In this large cohort of critically ill patients who had undergone urgent abdominal surgery for intra-abdominal infection, the prevalence of IAC was 44%. The study's findings revealed that a pBDG threshold of 45 pg/ml exhibited an NPV of 82.3% (area under the curve (AUC) of 0.63). When combining peritoneal BDG < 45 pg/ml and low serum BDG with a peritonitis score < 3, the negative predictive value reached an impressive 100%.

No demographic data or Candida risk factors emerged as predictors for IAC in our study, consistent with previous research [7]. Only the nosocomial origin was associated with more IAC occurrences. This could be explained by the high rate of postoperative infection and prior antibiotic exposure in the cohort, two recognized risk factors for IAC [20]. Interestingly, we did not observe a higher rate of supramesocolic origin in the IAC group. This could be attributed to a lower prevalence of gastrointestinal origin (16%) and more colorectal origin. De Ruiter et al. reported high rates of Candida during the initial week of intra-abdominal infection originating from colorectal origin [31], aligning with the timeframe of post-operative peritonitis. Additionally, we noted a balanced distribution of community origin between IAC and non-IAC patients, consistent with previous findings [12, 31, 32]. This underscores the importance of considering community origin when contemplating the initiation of empirical antifungal therapy, particularly in the presence of immunosuppression [19] and/or septic shock [22]. Additionally, because we only included patients with intra-abdominal infection requiring surgery, the prevalence of IAC increased. Indeed, Dupont et al. reported a 30% prevalence of IAC in a population with complicated intra-abdominal infection requiring surgery [11]. In contrast, Nourry et al. included patients with intra-abdominal infection managed by radiology and reported a prevalence of 21% [12]. Recent literature consistently supports the idea of focusing on selected ICU populations when evaluating the diagnostic performance of biomarkers for IAC such as pBDG [1, 9, 21]. Consequently, studying ICU populations that necessitate surgical source control emerges as a promising approach for assessing biomarkers for IAC.

Two studies with an IAC prevalence of 21% reported promising results when measuring pBDG in critically ill patients to rule out IAC [10, 12]. In the first retrospective study including 33 nosocomial secondary peritonitis cases, the authors reported a 98% NPV [10]. Recently, Nourry et al. conducted a prospective study with 113 patients and reported a 100% NPV [12]. In our study using the WT, the NPV of pBDG was notably lower, at 82.3%. First, as the NPV depends on the prevalence of the disease, its value decreases with increasing prevalence. Additionally, it is important to consider that in previous studies, the actual number of patients with IAC whose pBDG concentrations fell below the obtained threshold was low, at 3 patients in the first study and 12 patients in the second study. Furthermore, in the study by Nourry et al., the authors acknowledged that 21 samples had been exposed to antifungal treatment, which could lead to negative fungal culture results and/or reduced yeast quantities [12]. In our study, using pBDG alone with a threshold of 42 pg/ml would have led to nine missed cases. Notably, the mean time to positivity of the Candida culture in these nine cases was four days, suggesting a low inoculum. Thus, the quantity of yeast present might have influenced the diagnostic performance of pBDG.

The pBDG levels were higher in cases of polymicrobial IAC and lower in negative samples, which is consistent with previous findings [10, 12]. Similarly to previous study [4, 31], we reported a high bacterial documentation of 72%. Bacteria, especially gram-negative and enterococci are known to be associated with false-positive results for BDG [33], which could explain the high rate of false-positive peritoneal BDG results observed in our study. With a threshold of ≥ 45 pg/ml, 68 patients would have received unnecessary antifungal treatment.

Thus, we reaffirmed the limited positive predictive performance of both peritoneal and serum BDG. In the ICU setting, numerous confounding factors for sBDG exist, including antibiotics, albumin infusion, and transfusions, among others, which are known to be associated with false-positive results [33]. Additionally, Szyszkowitz et al. highlighted increased levels of sBDG in the peri- and post-operative period, diminishing the significance of its positive value [34]. Conversely, the risk of false-negative values is less likely to occur, particularly in the peritoneal fluid. False negatives are typically caused by a low inoculum or prior exposure to antifungals [35]. In the case of IAC, clinical studies have reported a low rate of antifungal initiation before surgical incision [5], and the peritoneal diffusion of antifungals has been demonstrated to be low (approximately 30%) [36].

The results of our study regarding sBDG differed from previously published data [11]. In the study by Dupont et al., the sBDG determined with the FA showed an AUC of 0.52, P = 0.77, indicating poor diagnostic performance. In our study, both tests demonstrated significantly lower levels of sBDG in non-IAC cases, with reported AUCs of 0.63 and 0.72 for the FA and WT, respectively. Previously, sensitivity values of approximately 70% have been documented for both tests in ICU patients with IAC and no concurrent candidemia [25]. Interestingly, the diagnostic performance of the WT alone and combined with the other markers was superior to that of the FA in our study. Additionally, our study confirmed the superior sensitivity of the WT when using a lower cutoff of 3.3 pg/ml compared to 7 pg/ml, as demonstrated previously in critically ill patients with noncandidemic IAC [25].

While our study has highlighted the limitations of using serum and pBDG and the Peritonitis score as standalone markers for initiating antifungal therapy, it has also emphasized the importance of combining tests to rule out IAC [9, 37]. Specifically, the combination of low pBDG and sBDG with a peritonitis score < 3 demonstrated a sensitivity and NPV of 100% and could be used for discontinuing unnecessary treatment in patients with IAC. In our study, we corroborated the limited performance of the peritonitis score when used on its own [11, 12, 38]. However, its combination with BDG demonstrated the potential to enhance diagnostic efficacy (Additional file 1: Table S6). This assessment tool is not only easy-to-use at the bedside but also cost-effective and globally accessible. Furthermore, the current BDG tests (WT or Fungitell STAT assay®) permit individual patient testing with a swift turnaround time (less than 90 min) [15]. Consequently, an algorithm grounded in the peritonitis score and individual BDG tests could swiftly exclude IAC within two days, averting unnecessary antifungal exposure linked to escalating antifungal resistance and elevated costs [39]. It is noteworthy that this timeframe necessitates further clinical validation in real-life ICU conditions.

The major strengths of our study include its sample size, multicenter nature, and high number of consecutively included patients with pBDG results, allowing for a high prevalence of IAC. Certainly, there are some limitations to consider. Firstly, there is a risk of misclassification between IAC and non-IAC due to the sensitivity of the peritoneal culture. However, all participating centers followed a local protocol that recommended direct inoculation of the peritoneal sample into a favorable culture medium and storage of all cultures for up to 8 days to detect delayed positivity, which helps mitigate this risk to some extent. Secondly, the COVID-19 outbreak affected our study, and we had to extend the inclusion period. However, the management of IAC remained consistent during this time, which should minimize any potential impact on our results. Thirdly, the peritonitis score was the exclusive scoring system employed in this non-interventional study, aligning with the standard practice across all participating centers and in accordance with the French guidelines [19]. Consequently, the assessment of alternative Candida risk factor scores was precluded due to the prevailing routine in the involved centers.

Last, our study did not evaluate patient survival but rather focused on surrogate markers to optimize the detection of Candida in peritoneal fluid. The role of Candida as a true pathogen in IAC is still debated [40, 41], and the effectiveness of antifungal treatment in IAC remains inconclusive in previous studies [4, 42, 43].