ABSTRACT
The 1,3-beta-d-glucan (BDG) test is used for the diagnosis of invasive candidiasis (IC) in intensive care units (ICUs). However, its utility for patient management is unclear. This study assessed the impact of BDG test results on therapeutic decisions. This was a single-center observational study conducted in an ICU over two 6-month periods. All BDG test requests for the diagnosis of IC were analyzed. Before the second period, the ICU physicians received a pocket card instruction (algorithm) for targeted BDG testing in high-risk patients. The performance of the BDG test for IC diagnosis was assessed, as well as its impact on antifungal (AF) prescription. Overall, 72 patients had ≥1 BDG test, and 14 (19%) patients had an IC diagnosis. The BDG test results influenced therapeutic decisions in 41 (57%) cases. The impact of the BDG test was positive in 30 (73%) of them, as follows: AF abstention/interruption following a negative BDG result (n = 27), and AF initiation/continuation triggered by a positive BDG test result and subsequently confirmed IC (n = 3). In 10 (24%) cases, a positive BDG test result resulted in AF initiation/continuation with no further evidence of IC. A negative BDG result and AF abstention with subsequent IC diagnosis were observed in one case. The positive predictive value (PPV) of BDG was improved if testing was restricted to the algorithm’s indications (80% versus 36%, respectively). However, adherence to the algorithm was low (26%), and no benefit of the intervention was observed. The BDG result had an impact on therapeutic decisions in more than half of the cases, which consisted mainly of safe AF interruption/abstention. Targeted BDG testing in high-risk patients improves PPV but is difficult to achieve in ICU.
INTRODUCTION
Invasive candidiasis (IC), due to the pathogenic Candida sp. yeasts, is a frequent cause of nosocomial infection in intensive care unit patients and is still associated with high mortality rates (1–3). IC accounts for an important proportion of bloodstream infections (i.e., candidemia) but also of blood culture-negative infections, such as intra-abdominal candidiasis (IAC) (2, 4). IAC occurs in as many as 30% to 40% of patients with complicated abdominal surgery and represents an independent risk factor of mortality (5, 6). Early diagnosis and prompt initiation of antifungal (AF) therapy are crucial in the management of these life-threatening infections. As a consequence, the use of prophylactic or empirical AF therapy in intensive care units (ICU) is associated with important costs and an epidemiological shift toward more resistant Candida species (2, 7, 8).
Several strategies have been proposed to identify high-risk ICU patients for a targeted prophylactic or preemptive approach of AF drug prescription. Clinical scores (e.g., Candida score) or prediction rules have acceptable negative predictive value but poor positive predictive value (9, 10). The 1,3-β-d-glucan (BDG) test, which can detect a fungal polysaccharide in serum, has gained interest in the diagnosis of IC. The BDG test was found to be superior to the Candida score for the early diagnosis of candidemia or IAC in high-risk ICU patients (5, 11). Several studies have assessed the performance of the BDG test for the diagnosis of IC in ICUs with the proposition of different cutoffs and varied sensitivity and specificity results (9, 11–14). The BDG test seems to be more useful when combined with the Candida score or restricted to high-risk surgical patients (5, 15). The use of BDG testing in this setting can reduce the use of AF drugs (15, 16). Nevertheless, data on the real impact of the test on AF drug prescription in clinical practice are lacking. The purpose of this study was to assess how the BDG test influences therapeutic decisions and its impact in terms of early diagnosis of IC and the use of antifungals.
MATERIALS AND METHODS
Setting and patients.This observational study was conducted in the ICU of the University Hospital of Lausanne (Switzerland). The BDG test (Fungitell; Associates of Cape Cod, Inc., Falmouth, MA, USA) was introduced in our institution in February 2017. We screened all ICU patients for whom a BDG test was ordered over two distinct 6-month periods, February to July 2017 (period 1) and May to October 2018 (period 2). During period 1, the decision to order a BDG test and the result interpretation were left at the sole discretion of the ICU attending physician. Before period 2, instruction was provided to the ICU physicians as part of a quality control program for improved AF use practices. Indications to order a BDG test as part of the diagnostic workup for preemptive or empirical AF treatments were defined in an algorithm, which was distributed as a pocket card to all ICU attending physicians and the infectious disease consultants (Fig. 1). While recommendations were provided for the indications and interpretation of the BDG test, the decision to start or withhold AF therapy was left at the sole discretion of the clinician. There was no defined policy in terms of AF prophylaxis for complicated surgery (i.e., recurrent gut transection or anastomotic leakage) in this ICU.
