FormalPara Take-home message

In this randomized multicenter clinical trial on sepsis patients we observed an earlier but also immoderate administration of antifungals when therapy was guided by (1 → 3)-β-d-glucan-guidance in comparison to culture guidance. (1 → 3)-β-d-glucan-guidance did not affect mortality but definite conclusions are hampered by an unexpectedly low rate of invasive Candida infections in this study population.

Introduction

Sepsis is a leading cause of death in critically ill patients and is defined as a life-threatening organ dysfunction caused by a dysregulated host-response due to infection [1]. Incidence of invasive Candida infection (ICI) is increasing [2, 3]. Mortality rates reached 80% in patients with candidemia if no antifungal treatment is started within the first 24 h of septic shock [4].

Current guidelines suggest early antifungal treatment in critically ill patients with a high risk of ICI but give only little advice on the selection of appropriate patients [5,6,7]. (1 → 3)-β-d-Glucan (BDG) may be suitable to guide antifungal therapy as BDG serum concentrations are significantly elevated in patients with ICI [8]. BDG is a cell wall constituent of Candida spp. Previous trials suggested that BDG guided antifungal therapy is safe [9], that BDG serum concentrations may aid to stop antifungals [10, 11], and that BDG might select patients benefiting from empirical antifungal treatment [12]. Current guidelines are cautious to recommend BDG guidance to trigger antifungal treatment because of limited data [6, 7, 13]. We designed the CandiSep trial to test whether BDG-guidance shortens time to antifungal therapy and thereby reduces mortality of sepsis patients with a high risk of ICI.

Methods

Trial design

The CandiSep trial was an open, randomized, multicenter trial comparing a BDG-guided antifungal therapy versus standard of care in patients with sepsis and high risk for developing ICI conducted in 18 German intensive care units (eTable 1). The protocol has been published previously [14]. The ethics board of the Jena University Hospital, the corresponding institutional review boards of all participating centers and the Federal Institute for Drugs and Medical Devices approved the trial. The trial was registered at ClinicalTrials.gov (NCT02734550). Written informed consent was obtained from all patients or their legal representatives. The ethics committees approved a deferred consent for those patients unable to give informed consent but not having appointed a legal representative. As soon as the legal representative of the patient was available, written informed consent was obtained; otherwise, the patient was withdrawn from the study. Trial execution was overseen by an independent data and safety monitoring board.

Patient selection and randomization

Between September 2016 and September 2019, we included adult patients with sepsis and increased risk for ICI. The allowed time window between onset of sepsis and enrollment was 12 h but was extended to 24 h as of January 26, 2018 because of insufficient recruitment. Sepsis definitions were applied as reported previously [15]. However, in the light of the Sepsis-3 definitions under development when this study was designed [1], Systemic Inflammatory Response Syndrome was not required for sepsis diagnosis. Risk for ICI was defined as total parenteral nutrition, abdominal surgery within the last 7 days, previous antimicrobial therapy for more than 48 h, and previous renal replacement therapy [16]. We excluded patients with proven ICI, present or planned systemic antifungal therapy prior to inclusion, recent surgery with cardiopulmonary bypass, recent treatment with immunoglobulins, and immunosuppression. A complete list of inclusion and exclusion criteria is provided in the Supplement. After enrollment, the patients were randomized to either a BDG- or control group in a 1:1 ratio via a web-based central randomization service. The randomization list was prepared by an independent statistician via a computer-based algorithm and was stratified by study center.

Trial treatments

BDG-samples together with concomitant blood cultures were taken within 1 h after randomization and 24 h after enrollment. In the BDG-group, samples were immediately forwarded to the central laboratory. Site specific courier schedules assured that the first BDG result was available not later than 24 h after sampling. The treating physicians were informed about the BDG results via telephone and fax. Any patient with a BDG concentration of ≥ 80 pg/mL had to be treated with antifungals suitable for ICI. Antifungal treatment of patients with only one BDG concentrations ≥ 80 pg/mL was discontinued in case of blood cultures negative for Candida spp. irrespective which BDG sample was positive. Patients with BDG concentrations ≥ 80 pg/mL in both samples remained on antifungals irrespective of the blood culture results. In the control group, BDG was measured in batches. The treating physicians were blinded to the BDG results and patients were treated according to standard of care. In both groups, targeted treatment of ICI was performed according to European guidelines [17], and patients with microbiological growth of Candida spp. in any blood culture, biopsy or sample from normally sterile body fluids were treated with antifungals independent of the BDG result. Empirical antifungal therapy was discouraged in both groups.

