Febrile neutropenia is a medical emergency necessitating early initiation of broad-spectrum parenteral antibiotics. The role of inflammatory markers needs to be delineated while managing febrile neutropenia. Procalcitonin (PCT) is a 116-amino acid peptide encoded by the CALC-1 gene on chromosome 11 [1]. Recent systematic reviews have shown PCT to be more discriminatory than C-reactive peptide (CRP) while evaluating fever in oncology patients despite inconsistent results [1,2]. Guidelines for febrile neutropenia management highlight insufficient data in serum inflammatory markers’ ability to guide treatment decisions [3].

Episodes with pneumonia (chest X-ray, CT scan or BAL positive), hemodynamic instability (requiring fluid resuscitation and/ or vasoactive support), identified focus of infection (e.g., abscesses, neutropenic enterocolitis) or prolonged fever ³72 hours with no identifiable non-infective causes were classified as CSI. Any organism identified (by culture, PCR, IgM serology or galactomannan) from sterile body fluids qualified for MDI. Documented bacterial infection was any bacterium identified from sterile body fluids, sputum, BAL or sinus aspirate samples. Documented fungal infection was defined as any fungus identified from above mentioned samples or serum galactomannan level ³1. Febrile neutropenia was defined as single oral temperature measurement of >101ºF or a temperature of >100.4ºF sustained over a 1 hour period with absolute neutrophil count (ANC) <1000/µL [3].

TABLE I	Clinical and Laboratory Characteristics of Febrile Episodes (N=821)

Median PCT (IQR) was higher in episodes of CSI [0.80(0.34–3.17) vs. 0.28 (0.16–0.50), P<0.001] and in episodes of MDI [0.71 (0.28–3.63) vs. 0.34 (0.19–0.76), P<0.001]. Median PCT (interquartile range) was higher in episodes leading to mortality [1.98 (0.46–15.9) vs. 0.38 (0.19–0.94), P<0.001]. ROC curve showed that PCT cut-off level ³0.7 ng/mL better discriminated episodes with CSI from those without CSI (AUC=0.740). PCT ³0.7 ng/mL was also optimum to discriminate episodes with MDI (AUC=0.636). PCT ³5.0 ng/mL had better discrimination for episodes resulting in death (AUC=0.751).

The diagnostic accuracy of PCT at a cut-off of ³0.7 ng/mL is tabulated in Table II. PCT had a poor sensitivity but a good specificity and NPV. PCT ³0.7 ng/mL adequately predicted CSI (PPV around 70%), but not MDI and DBI (PPV 20-30%). We compared the prevalence of significant infections among different ranges of PCT (normal PCT range <0.7 ng/mL); mild (0.7-1.4 ng/mL), moderate (1.5-4.9 ng/mL) and significant elevation of PCT (³5 ng/mL). The prevalence of CSI, MDI, documented bacterial infection and deaths increased significantly with the degree of elevation of PCT level (P<0.001).

TABLE II Diagnostic Accuracy of PCT ≥0.7 ng/mL in Detecting Significant Infections

Among 27 episodes with documented viral infections, 15 (55.5%) had PCT ³0.7 ng/mL (one had co-existent Documented bacterial infection). Viral etiology included cytomegalovirus, Epstein-Barr virus, herpes simplex, dengue and hepatitis viruses. Among 47 episodes with documented fungal infections, 23 (49%) had PCT ³0.7ng/mL (6 had co-existent documented bacterial infection). Isolated fungi included Aspergillus, Candida and Trichosporon asahii.

Among 24 deaths, 16 (66.7%) had neutropenia at admission, 7 (29.2%) had documented bacterial infection (6 were gram negative) and 3 (12.5%) had documented fungal infection. Mortality was higher in episodes with PCT ³5 ng/mL (12.4% vs.1.8%, P<0.001, relative risk of mortality 7.1 (95% CI=3.2–15.1)).

The optimum cut-off of PCT identified in our study (³0.7ng/ml) falls within the range identified in previous studies (0.5-0.8ng/mL) [1]. At this level, we found that PCT at admission is poorly sensitive (45-55%), both in neutropenic and non-neutropenic febrile episodes. Thus, a normal PCT level at admission does not exclude the possibility of an underlying significant infection. The good PPV for CSI (69-71%) as against that for MDI (30-31%) highlights the limited capability of currently available investigations to isolate the microorganisms, and elevated PCT at admission warns the clinician early about the likelihood of an underlying clinically significant infection. PCT ³5ng/mL at admission increased the risk of mortality by 7 fold. This warns the clinician to aggressively manage these episodes. Elevated PCT in viral and fungal infections (despite no bacterial co-infections) suggests that PCT is unable to discriminate bacterial infections from others. Our findings suggest that PCT correlates with the severity of infection rather than the type of infection.

Another prospective study of 230 febrile episodes in children receiving chemotherapy found a sensitivity of 93% at PCT ³0.4 ng/mL, but the specificity decreased to 45% at this cut-off [5]. We found PCT ³0.7 ng/mL as the optimum cut-off, which has reasonable sensitivity and specificity.

To conclude, a cut-off PCT levels of ³0.7 ng/mL and ³5.0 ng/mL at admission were optimum to predict significant infections and mortality in pediatric oncology patients, respectively. PCT at admission identifies febrile pediatric oncology patients with high risk for significant clinical course and death. However, it does not select a low risk group in whom hospitalization and parenteral antibiotics can be avoided. PCT correlates with the severity of infection and not the type of infection.