Pocket cards distributed to clinicians during the second period of study.
Demographic, clinical, and microbiological data were collected from medical records and laboratory databases. Only patients for whom a BDG test was ordered to test for IC were included in the study (i.e., excluded were cases for which the test was ordered for suspicion of Pneumocystis jirovecii pneumonia or for the diagnosis of invasive fungal infections in the setting of hematologic cancer and/or neutropenia with an absolute neutrophil count of <500 cell/mm3). Patients were followed for the entire duration of their hospital stay for the incidence of IC according to the below-described criteria.
BDG testing and microbiological cultures.BDG concentration was measured in serum samples by the Fungitell assay, according to the manufacturer’s instructions (17). Samples were analyzed in batches twice per week according to the routine procedure of our institution. BDG results were provided to the clinicians quantitatively and interpreted according to the manufacturer’s cutoff, with a test considered positive, intermediate, or negative for values of ≥80 pg/ml, 60 to 79 pg/ml and <60 pg/ml, respectively. For all positive BDG tests, a system of automatic email alerts for the principal study investigators was set up. In case of positive BDG results, a second test of confirmation on a distinct serum sample was usually required to exclude analytical errors or contamination. Blood cultures were incubated in an automatic system (Bactec Plus aerobic and anaerobic/F bottle; Becton, Dickinson, Sparks, MD). Fungal cultures other than blood were performed on Sabouraud dextrose agar with selective antibiotics and daily reading for 4 days. Identification of yeasts was performed by matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS; MALDI BioTyper, Bruker Daltonik GmbH, Leipzig, Germany).
Definitions.Candidemia was defined as isolation of a Candida sp. from at least one blood culture bottle in the presence of clinical symptoms/signs of infection (18). Intra-abdominal candidiasis (IAC) was defined according to the recommendations of an international expert consensus, as the presence of Candida spp. isolated from an intra-abdominal sample obtained by surgery, by percutaneous aspiration, or from a drain inserted less than 24 h from sampling (4, 19). Other IC were defined, by extension of the previous definition, as the presence of Candida spp. isolated from any sample from a normally sterile site, with the exception of urine (e.g., pleural fluid) obtained under the above-mentioned conditions. The reasons for starting antifungals were classified as follows: targeted AF therapy (after documentation of IC according to the above-mentioned criteria), empirical AF therapy (in case of signs of infection in the absence of any microbiological documentation of IC and irrespective of BDG results), preemptive (on the basis of a positive BDG and in the absence of documented IC by cultures), and prophylaxis (on the basis of high-risk conditions, in the absence of any sign of infection and any microbiological documentation of IC).
Analyses.The influence of BDG test results on AF therapeutic decisions (yes versus no) was assessed by two infectious diseases specialists according to the decisional tree presented in Text S1 in the supplemental material, taking into account (i) the presence or absence of IC criteria (according to the above-described definitions) at the time of BDG result, (ii) the documentation in medical records about the reasons triggering AF therapeutic decisions, and (iii) the discretion of the investigator if AF therapeutic decisions were in accordance with the algorithm (Fig. 1) without any alternative explanation. The impact of BDG results on therapeutic decisions was then evaluated as positive, undetermined, or negative according to the final diagnosis of IC or not in follow-up (Fig. 2). In addition to the actual impact of the BDG test on AF prescription, whether a virtual or potential impact, on the hypothesis of a strict adherence to the algorithm was calculated.
Impact of BDG results on therapeutic decisions. The impact of BDG result was assessed according to the following definitions: (i) Positive (green). A positive BDG result led to AF therapy initiation or continuation with IC subsequently confirmed, or a negative BDG result led to AF therapy abstention or interruption without subsequent diagnosis of IC. (ii) Undetermined (gray). A positive BDG result led to AF therapy continuation or initiation without subsequent confirmation of IC diagnosis (whether early start of AF therapy was unnecessary or could have prevented further development of IC was not evaluable in this scenario). (iii) Negative (red), a negative BDG result led to AF abstention or interruption with subsequent diagnosis of IC.
The performance of the BDG test for the diagnosis of IC was assessed for different cutoffs in terms of sensitivity, specificity, and positive and negative predictive values. The gold standard was IC, including candidemia, IAC, and other IC, as defined above. The performance of the test was compared between the entire population (i.e., all patients for whom a BDG test was ordered for IC diagnosis) and the restricted population of high-risk patients fulfilling the criteria of the algorithm (Fig. 1).
