Procalcitonin levels in candidemia versus bacteremia: a systematic review

Background Procalcitonin (PCT) is a biomarker used to assess systemic inflammation, infection, and sepsis and to optimize antimicrobial therapies. Its role in the in the differential diagnosis between candidemia and bacteremia is unclear. The aim of this systematic review was to summarize the current evidence about PCT values for differentiating candidemia from bacteremia. Methods PubMed and EMBASE were searched for studies reporting data on the diagnostic performance of serum PCT levels in intensive care unit (ICU) or non-ICU adult patients with candidemia, in comparison to patients with bacteremia. Results We included 16 studies for a total of 45.079 patients and 785 cases of candidemia. Most studies claimed to report data relating to the use of PCT values for differentiating between candidemia and bacteremia in septic patients in the intensive care unit. However, the studies identified were all retrospective, except for one secondary analysis of a prospective dataset, and clinically very heterogeneous and involved different assessment methods. Most studies did show lower PCT values in patients with candidemia compared to bacteremia. However, the evidence supporting this observation is of low quality and the difference seems insufficiently discriminative to guide therapeutic decisions. None of the studies retrieved actually studied guidance of antifungal treatment by PCT. PCT may improve diagnostic performance regarding candidemia when combined with other biomarkers of infection (e.g., beta-d-glucan) but more data is needed. Conclusions PCT should not be used as a standalone tool for the differential diagnosis between candidemia and bacteremia due to limited supporting evidence. Electronic supplementary material The online version of this article (10.1186/s13054-019-2481-y) contains supplementary material, which is available to authorized users.


Background
Early diagnosis of candidemia is challenging [1][2][3]. The absence of sensitive and specific clinical signs and symptoms and radiological findings as well as the prolonged time of blood culture growth hamper early identification of candidemia [2,4]. Adding to this is the need to differentiate between bacterial and fungal infections, which often have similar clinical manifestations. For these reasons, risk factor clinical characteristics, scoring systems, and microbiological techniques (culture-and nonculture-based) are all being used to optimize early treatment and reduce unnecessary antifungal therapy [4][5][6][7][8][9][10][11][12][13].
Procalcitonin (PCT) has been proposed as a useful tool to characterize systemic inflammation, infection, and sepsis [14][15][16]. Findings from several randomized controlled trials indicate that the use of a PCT-guided antibiotic treatment algorithm (i.e., PCT guidance) is likely to reduce antibiotic exposure in septic patients, without an adverse effect on health outcomes [17]. PCT production is promoted by lipopolysaccharides and cytokines, which are expressed in pro-inflammatory conditions [18]. Although some non-bacterial inflammatory conditions increase PCT levels, bacterial infections typically show higher PCT serum concentration [14,18,19]. Some studies reported lower PCT serum levels in patients with candidemia compared to bacteremia [20,21]. Although the mechanism for this finding is unclear, patients with invasive candidiasis showed signs of impaired inflammatory response, immune cell exhaustion, and reduced production of positive co-stimulatory molecules [22][23][24]. Thus, the serum levels of PCT may differ in patients with bacterial and Candida infections [1,20,21]. The aim of this systematic review was to summarize the current evidence about PCT values for differentiating candidemia from bacteremia.

Search strategy and selection process
For the purpose of this review, a search was conducted in PubMed and EMBASE (see Additional files 1 and 2). The terms used included "Candida" OR "fungi" AND "Procalcitonin" (see full search strategy in Additional file 1). We considered only articles published in peer-review journals in the English language. We excluded conference proceedings and case reports.
We selected studies reporting data on the values and diagnostic performance of PCT in intensive care unit (ICU) or non-ICU nonimmunosuppressed adult patients with microbiologically confirmed candidemia in comparison to patients with bacteremia. We also included studies in which data about PCT where reported separately for patients with candidemia from those with other fungal infections. If several samples of PCT were taken, we selected the value of the first available PCT sampled during the diagnostic process.
Two searches were run: the first in 5 October 2018 and the last in 20 February 2019. Two authors (AC, GM) independently screened all titles and abstracts to select potentially relevant papers. Papers selected for full review also underwent screening of their list of references by the same authors to identify additional potential studies of interest. Discrepancies between the two reviewers on relevance at any stage were adjudicated by two other authors (ES, AG). Papers selected for full review underwent data extraction if both reviewers (AC, GM) agreed on their relevance. In case of doubt at any stage, we contacted the corresponding authors of the manuscripts. Figure 1 describes paper inclusion/exclusion process.

Characteristics of the included studies
The searches yielded overall 1175 articles (see Additional files 1 and 2). Among these, 43 were selected for full review but only 16 were ultimately selected for inclusion. These 16 studies included overall 45.079 adults and yielded of 785 cases of candidemia. Of these studies, 10 specifically referred to ICU patients. Twelve of the 16 included studies had at least sepsis as inclusion criteria; three studies did not report this information; in one study, the majority of patients were at least septic, but sepsis was not an inclusion criterion ( Table 1). All studies were retrospective, except for one secondary analysis of a prospectively collected dataset. Table 1 presents data from the included studies, including study design, patient characteristics, microbiological findings, assays used for dosing, and the information given on the diagnostic performance of PCT. Following qualitative synthesis of the data, a decision was made to not to proceed to meta-analysis because of the heterogeneity found in patient populations (study and control groups) and the assays used, as well as the amount of missing data (i.e., large risk of bias). Instead, we hereby summarize the evidence from included studies.

