Monitoring Procalcitonin in Febrile Neutropenia: What Is Its Utility for Initial Diagnosis of Infection and Reassessment in Persistent Fever?

Background Management of febrile neutropenic episodes (FE) is challenged by lacking microbiological and clinical documentation of infection. We aimed at evaluating the utility of monitoring blood procalcitonin (PCT) in FE for initial diagnosis of infection and reassessment in persistent fever. Methods PCT kinetics was prospectively monitored in 194 consecutive FE (1771 blood samples): 65 microbiologically documented infections (MDI, 33.5%; 49 due to non-coagulase-negative staphylococci, non-CNS), 68 clinically documented infections (CDI, 35%; 39 deep-seated), and 61 fever of unexplained origin (FUO, 31.5%). Results At fever onset median PCT was 190 pg/mL (range 30–26'800), without significant difference among MDI, CDI and FUO. PCT peak occurred on day 2 after onset of fever: non-CNS-MDI/deep-seated-CDI (656, 80–86350) vs. FUO (205, 33–771; p<0.001). PCT >500 pg/mL distinguished non-CNS-MDI/deep-seated-CDI from FUO with 56% sensitivity and 90% specificity. PCT was >500 pg/ml in only 10% of FUO (688, 570–771). A PCT peak >500 pg/mL (1196, 524–11950) occurred beyond 3 days of persistent fever in 17/21 (81%) invasive fungal diseases (IFD). This late PCT peak identified IFD with 81% sensitivity and 57% specificity and preceded diagnosis according to EORTC-MSG criteria in 41% of cases. In IFD responding to therapy, median days to PCT <500 pg/mL and defervescence were 5 (1–23) vs. 10 (3–22; p = 0.026), respectively. Conclusion While procalcitonin is not useful for diagnosis of infection at onset of neutropenic fever, it may help to distinguish a minority of potentially severe infections among FUOs on day 2 after onset of fever. In persistent fever monitoring procalcitonin contributes to early diagnosis and follow-up of invasive mycoses.


Introduction
Febrile neutropenia is a frequent complication in patients with hematological malignancies requiring prompt empirical broadspectrum antibacterial therapy. Cultures and clinical examination fail to detect a pathogen and/or an infectious focus in 30-60% of episodes, which are classified as fever of unknown origin (FUO) [1]. As infectious and non-infectious etiologies of FUO cannot be differentiated, all patients receive maybe unnecessarily prolonged empirical antibacterial therapy. Persistence of fever despite ongoing antibiotics may suggest uncontrolled and potentially lifethreatening infection due to resistant bacteria or fungi. Unresolved fever represents a major challenge for diagnosis and therapy, often resulting in multiple investigations and empirical modifications of antibacterial and/or addition of antifungal therapy [1]. New diagnostic tools are thus needed for guiding management of febrile neutropenic patients when initial microbiological and clinical documentation is lacking and in persistent fever, when resistant bacterial infection or invasive fungal disease (IFD) is suspected.
Procalcitonin (PCT) is a prohormone of calcitonin produced during systemic infection in response to circulating microbial toxins and host inflammatory mediators [2,3]. Previous studies in non-neutropenic patients have reported the utility of PCT for differentiating bacterial from viral infections or non-infectious inflammatory conditions [2,3,4,5,6]. PCT concentrations thresholds have been proposed for distinguishing systemic from localized infections and infections due to virulent pathogens from those due to pathogens with low virulence (e.g. coagulase-negative staphylococci, CNS). Christ-Crain et al have shown that a PCT-based management algorithm reduces antibiotic use in lower respiratory tract infections [7,8]. A high variability of the diagnostic performance of PCT has been reported in febrile neutropenic patients, in whom the utility of this blood marker remains unclear [9]. Few data are available on PCT in persistent neutropenic fever and suspected IFD. We aimed at prospectively monitoring the PCT kinetics in febrile neutropenic patients for the initial assessment of the etiology of fever and the diagnostic reassessment in persistent fever.

Patients
This study was conducted in the Isolation Ward of the Infectious Diseases Service of a 900-bed University Hospital during an 18-month period. Consecutive adult patients undergoing induction or consolidation myelosuppressive chemotherapy for acute leukemia, or intensive chemotherapy followed by autologous hematopoietic peripheral stem cell transplantation were enrolled after written informed-consent. All patients were hospitalized in single-bed isolation rooms with positive pressure and HEPAfiltered ventilation. The protocol was reviewed and accepted by the local HREC (Commission d'éthique de la recherche clinique, Faculté de Médecine et de Biologie, Université de Lausanne, approval number 179/01 on the 5th November 2001). All patients gave written informed consent prior to study enrolment.

