison of BDG with mannan Ag detection showed that the levels of both markers correlated in candidemia (P < 0.001), when the extreme outliers (BDG of >10,000 pg/ml; mannan Ag of >20,000 pg/ml) were excluded. There was no correlation between BDG (P = 0.913) or mannan Ag levels (P = 0.608) and the time-to-positivity of blood culture. In patients without candidemia, BDG levels increased with the time in the ICU (P = 0.001). This was not the case for mannan Ag (P = 0.773). However, the median BDG level of ICU patients was 20 pg/ml and therefore clinically not relevant.
DISCUSSION
We conducted a retrospective study using the Fungitell assay, the Platelia Candida Ag and Ab Plus assays and the Cand-Tec latex agglutination test to evaluate BDG, mannan Ag, mannan Ab, and CA as biomarkers to aid in the diagnosis of candidemia.
Due to its low sensitivity (13.0%), the benefit of the Cand-Tec latex agglutination test in clinical routine is limited, and it should not be used on its own. Of the remaining biomarkers, BDG and the combination of mannan Ag plus Ab showed the highest sensitivities (87.5% and 89.3%, respectively). While BDG measurement produced a reasonable specificity of 85.5%, the specificity of mannan Ag plus Ab was only 63%. Responsible for this poor result is the integration of the antibody component, with a specificity of 65% when used exclusively. A similarly high sensitivity but with a specificity of 85% is achieved by combining BDG with mannan Ag. However, it is questionable if an increase in sensitivity of 1.8% justifies the extra costs of combination testing. Mannan Ag detection alone showed an excellent specificity of 97.5%, but this is offset by an intolerably low sensitivity of 58.9%. The ROC analysis revealed that by lowering the Platelia Candida Ag Plus cutoff from 125 pg/ml to 50 pg/ml, the sensitivity would increase to 69.6% without severely affecting the specificity (93.0%). In contrast, sensitivities and specificities of BDG measurement and measurement of the combination of mannan Ag plus Ab did not improve substantially by utilization of an optimized cutoff.
The lower sensitivity of mannan Ag detection is at least partially explained by our observation that Candida parapsilosis and Candida guilliermondii were not detected by the Platelia Candida Ag Plus assay. These two species comprised 12.5% of all Candida isolates. The sensitivity of mannan Ag in the other species was as follows: Candida albicans, 56%; Candida glabrata, 82%; and Candida tropicalis, 100%. Similar results were obtained with the predecessor of the current test. Fujita et al. described lower sensitivities for Candida krusei (0%), Candida parapsilosis (15%), Candida guilliermondii (27%), and Candida glabrata (36%) than for Candida albicans (78%) and Candida tropicalis (67%) (16). In an in vitro study, Rimek et al. analyzed the cross-reactivity of 63 fungi in the Platelia Candida Ag assay (17). Although they were using highly concentrated culture extracts, they were unable to detect Candida krusei, Candida parapsilosis, and a number of less frequently encountered Candida species. The reason for this nonhomogeneity in detection of different Candida species probably lies in the monoclonal EBCA-1 antibody, which was generated by immunization of rats with Candida albicans. This antibody binds to oligomannoside antigens that are abundant in the cell wall of Candida albicans, Candida glabrata, and Candida tropicalis but have been found to a lesser extent in Candida krusei and Candida parapsilosis (18). In addition, the variable antigen content may partially explain the missing correlation between the time-to-positivity of blood cultures and mannan Ag levels in our study. Existing antimycotic therapy, different sampling locations (catheter/peripheral vein), and the variable growth rates of Candida species (19) may further confound the time-to-positivity.
So far, there has been only one retrospective study using the novel Platelia Candida Ag and Ab Plus assays (13). In 21 patients with invasive candidiasis, Lunel et al. found an overall per patient sensitivity and specificity, respectively, of 61.9% and 43.3% for mannan Ag and 47.2% and 86.7% for mannan Ab (13). Because they followed a screening approach (multiple sera per patient) and used lower cutoffs (62.5 pg/ml and 5 AU/ml) and because a positive patient needed to have at least two positive sera, a direct comparison of their results and our own is not possible. However, the low specificity for the antigen and the high specificity for the antibody are surprising and in contrast to our results. Based on logistic regression analysis, Lunel et al. argued that the low antigen specificity was due to detection of mannan from patients with superficial candidiasis (n = 4). In our study, Candida species were cultured from samples of nonsterile sites in 40 of the control patients. Ten of these patients had a positive oral swab, and five patients suffered from thrush. However, none of these patients had a positive mannan Ag test, arguing against superficial candidiasis as a cause of elevated mannan Ag levels.