Data were analyzed for the entire study (period 1 and period 2 pooled) and separately for each period to assess the impact of the intervention.
Statistical analysis.To assess for differences between groups, categorical data were compared using Fisher’s exact test, and continuous data were compared using Student's t test. Statistical significance was set at a P value of ≤0.05.
Ethical statement.The study protocol was approved by the ethics committee of Canton de Vaud (project number 2017-00867).
Data availability.All data supporting our findings are available from the corresponding author upon reasonable request.
RESULTS
Patient characteristics.During the study period, 127 ICU patients had at least one BDG test. Fifty-five patients were excluded from the analysis because the BDG test was performed for the diagnosis of fungal diseases other than IC. Of the 72 remaining cases that were tested for BDG in search of IC, 45 (63%) cases had a Candida score of ≥3, and 19 (26%) cases fulfilled the indications for BDG testing as defined by our algorithm (Fig. 1). A total of 139 BDG tests were performed for these patients (median, 2 tests; range, 1 to 5 tests). The median turnaround time from sampling to result was 2 open days. Diagnosis of IC was established by cultures in 14 (19%) of them (4 candidemia and 10 noncandidemic IC [8 IAC and 2 other IC]). C. albicans represented the most frequent etiological agent of IC (64%).
Demographic and clinical characteristics of the IC versus non-IC cases are shown in Table 1. Patients with IC were predominantly males (P < 0.01) and were older (P = 0.02) with a higher Simplified Acute Physiology Score (SAPS II) (P = 0.04) and higher prevalence of septic shock (P < 0.01) than in patients without IC. Total parenteral nutrition was the only intervention significantly associated with IC (P = 0.02).
Patient demographics and clinical characteristics
AF therapy was initiated in all patients with an IC diagnosis and in 34 (59%) patients without IC. Most patients (n = 31) received an echinocandin, and the remaining (n = 17) patients were treated by fluconazole. Initially, the reason for AF therapy was targeted treatment of IC in 7 (14.5%) cases, empirical therapy in 34 (71%) cases, and preemptive therapy in 7 (14.5%) cases. There was no patient receiving antifungal prophylaxis.
Impact of BDG test.The results of the BDG test were considered to have influenced therapeutic decisions in 41/72 (57%) cases (Fig. 2). In 7 (10%) cases, a diagnosis of IC was already confirmed at the time of the BDG test result. The 65 remaining cases for which BDG could have influenced therapeutic decision (i.e., no IC diagnosis at time of BDG test result interpretation), were classified according to the following two approaches: (i) a preemptive approach (n = 31) in which AF therapy was not yet started and the BDG test result was used to trigger the decision to start or not to start AF therapy, and (ii) an empirical approach (n = 34) in which AF therapy was already started and the BDG test result was used to trigger the decision to continue or interrupt AF therapy. The BDG test result had an influence on therapeutic decisions in 23/31 (74%) cases in the preemptive group and 18/34 (53%) cases in the empirical group (P = 0.25) (Fig. 2).
In the preemptive group, the impact of the BDG test result consisted of AF therapy initiation in 5/23 (22%) cases and abstention of AF therapy in 18/23 (78%) cases. The impact of the BDG test result in this group was considered positive in 19 (83%) cases, undetermined in 3 (13%) cases, and negative in 1 (4%) case (see details in Fig. 2).
In the empirical group, the impact of the BDG test result consisted of AF therapy continuation in 8/18 (44%) of cases and AF interruption in 10/18 (56%) cases. The impact of the BDG result was considered positive in 11 (61%) cases and undetermined in 7 (39%) cases, without negative impact (see details in Fig. 2).
In total, the BDG test result had a positive impact in 30/41 (73%) cases, which consisted mainly of AF therapy abstention or discontinuation (90% of them) without subsequent IC diagnosis. Discontinuation of empirical AF therapy on the basis of a negative BDG test result occurred at a median of 6 days (range, 3 to 10 days) after the start of AF therapy. The impact of the BDG test result was considered undetermined in 10/41 (24%) cases, for which AF therapy was started or continued on the basis of a positive BDG test result. While these patients did not fulfill the criteria of IC in follow-up, it is not possible to assess whether AF therapy may have prevented/treated early IC or was unnecessary in these cases. No adverse events of AF therapy were observed in these cases. Finally, a negative impact (AF abstention on the basis of a negative BDG test and subsequent IC diagnosis) was observed in one (2%) case. This patient had IAC following a necrotizing pancreatitis. AF therapy was, however, started 2 days later on the basis of a positive culture for Candida spp., and the patient survived.