Febrile episodes
Neutropenia was defined as a neutrophil count ,500/mm 3 . A febrile neutropenic episode (FE) was defined as an axillary temperature .38.5uC once or $38uC on two or more occasions (at least one hour apart) during a 12-h period [1]. A new febrile episode was defined as a relapsing fever after more than 72 hours of apyrexia (,38uC). For each FE the initial diagnostic work-up consisted of a thorough physical examination, two sets of aerobic and anaerobic blood cultures drawn by phlebotomy and via each line of the central venous catheter, urine culture, and chest X-ray. Additional specific investigations were performed according to the clinical presentation (e.g.: PCR for HSV on oropharyngeal swabs in patients with severe mucositis, screening of respiratory viruses with specific PCRs in patients with suggestive symptoms/signs). In persistent fever during more than 3 days, the diagnostic reassessment included repeated blood cultures as described above, 2-weekly measurement of galactomannan antigenemia (results were not available in real-time for clinical decisions, and only used for diagnostic classification of IFD: see next paragraph), a thoracoabdominal CT-scan, a bronchoscopy with BAL in presence of a lung infiltrate, and biopsy for culture and histopathology at any clinically suspected site of infection.

Classification of febrile episodes
Based on clinical, microbiological and radiological data and blinded of PCT results, the etiology of fever was classified on day 2 after onset of fever according to standard definitions of the International Immunocompromised Host Society (ICHS) and to the PCT manufacturer's recommendations derived from previous PCT studies: i) microbiologically documented infection with (MDIB) or without (MDInB) bacteremia [all MDI were differentiated in those due to coagulase-negative staphylococci (CNS) and those due to other pathogens (non-CNS)] [9,10,11], ii) clinically documented infection [CDI, subdivided into superficial (upper respiratory tract infections, mucositis grade 3, skin and soft tissue infections, lower urinary tract infections) and deep-seated infections (lower respiratory tract infections, enterocolitis, mucositis grade 4, hepato-splenic infection, arthritis, osteomyelitis, upper urinary tract infection, and CNS infection) ] [9,12,13,14], iii) and fever of unexplained origin (FUO) [15]. Standard definitions of the focus/i of infection were used in CDI [16]. Mucositis with a WHO score .2 [17] and diarrhea with a frequency.86/day were considered as CDI [18,19] Diagnosis of proven, probable or possible invasive fungal disease (IFD) was based on the definitions of the European Organization for the Research and Treatment of Cancer-Mycoses Study Group (EORTC-MSG) [20]. Galactomannan (Platelia, Biorad, France) and beta-glucan (colorimetric assay, Wako-Maruha, Japan [21]) antigenemias were measured on the same days as the PCT according to the recommendations of the manufacturers. IFD were included in the classification of infection as MDIB (fungemia), MDINB (tissue invasion documented by culture) or CDI (probable or possible IFD without microbiological evidence).

Antimicrobial therapy
Broad-spectrum beta-lactam antibiotic monotherapy was started within 1 hour after onset of fever and adjusted after 48 hours according to microbiological results and clinical course, as recommended [1]. If a patient remained febrile during more than 3 days despite appropriate antibacterial therapy and/or presented signs of IFD, antifungal treatment was added as recommended [1]. Clinical, laboratory, radiological, microbiological and histopathological data and information on antimicrobial therapy were prospectively collected. Ongoing antimicrobial therapy was defined as inappropriate if a microbiologically documented (e.g. positive culture or PCR) or presumed (e.g. probable or possible IFD based on EORTC-MSG criteria) pathogen was not covered [22]. Antibacterial and antifungal prophylaxis were not used. Antiviral prophylaxis (valacyclovir) was administered to patients with positive HSV serology during the neutropenic period.

Procalcitonin measurements
For PCT measurements, 5 ml-blood samples (Li-heparinate, MonovetteH, Sarstedt, Numbrecht, Germany) were collected for each FE at the same time as blood cultures at onset of fever (day 0) and then on day 1, 2, 3, 4, and 7. Follow up after day 7 included twice-weekly blood sampling until occurrence of a subsequent FE (measurements according to the above sequence) or discharge (end of study). The same screening strategy was applied regardless of the resolution or the persistence of fever. Tubes were centrifuged during 10 minutes at 3000 rpm.min 21 and stored at 280uC until processed. PCT was measured by an amplified cryptate emission technology assay (Kryptor PCT, BRAHMS Diagnostica, Hennigsdorf bei Berlin, Germany) with a functional sensitivity of 60 pg/ml (corresponding to 0.06 ng/ml) and an analytical range of 60 pg/ml to 5000 pg/ml (i.e. 0.06 to 5 ng/ml). The test's intraand inter-assay coefficient of variation was below 10% [23] Measurements of PCT were performed in batch at the end of study and were not available for patients management, nor for classification of the etiology of fever and localization of infection, as described above.