PCT levels for differentiating candidemia from bacteremia Studies in the ICU
In a retrospective cohort study, Charles et al. evaluated 50 non-surgical septic ICU patients with bloodstream infection (BSI). They found significantly lower PCT levels in patients with candidemia (median 0.65 ng/ml [range 0.08-1.56], n = 15) compared to those with bacteremia (median 9.75 ng/ml [range 1.00-259.5]). PCT levels < 5.5 ng/ml had a negative predictive value (NPV) of 100% and a positive predictive value (PPV) of 65% for Candida spp. sepsis [25]. Martini  In cases with candidemia (n = 11), the serum PCT concentration was 2.1 ng/ml (SD 1.8), significantly lower than in Gram-positive or Gram-negative BSI [38].
Thomas-Rüddel et al. performed a secondary analysis of a prospectively collected dataset involving 4858 septic patients with at least one related organ dysfunction from the ICUs of 40 hospitals [40]. PCT values at sepsis onset were analyzed in patients with bacteremia or candidemia but mixed infections were excluded. PCT values were significantly higher in patients with Gram-negative (26 ng/ml [IQR 7.7-63.1]) than Gram-positive bacteremia (7.1 ng/ml [IQR 2.0-23.3]) or candidemia (4.7 ng/ml [IQR 1.9-13.7], n = 63).

Studies in wards or including hospitalized patients
Pieralli et al. retrospectively compared 64 cases with sepsis due to Candida spp. and 128 cases with sepsis due to bacteria in 3 internal medicine wards [36]. PCT [39]. In mixed infections, PCT levels were 2.1 ng/ml (SD 10.0) and 0.1 ng/ml (SD 0.1) for Candida spp. with Grampositive and Gram-negative bacteria, respectively.

PCT use in association with other biomarkers
PCT has been also evaluated in combination with other biomarkers for improving performance in diagnosis of IC [29,35].
Giacobbe et al. retrospectively assessed the combination of PCT and beta-D-glucan (BDG) in 166 critically ill ICU patients for early differentiation between bacteremia and candidemia [35]. Compared to patients with bacteremia, the levels of PCT were lower (median 0.76 vs. 4.32 ng/ml, p < 0.001) and those of BDG were higher (median > 500 vs. < 80 pg/ml, p < 0.001) in patients affected by candidemia. Combining the standard BDG cut-off level (≥ 80 pg/ml) with the rounded optimal PCT cut-off level (< 2 ng/ml) yielded a higher PPV for identifying the presence of candidemia than the PPV of either test alone. Held et al. similarly reported that the combination of BDG and PCT increased specificity (from 89.4 to 96.2%), but this was accompanied by loss of sensitivity (from 86.7 to 51.7%) for candidemia in 56 hospitalized patients [29].
Fu et al. found that the combination of PCT (cut-off 8.06 ng/ml), CRP (cut-off value 116 mg/l), and IL-6 (cutoff 186.5 pg/ml) increased the sensitivity and specificity for early diagnosis of candidemia (n = 23) and its distinction from Gram-positive/negative bacteremia (AUC to 0.912) in 85 ICU septic patients [27]. However, PCT showed the best diagnostic performance, when compared to CRP or IL-6.

Discussion
In this systematic review of the value of PCT for differentiating between candidemia and bacteremia, we found that PCT has been studied in only 785 cases of candidemia. We limited our analysis to adult nonimmunosuppressed patients with bloodstream infections related to Candida spp. to reduce clinical heterogeneity.
Most of the studies identified evaluated the use of PCT for differentiating between candidemia and bacteremia in septic patients in the ICU. We found no study specifically evaluating PCT levels as a tool for monitoring the effect of antifungal treatment.
Although most of these studies showed lower PCT values in patients with candidemia compared to bacteremia, the evidence supporting this observation is of low quality. Moreover, this difference seems to be insufficiently discriminative to guide therapeutic decisions.
PCT may improve diagnostic performance when combined with other biomarkers of infection. Of note, the association with BDG may be of interest due its widespread use and specific role in this setting [2,41]. However, this finding requires additional confirmation.
Our systematic review has several limitations. We could not proceed with meta-analysis because the studies identified were clinically very heterogeneous, involving different assessment methods and comparators. This may limit the impact of our findings but should be mostly seen as a limitation of the available evidence rather than of the review. Another limitation is the inability to separate the results and conclusions according to septic state (e.g., sepsis, septic shock). However, most studies did use sepsis as inclusion criteria or included mostly septic patients (13 out of 16 studies). We were unable to select studies where a surrogate of fungal infection (e.g., beta-D-glucan) was sampled alongside PCT since only one study included such data. The timing of blood sampling for PCT levels varied among the included studies. However, for all studies, we considered the value of the first available PCT sampled during the diagnostic process.

Conclusions
PCT should not be used as a standalone tool for the differential diagnosis between candidemia and bacteremia due to limited supporting evidence. In this setting, PCT values seem to be insufficiently discriminative to guide therapeutic decisions. PCT should be further investigated in antifungal stewardship programs, in association with other biomarkers or non-culture diagnostic tests.