Measurements

Blood samples were handled according to standardized procedures [14]. Briefly, serum samples for BDG-measurement were obtained within the first hour after randomization and after 24 h using BDG-free collection tubes. BDG serum concentrations were measured in a central laboratory at the Microbiology Institute of the University Hospital Erlangen, Germany using the Fungitell® assay (Associates of Cape Cod Inc., East Falmouth, MA, USA) according to the manufacturer’s instructions [14]. The Fungitell® assay is a Limulus amebocyte lysate-based test with a detection-limit at 30 pg/mL and coefficient of variation of 9.6%. Aerobic and anaerobic blood cultures were obtained by sterile venipuncture no more than 3 h after randomization if no blood cultures had already been taken up to 6 h before randomization. Additional blood samples with tubes containing ethylenediaminetetraacetic acid (EDTA) were obtained for Candida detection via polymerase chain reaction (Candida-PCR) via the same venipuncture. Microbiological samples were taken from nose/throat, skin (axillar region), rectum/feces, urine, and tracheal or bronchial secretion to determine the Candida Colonization Index (CCI) [18]. Blood cultures were repeated on the day after randomization and microbiological samples for the CCI were repeatedly obtained on day 7 and day 14 if the patient was still on the ICU. Candidemia was defined as growth of Candida spp. in at least one blood culture and ICI was defined as any microbiological proof of Candida spp. in primary sterile specimens [19]. The PCR was performed in batches in the National Reference Center for Invasive Mycoses (Jena, Germany). Candida spp. were detected by a semi-nested PCR assay amplifying the internal transcribed spacer region 2 [14]. PCR results were not reported to the treating physicians.

Outcomes

The primary outcome of the study was all-cause mortality by 28 days after inclusion. Secondary outcomes were antifungal-free survival within 28 days after inclusion, Candida colonization according to the Candida Colonization Index (CCI), time to antifungal therapy, costs of antifungal therapy, duration of organ support (including ventilation, vasopressor, and renal replacement therapy) until day 28, mean total score for Sequential Organ Failure Assessment (SOFA) calculated as the sum of daily SOFA scores divided by the number of study days on intensive care unit (ICU), ICU and hospital length of stay, ICU and hospital mortality, and frequency of adverse events.

Statistical analysis

Complete details regarding the sample size justification and statistical analysis have been reported previously [14]. Sample size was based on an estimated rate of the primary end point of 49.8% in the control group and 34.2% in the BDG group. A sample size of 156 patients per group would provide more than 80% power to show the superiority of BDG-guidance over standard of care using a two-by-two chi-square test at a two-sided alpha level of 0.05. Studies with a similar patient population have demonstrated a 10% drop-out rate [20, 21]. We therefore expected to randomize 348 patients to achieve 312 evaluable patients. An interim analysis was performed after one year of recruitment, with data available only to safety monitoring board. Based on this analysis, the data and safety monitoring board recommended to continue the trial as planned.

Study objectives were analyzed in the intention-to-treat population (ITT), which consisted of all randomized patients with informed consent. The primary objective was analyzed by using the chi-square test. Kaplan–Meier estimates and log rank tests were also applied. Twenty-eight day mortality was also analyzed in a per-protocol analysis and for the following pre-defined subgroups: septic shock at randomization, CCI ≥ 0.5, Candida-PCR positive, blood culture positive for Candida spp., BDG serum concentration ≥ 80 pg/mL in both samples, ≥ 2 risk factors for ICI. Patients were excluded for the per-protocol population if antifungal therapy deviated from the recommendation based on the BDG results (BDG-group), and if the patients received systemic antifungal therapy in absence of proven ICI (control group). Secondary objectives were analyzed according to their scales with the Chi-square or Fisher’s ex!ct test and Mann–Whitney-U-test, respectively.

Results

Patients

Patients were recruited from September 2016 to September 2019. We identified 2324 patients eligible of whom 342 underwent randomization. Seventy-eight patients were enrolled before the amendment. Three patients withdrew informed consent resulting in 339 patients, of whom 172 patients in the BDG group and 167 patients in the control group were included in the analysis of the primary endpoint (Fig. 1). Baseline characteristics were well balanced between the two groups (Table 1) with recent abdominal surgery being the most frequent ICI risk factor. Initial blood cultures were positive for Candida spp. in 15 patients (4.4%). ICI was diagnosed in 48 patients (14.2%) within the first 96 h after enrollment (see eTable 2 for characteristics of patients with ICI). The most frequent pathogen was Candida albicans (see eTable 3 for characteristics of ICI). Six patients (1.8%) were positive for Candida spp. in the PCR but only three of these patients had microbiological proof of Candida spp. and were considered having ICI. Sensitivity and specificity of two positive BDG results to predict candidemia were 64.3% and 63.7% and to predict ICI diagnosis within the first 96 h were 54.4% and 65.2%, respectively (see eTable 4 for association of BDG results and ICI). The per-protocol population consisted of 101 patients in the BDG- and 127 patients in the control-group (see eTable 5 for characteristics of patients).