Analyses of PCT results for initial diagnostic assessment of neutropenic fever
Daily PCT values over 7 days after onset of fever were compared among different etiologies of fever, as defined above. The time point with the highest PCT value (i.e. peak value) was considered for analyses of the performance of PCT cut-offs ranging 100-2000 pg/mL for distinguishing etiologies of fever in receiver operating characteristic (ROC) curves and by calculation of sensitivity, specificity, positive and negative predictive values and likelihood ratios, and efficiency. The best PCT cut-off was defined on the basis of the highest calculated efficiency.
Analyses of PCT results for diagnostic reassessment in persistent neutropenic fever PCT kinetics was analyzed in patients with persistent fever during more than 3 days focusing on the assessment of the appropriateness of ongoing antimicrobial therapy and the diagnosis of proven, probable or possible IFD.

PCT for initial diagnostic assessment of neutropenic fever
The PCT kinetics over 7 days in FE of different etiologies is shown in Figure 1 The PCT peak was observed at a median of 2 days after the onset of fever (range 0-20). When compared with PCT values on day 2 in FUO (median 204, range 33-771), those in non-CNS MDIB (2155 pg/ml, range 80-86350; p,0.001), non-CNS MDInB (336, range 117-12090; p = 0.007) and deep-seated CDI (median 520, range 107-13000; p,0.001) were significantly higher. However, no significant difference in PCT values on day 2 was observed in CNS MDIB/MDInB and superficial CDI when compared with FUO ( Figure 2). No difference in PCT values in the various etiologies of fever was observed by stratifying for the type of chemotherapy regimen, i.e. induction/consolidation chemotherapy for acute leukemia vs. conditioning chemotherapy followed by autologous hematopoietic stem cell transplantation (data not shown).
PCT cut-offs ranging between 100 and 2000 pg/mL were analyzed on day 2 after the onset of fever. Proportions of PCT values higher than above cut-offs in different FE etiologies are reported in Table 2. The performances of these PCT cut-offs for discriminating different subsets of documented infections from FUO were calculated. The areas under the ROC curves were 0.695 (all documented infections vs. FUO), 0.803 (non-CNS MDI and deep-seated CDI vs. FUO), and 0.860 (non-CNS MDIB vs. FUO), respectively ( Figure 3). The best diagnostic efficiency was observed with a PCT cut-off of 500 pg/mL ( Table 3). Efficiency of PCT was not different in first vs. recurrent FE, in those occurring with and without ongoing antibiotic therapy at time of onset of fever (data not shown).
The patients with IFD received an antifungal therapy within a median of 4 days (range 0-10) after onset of fever: all survived and were afebrile at time of discharge. The proportions of patients with PCT higher than 500 pg/mL and with fever during the course of IFD are shown in Figure 5. The PCT decrease below 500 pg/ mL preceded the resolution of fever: median 5 days (range 1-23) vs. 10 days (range 3-22), respectively (p = 0.026).