In addition to the actual impact of the BDG test, we estimated the potential impact among the 24 cases for which BDG results could have influenced therapeutic decisions but were not taken into account. For 13 (54%) of these cases, AF therapy was not initiated despite a positive BDG test result. A subsequent diagnosis of IC was established by cultures in 2 (22%) of them within 4 to 5 days of BDG testing. For 11 (46%) cases, a negative BDG test result represented a missed opportunity to interrupt AF therapy. A subsequent diagnosis of IC was established in one (9%) of them. This patient had a noncandidemic IC with Candida spp. recovered by culture in an empyema.
Overall, strict adherence to the algorithm regarding the interpretation of BDG test results for therapeutic decisions would have resulted in a positive impact in 42/65 (65%) cases, of which 88% would involve abstention/interruption of AF therapy without subsequent IC, and 12% would involve initiation/continuation of AF therapy with subsequent IC diagnosis. For 2 (3%) cases, the test would have negatively impacted the therapeutic decision with potential life-threatening consequences (interruption or abstention of AF therapy on the basis of negative BDG result with subsequent IC). For the 21 (32%) remaining cases, AF therapy would have been started on the basis of a positive BDG result without documented IC in follow-up according to culture results (undetermined impact).
BDG diagnostic performances.The diagnostic performance of the BDG test for the diagnosis of IC using the recommended cutoff (e.g., ≥80 pg/ml) is shown in Table 2, and the receiver operating characteristic (ROC) curve for different cutoffs is shown in Fig. 3. The specificity and positive predictive value (PPV) of the BDG test were much better when the test was performed in selected high-risk patients according to the algorithm (80% versus 36% for the entire population). However, the negative predictive value (NPV) was somewhat decreased (90% to 79%). The ROC curve confirmed the overall better performance of the test within the targeted population at risk of IC (area under the curve [AUC], 0.77 versus 0.73 for the entire study population), with an optimal sensitivity/specificity ratio close to the recommended cutoff (60 to 80 pg/ml).
BDG test performance based on cutoff and indication for testing
Receiver operating characteristics (ROC) curves for diagnosis of invasive candidiasis (IC) according to testing indications.
Comparison of the two study periods.A total of 67 BDG tests in 32 patients were ordered during period 1 (before introduction of the algorithm) compared to 72 tests in 40 patients in period 2 (after introduction of the algorithm). Indications for BDG testing were in accordance with the algorithm’s criteria in 34% and 20% of cases for periods 1 and 2, respectively (P = 0.12). There were no differences in terms of the impact of BDG test results on antifungal drug prescription (69% versus 45% for periods 1 and 2, respectively) and performance of the BDG test.
DISCUSSION
In this study, we assessed the actual impact of the BDG test on therapeutic decisions for the management of ICU patients at high risk of IC. Multiple studies have assessed the performance of the BDG test for IC diagnosis with a wide range of results (sensitivity, 65% to 93%; specificity, 74% to 94%; PPV, 42% to 68%; NPV, 77% to 99%) and different proposed cutoffs (80 to 350 pg/ml) (5, 12, 13, 15, 20). However, data supporting the real utility of the test in clinical practice are lacking. Indeed, the role of BDG for detecting IC is controversial. Some studies suggest that the performance of the BDG test could be better in selected subgroups of ICU patients exhibiting high risk for IC, such as complicated abdominal surgery or a Candida score of ≥3 (5, 11, 21). Some recent reports suggest that the BDG test may be particularly useful for its negative predictive value and reduction of AF drug use (15, 16, 22, 23). However, delayed turnaround times also represent a limitation, and it is difficult to assess how the BDG test can directly influence therapeutic decisions under real-life conditions (22).
Compared to previous analyses (5, 15, 16), our study has the originality to address the actual impact of BDG testing on AF drug prescriptions in the ICU under real clinical conditions with the application of an algorithm distinguishing the application of the test in an empirical and a preemptive strategy. It is, however, limited by the small sample size and monocentric design.
Our analysis found that BDG test results had an impact on therapeutic decisions in 57% of cases. However, the test was considered to be used in the appropriate setting in only 26% of cases, and efforts to improve adherence to the algorithm for targeted BDG testing in high-risk patients were unsuccessful. ICU clinicians used to order BDG testing in both preemptive and empirical strategies. Overall, the impact of BDG test results on therapeutic decisions was considered positive in 57% of cases, which could be slightly increased (65%) when considering the potential impact. This benefit consisted mainly of interruption or abstention of unnecessary antifungal therapy in the case of a negative BDG test result, which is consistent with previous studies (15, 16). However, the NPV in our study was not optimal. Two patients had a final IC diagnosis despite a negative BDG test result, and AF therapy was delayed in one of them on the basis of this false-negative result. This is concerning in consideration of the potential fatal outcome of IC. Therefore, clinical conditions should always be taken into account, in particular in patients with severe sepsis or septic shock and no alternative infectious diagnosis.