In candidemia patients with central venous catheters, mannan Ag levels were significantly elevated. This may be a consequence of a higher fungal load of the blood in these patients; alternatively, the blood may have been drawn through a Candida-colonized central venous line. The latter would point to an interesting question. Are different mannan Ag levels in blood drawn from a central venous catheter and a peripheral vein indicative of catheter infection? Further studies are needed to answer this question.
BDG has proved to be a sensitive biomarker for invasive fungal infections; however, because of its panfungal nature and various confounding factors, its specificity has always been an issue. While IVIG and albumin are definitely sources of false-positive results, other factors remain controversial (14). Among these possible confounding factors are bacteremia, certain antibiotics, Candida colonization, and treatment in an ICU. We compared the BDG levels of clinical subgroups and could not find significant differences between bacteremia versus no bacteremia, antibiotic therapy versus no antibiotic therapy, β-lactam therapy versus no β-lactam therapy, and Candida colonization versus no colonization. However, BDG levels were significantly elevated in patients with Enterococcus faecalis bacteremia and in the month after abdominal surgery. Of 10 patients with Enterococcus faecalis bacteremia, five had a positive BDG result. Because we were not able to detect relevant amounts of BDG in Enterococcus faecalis culture supernatants (data not shown), it is unclear whether Enterococcus faecalis itself is really the BDG source. One alternative explanation would be that Enterococcus faecalis bacteremia indicates only a gut barrier loss, resulting in an increased permeability for BDG. The same is the case for patients in the month after abdominal surgery. It is known that surgical exposure to sponges or gauze can lead to elevated BDG levels (20). However, baseline levels seem to be restored after 3 days (21), and the fact that these patients were still in the hospital several weeks after the operation suggests that they suffered from potential complications like intestinal perforation with subsequent BDG translocation. Further studies are clearly needed to elucidate the impact of the various confounding factors on BDG measurement, always keeping in mind that BDG is not specific for Candida species but is also detected in other fungal infections like aspergillosis or pneumocystosis.
Besides the statistically significant differences, comparison of the analyte levels in the various subpopulations also showed some interesting trends. In particular, the higher BDG and mannan Ag levels in nonsurvivors point to a possible prognostic relevance of these biomarkers.
An approach to improve the performance of antigen assays is the combination with other diagnostic tools. One example is a study by Posteraro et al., who used BDG measurement together with the Candida score in 14 patients with invasive candidiasis and increased the sensitivity from 92.9% to 100%. However, this gain in sensitivity came at the price of a loss of specificity from 93.7% to 83.5% (22). The number of false positives in our study was particularly high in patients with bacteremia. Because there is evidence that PCT can be used to distinguish between bacteremia and candidemia (15), we decided to examine the effect of combining PCT with BDG in order to increase specificity. By doing so, specificity increased considerably, but the benefit of the combination did not outweigh the drastic loss of sensitivity. One reason for this was certainly that, in our study, PCT levels of bacteremia patients were not significantly higher than those of candidemia patients (P = 0.36).
Altogether, our data show that BDG, mannan Ag, and mannan Ab levels were significantly elevated in patients with candidemia. Candida parapsilosis and Candida guilliermondii were not detected by the Platelia Candida Ag Plus assay, reducing the sensitivity of mannan Ag measurement. Of the four tests compared, BDG and the combinations of BDG plus mannan Ag and of mannan Ag plus Ab showed the highest sensitivity. However, BDG with or without mannan Ag revealed a moderate specificity, and mannan Ag plus Ab had an unacceptably low specificity. The combination of PCT with BDG was of no essential benefit. BDG and mannan Ag seem to be the superior biomarkers for the diagnosis of candidemia, depending on whether a sensitivity-driven or specificity-driven approach is used.
ACKNOWLEDGMENTS
We thank Christian Schneider and Tanja Schmidt for the data mining and Manfred Olschewski for the assistance with the statistical analysis.
This work was not specially funded. Some of the results were generated during routine diagnostic activities.
No commercial relationship or potential conflict of interest exists.
FOOTNOTES
- Received 16 September 2012.
- Returned for modification 26 November 2012.
- Accepted 23 January 2013.
- Accepted manuscript posted online 30 January 2013.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.