Discussion
The kinetics of PCT was prospectively studied in a large cohort of febrile neutropenic patients with haematological   malignancies at high risk for severe bacterial and fungal infections. A close monitoring of this blood marker at different time points during the course of a febrile episode and a thorough analysis of a broad range of PCT cut-offs provided reliable information on its utility for the initial assessment of the etiology of fever and for the diagnostic reassessment in persistent fever. At onset of fever, PCT was low to undetectable (i.e. below the threshold of positivity), although the values in non-CNS MDIB were significantly higher than those in all other etiologies of fever. A PCT peak was observed in non-CNS MDI and deep-seated CDI 2 days after the onset of fever, i.e. the time at which a reevaluation of the empirical antibacterial therapy based on clinical course and microbiological results is recommended. The PCT threshold of 500 pg/mL allowed the best discrimination of these severe infections from infections due to CNS, superficial infections or FUO. Interestingly, PCT values higher than this threshold were observed in a minority of patients with FUO, suggesting that these FE may have been due to deep-seated bacterial infections with lacking clinical and microbiological documentation. However, this could not be demonstrated in these few patients in whom fever was the only clinical sign suggesting infection and who had received empirical antibacterial therapy with prompt defervescence.
Monitoring PCT for the diagnostic reassessment in patients with persistent neutropenic fever lasting more than 3 days revealed that a delayed PCT peak higher than 500 pg/mL was more frequently observed in patients receiving inappropriate antimicrobial therapy. This finding was true in the majority of cases of IFD and often Table 3. Diagnostic performance of PCT at the 500 pg/mL cut-off on day 2 after onset of fever in i) documented infections (MDIB, MDInB and CDI) vs. FUO, ii) non-CNS MDIB, non-CNS MDInB and deep-seated CDI vs. FUO, and iii) non-CNS MDIB vs. FUO.  preceded diagnosis of IFD based on EORTC-MSG criteria. In follow-up PCT decline reflecting the response of IFD to antifungal therapy significantly anticipated defervescence. Previous investigations in febrile neutropenic patients have reported controversial results regarding the performance of PCT for the initial assessment of the etiology of fever by distinguishing bacteremia due to pathogens other than CNS from nonbacteremic infections, and systemic from localized infections or FUO [9,10,11,12,24,25,26,27,28,29,30].In the present study, the low PCT values observed at the time of onset of fever in severe documented infections do not support the use of PCT for the decision to start or withhold prompt empirical antibacterial therapy. This finding differs from previous reports in which the timing of the first PCT measurement was variable [9,12,26,28,29], i.e. it corresponded either to the onset of fever, the onset of antimicrobial therapy [11] or the time of blood cultures [10]. Moreover, different types of neutropenic populations (i.e. hematological vs solid tumors, adults vs children) and delayed PCT measurements in outpatients vs inpatients may be responsible of the heterogeneity of the results in some studies. On the other hand, the lack of a reliable diagnostic tool for distinguishing infections from non-infectious causes of fever when microbiological and clinical documentation is lacking, i.e. in FUO, remains a major issue in management of febrile neutropenic patients. The utility of PCT for identifying non-documented bacterial infections in patients with FUO has been suggested by Giamarellos-Bourboulis et al. who reported higher PCT values in those responding to antibacterial therapy when compared with those without response [12]. In the present study, PCT values similar to those observed in deep-seated CDI were measured on day 2 after onset of fever in a small subset of patients with FUO, which might help to identify febrile episodes due to occult severe bacterial infections.

Etiology of fever
New findings are provided on the role of PCT for the diagnostic reassessment in patients with persistent neutropenic fever and suspected IFD. Few data have been reported in this specific setting, which represents a major challenge for the clinical management. Small case series have described controversial results on the utility of PCT for the diagnosis of IFD [31,32,33,34]. Some did not report any increase of PCT concentrations in life-threatening IFD [31,32,33], while others associated elevated PCT values with poor outcome in IFD [34]. In 5 cases of invasive aspergillosis in allogeneic hematopoietic stem cell transplant recipients, PCT values exceeding 3000 pg/mL have been described [35,36]. In contrast with the paucity of available data in the literature, sequential PCT measurements in a large number of episodes of persistent fever (n = 63), including the highest number of IFD reported in a PCT study (n = 25) are analyzed in the present investigation. The results suggest that a delayed PCT peak higher than 500 pg/mL observed beyond 3  days of persistent fever is helpful for the early diagnosis of IFD. This blood marker could be combined with the conventional radiological and microbiological tools for improving and anticipating the detection of these life-threatening infections. Moreover, a PCT decrease to normal levels significantly preceded the resolution of fever in follow-up of IFD responding to antifungal therapy. Monitoring PCT may thus help to avoid unnecessary investigations and modifications of antimicrobial therapy in IFD with prolonged fever.
In contrast with previous studies, which have shown that persistently elevated or increasing PCT values predict the outcome in neutropenic patients with severe sepsis or septic shock, a very low rate of severe septic complications was observed in the present investigation. Prompt and appropriate empirical antibacterial therapy in an epidemiological setting of low bacterial resistance has probably influenced the PCT kinetics in the majority of potentially life-threatening infections, thus precluding the evaluation of PCT as a predictor of infection-related severe morbidity and mortality [9,12,37,38,39]. The lack of data in allogeneic hematopoietic transplant recipients and the small number of proven IFD are other limitations of this single-center study.
In conclusion, when recommended clinical, radiological and microbiological investigations are used, the utility of PCT for the initial assessment of the etiology of febrile neutropenia is limited. While PCT measured at the onset of fever cannot be used for the decision to start or withhold immediate empirical antibacterial therapy, a PCT value higher than 500 pg/mL in FUO on day 2 might suggest an occult bacterial infection, i.e. with lacking microbiological and clinical documentation. In contrast, PCT monitoring is useful for diagnostic reassessment in persistent fever. A delayed PCT peak higher than 500 pg/mL contributes to the early diagnosis of invasive fungal disease and PCT decrease in follow-up promptly reflects response to antifungal therapy when compared with the slow resolution of fever.