The role of the test in a preemptive BDG-driven strategy to start AF therapy is more debated because of the limited PPV (5). On one side, this approach could lead to antifungal drug overconsumption. On the other side, it is well known that the performance of standard cultures for the diagnosis of IC is particularly low in the ICU setting. Blood cultures detect about 20% of IAC cases, and cultures from intra-abdominal samples require invasive procedures, which are often delayed (4). The BDG test may detect the missed cases and allow early start of AF therapy and clearance of the pathogen before cultures turn positive (24). Therefore, we could not assess the impact of BDG when AF therapy was triggered by a positive test without evidence of IC in follow-up. The fact that no IC diagnosis was established for the majority (88%) of patients with a positive BDG test and no further AF therapy suggests that a BDG-driven preemptive approach would result in a significant risk of antifungal overconsumption. However, the PPV of the BDG test increased drastically when the test was performed among the appropriate high-risk subset of patients defined by our algorithm (from 36% in the entire tested population to 80% among the targeted high-risk population). The small number of this subgroup does not allow us to draw firm conclusions, but these results suggest that the performance and utility of the BDG test in ICUs is dependent on the appropriate use of the test.
Education of ICU practitioners for the identification of high-risk patients and implementation of BDG-driven preemptive and empirical strategies are warranted and could serve as a basis for improved antifungal use policies (21). However, these approaches are hampered by several limitations. First, the current BDG testing kit of Fungitell is designed to test multiple samplings in a single microtiter plate. Therefore, patient samples are usually tested per batch (e.g., twice per week), which may delay results. In our study, test results were obtained at a median of two open days from the sampling time, but weekend periods may have prolonged this time interval. Indeed, the median duration of empirical AF therapy before interruption based on negative BDG results was 6 days. Moreover, many institutions do not perform in-house BDG testing and send the samples to a reference lab, which may prolong the turnaround time and alter the feasibility of such a BDG-driven antifungal strategy.
Second, as outlined above, adherence to an algorithm for targeted BDG testing and AF therapy is very difficult to achieve. Because of the frequent turnover of attending physicians in ICUs and the rotation of fellows or residents, it is not always feasible to provide the same instructions to everyone at the same time and to ensure the continuity of this information over time. Antifungal stewardship policies are also difficult to implement in the microbiology laboratory. In this study, an email alert system was set up for all positive BDG tests. However, a real-time monitoring and assessment of all BDG testing requests and/or results by a microbiologist or infectious diseases (ID) specialist, which cannot be substituted by an automatic system, is difficult to maintain in the long term.
In conclusion, despite intrinsic limitations partly due to the small number of cases, this study is among the rare ones to address the proof of concept of the actual impact of the BDG test on therapeutic decisions in ICUs and to consider its utility in two distinct approaches (empirical and preemptive) via a simple algorithm. Our results suggest that the BDG test, when used adequately on a targeted population, can be useful for both approaches. However, under real-life conditions, the indication to perform the test was considered appropriate in only 26% of cases, and BDG results guided therapeutic decisions in only 57% of cases. Inappropriate use of the BDG test may result in low PPV and overconsumption of antifungals, which counterbalance the benefit of antifungal drug sparing associated with the acceptable NPV. Continuous efforts are warranted to implement BDG testing algorithms for guiding antifungal drug prescriptions in ICUs.
ACKNOWLEDGMENTS
We are grateful to Sarah Chappuis for technical assistance with laboratory data extraction. We acknowledge all attending physicians of the intensive care unit of the University Hospital of Lausanne for their collaboration during the study period.
F.L., A.K., and J.P. conceived and designed the study. A.K., J.P., and J.-L.P. acquired the data. A.K. and F.L. analyzed and interpreted the data and drafted the article. A.K., J.P., A.C., P.-Y.B., J.-L.P., and F.L. revised the manuscript. The final version of the manuscript was reviewed and approved by all authors.
We declare no competing interests.
FOOTNOTES
- Received 4 December 2019.
- Returned for modification 7 January 2020.
- Accepted 10 March 2020.
- Accepted manuscript posted online 1 April 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