Fig. 1
figure 1

Participant flow in the CandiSep-trial. Categories of screening failures were not mutually exclusive; participants could have more than one reason. BDG group: (1 → 3)-β-d-glucan-guided group

Full size image
Table 1 Characteristics of the patients at baseline
Full size table

Trial regimens

In the BDG-group, 84 of 172 (48.8%) patients received antifungal therapy, 74 of 172 (43%) patients because of elevated BDG concentrations, 3 (1.7%) patients because of microbiological findings and 7 (4.1%) patients were treated empirically. During the same period, 10 out of 167 (6%) patients in the control group received empirical antifungal treatment, 9 (5.4%) patients because of microbiologically proven candidiasis, and one (0.6%) patient without specified reason. Protocol deviations are listed in the Supplement.

Primary outcome

After 28 days of randomization, 58 of 172 (33.7%) patients assigned to the BDG group and 51 of 167 (30.5%) patients assigned to the control group had died [relative risk, 1.10; 95% confidence interval (CI) 0.80–1.51; p = 0.53; Table 2]. There was no significant difference between the groups concerning the rate of death in the time-to-event-analysis during the 28 days after randomization (Fig. 2). The analysis of pre-specified subgroups yielded similar results with no subgroup showing a benefit of BDG-guidance. In the per protocol analysis, 24 out of 96 (25.0%) subjects died until day 28 in the BDG group and 41 out of 126 (32.5%, p = 0.22) patients in the control group (eFigure 1). In an un-prespecified post-hoc analysis on patients having ICI, 12 of 25 (48%) patients of the BDG group and 9 of 23 patients (39.1%) of the control group died until day 28 (relative risk, 1.23; 95% CI, 0.64–2.36, p = 0.54).

Table 2 Primary and secondary outcomes
Full size table
Fig. 2
figure 2

Rate of survival and Risk of Death at day 28, according to subgroups. Upper panel shows Kaplan–Meier estimates of the survival rate according to the (1 → 3)-β-d-glucan-guided (BDG) and the control-group. The p value of 0.55 was calculated with the log-rank-test. Lower panel shows the relative risk (RR) of death at day 28 in the prespecified subgroups. The size of the square representing the relative risk reflects the relative numbers in each subgroup, and horizontal bars represent 95% confidence intervals

Full size image

Secondary outcomes

Secondary outcomes are shown in Table 2. Patients of the BDG group received antifungals within a median time of 1.1 [interquartile range (IQR) 1.0–2.2] days while time to antifungal therapy was 4.4 (IQR: 2.0–9.1, p < 0.01) days in the control group. In total, there was a higher frequency of antifungal therapy in the BDG group (99 of 172 patients, 57.6%) compared to the control group (46 of 167 patients, 27.5%; Fig. 3). Median duration of antifungal therapy was 10.5 (IQR: 5.6–14.7) days in the BDG group versus 10.0 (IQR 4.2–15.4, p = 0.69) days in the control group. Antifungal free survival at day 28 was significantly lower in the BDG group than in the control group (30.2% vs. 52.1%, p < 0.01) with an odds ratio of 2.97 (95% CI 2.1–4.2). This was associated with higher per patient costs of antifungal therapy in the BDG (4451 €, IQR 1385–6923) compared to the control group (2800 €, IQR 989–7097) but this difference did not reach statistical significance (p = 0.52). Candida colonization was generally low in the study population (median CCI: 0.2 in both groups) and did not change throughout the trial. There was no overt safety issue with BDG-guidance (see Supplement).

Fig. 3
figure 3

Rate of patients not on antifungal therapy. The figure shows Kaplan–Meier estimates of the rate of patients not on antifungal therapy according to the (1 → 3)-β-d-glucan-guided (BDG) and the control-group. The p value of < 0.01 was calculated with the Log-rank-test

Full size image

Discussion

In this multicenter randomized trial, BDG guidance in critically ill patients with sepsis or septic shock and at high risk for ICI resulted in a more frequent and earlier initiation of antifungal therapy but did not improve 28-day mortality. None of the pre-defined subgroups showed a benefit from the BDG guided early antifungal therapy. However, we observed a higher antifungal free survival at day 28 in the control group compared to the BDG-group. Thus, BDG-guided patients were more likely to receive antifungal therapy and thereby might generate higher costs for antifungal therapy.

International guidelines [5,6,7] recommend antifungal therapy in critically ill patients without proven Candida infection but with inherent risk factors. This recommendation is based on the high risk of developing ICI [18, 22] and the high mortality rate in case of untreated ICI [4] in this patient population. Some studies suggested a benefit for risk-based antifungal therapy [23, 24] while other randomized controlled trials did not show a benefit for such an approach [12, 25,26,27]. Our study now demonstrates that a BDG-guided initiation of early antifungal therapy was not superior to a wait-and-see culture-based approach in a sepsis population with a high risk of ICI.

Specificity of BDG was low in our study. Various factors have been discussed for triggering false positive BDG results including administration of antimicrobials [28], blood products including albumin [29], and immunoglobulins [30], all of which may be applied in the care of sepsis patients. However, significance of these factors is not undisputed [31]. Previous abdominal surgery was the most prevalent risk factor predisposing for ICI in our trial. However, open gut surgery results in elevated serum BDG concentrations for up to 5 days which might have impacted our BDG-measurements [32]. Although BDG was eliminated within 4 days in pediatric patients [33], animal experiments described a BDG-clearance of 9 days [34]. The most recent Cochrane analysis concluded that it remains unclear whether BDG can identify ICI early [35]. This lack of specificity likely caused an overuse of antifungals in the BDG-group as many patients without microbiological proof of ICI were treated without having a benefit. We aimed to improve BDG-specificity by excluding patients treated with immunoglobulins [30] or cardiopulmonary bypass surgery [36] and by requiring two positive BDG measurements from two consecutive days [37, 38]. Although a BDG cut-off of 80 pg/mL has been successfully used to diagnose candidemia and intraabdominal candidiasis [8, 39, 40], a higher cut-off such as 200 pg/mL has been suggested to increase specificity [41]. Sensitivity was only 54.3% in our study. A sensitivity as low as 20% has been observed in several studies [35] and may be associated with a low pathogen load [39]. Current data rather suggest that BDG can be used as a surrogate for withdrawal of empirically started antifungal treatment rather than early initiation [10, 11, 42].

Our sample size calculation was based on a mortality rate of 49.8% considering the high-risk profile of this patient population [14]. However, the mortality rate in the control group was 30.5% and, thereby, lower than expected. This might be partly explained by the high frequency of C. albicans-infections being associated with a lower mortality than infections with C. glabrata or C. krusei [43]. In addition, our inclusion criteria aimed for a patient population with a high risk for ICI, but surprisingly, we only observed fifteen cases of proven candidemia, 48 patients with ICI and a low Candida colonization rate. A low rate of Candida colonization is associated with lower risk for developing ICI [44]. This was accompanied by a low positivity rate of the Candida-PCR. Therefore, the inclusion criteria did not result in a pre-selection of a population with high risk of ICI, and BDG may thus not unfold its full diagnostic benefit. Interestingly, this observation does not align with studies with a similar patient profile but a higher proportion of patients with candidemia [11, 40, 45]. Prediction of ICI solely based on risk factors may be too inaccurate to select patients for antifungal therapy.

Strengths and Limitations

The strengths of our trial include the randomized, controlled design, and its multicenter approach. The patients were efficiently enrolled shortly after onset of sepsis. Randomization resulted in an equal distribution of possible confounders between the two groups. Our trial achieved a sufficient separation between the groups regarding timing and frequency of antifungal therapy. Nevertheless, the trial also had limitations. Quality of BDG measurement required a central lab. Compared to in-house measurement, this approach resulted in a delay of measurement and reporting despite an established express delivery service. We increased the time window for recruitment from 12 to 24 h to facilitate sufficient enrollment. Both factors possibly inflicted a delay in antifungal therapy to the disadvantage of the BDG group. The strict time window and the exclusion criteria caused an enrolment rate of 14.7% of the screened patients. Such a low rate might affect the generalizability of the trial. We had a significant number of protocol deviations mainly addressing the duration of the antifungal therapy in the BDG group. The lower mortality in the BDG-group compared to the control group in the per-protocol analysis suggests that an improvement in algorithm adherence might have impacted the outcome. However, a per-protocol-analysis does not maintain internal validity as obtained by randomization and adequate sample size. Frequency of ICI was unexpectedly low. BDG-guidance might perform better in a study population with a higher rate of ICI. The investigated intervention did not allow for blinding. Therefore, this trial may be affected by performance and detection bias.

Conclusion

Early administration of antifungal therapy guided by serum BDG did not provide an advantage over standard of care as measured by 28-day mortality among critically ill patients with sepsis. These data do not support BDG guidance for initiation of antifungal therapy in this patient population if the observed rate of ICI is reasonably low. A possible benefit of serial BDG measurement in a more selected population at risk needs to be addressed in future studies.