Procalcitonin: Is This the Promised Biomarker for Critically Ill Patients?

Abstract

Objective Procalcitonin (PCT) has been increasingly used in the critical care setting to determine the presence of bacterial infection and also to guide antibiotic therapy. We reviewed PCT's physiologic role, as well as its clinical utility for the management of pediatric critically ill patients.

Findings PCT is a precursor of the hormone calcitonin. Its production is induced by inflammatory conditions, especially bacterial infections. Literature shows that PCT is a moderately reliable diagnostic test for severe bacterial infection in children. Synthesis of available adult studies suggests that the use of PCT-based algorithms to support medical decision making reduces antibiotic exposure without compromising safety in critically ill patients. However, no study has addressed the usefulness and safety of PCT to guide antibiotic therapy in severely ill children. In pediatric patients with acute lower respiratory tract infections, the use of PCT-based algorithms also led to a safe decrease in antibiotic treatment duration.

Conclusion PCT has demonstrated clinical utility in the pediatric critical care setting when used for the diagnosis of bacterial infections and to guide antibiotic use in children with acute lower respiratory tract infections. However, more research is needed in critically ill children to determine the utility of PCT-driven antibiotic therapy in this population.

Introduction

Over the last few decades, studies have shown procalcitonin (PCT) to be a potentially useful biomarker in the critical care setting. Initially used to diagnose severe bacterial infections, PCT has also been proposed to improve the appropriateness of antibiotic choice in bacterial infections and to guide the duration of antibiotic treatment in adult and pediatric patients.

Antibiotic overuse in intensive care units (ICUs) is a major problem. Between 40 and 72% of patients admitted to pediatric intensive care units (PICUs) receive antibiotics as part of an empiric treatment and in 9 to 33% of cases, inappropriately.^{1 2 3 4 5 6 7} As current recommendations about the optimal duration of antibiotic treatment in empiric situations and/or proven severe bacterial infections are vague and frequently based on expert opinion, ICU physicians must rely on their clinical judgment to decide when to stop antibiotics. In the last few years, the use of PCT-based algorithms has been successfully proposed to objectively guide antibiotic use in critically ill patients. In this article, we review the physiologic role of PCT, as well as its clinical utility for the management of adult and pediatric critically ill patients.

Biochemistry of Procalcitonin

PCT is a 116-AA polypeptide precursor of the hormone calcitonin, which is responsible for reducing calcium levels in response to hypercalcemia.⁸ PCT is a composite of three sections (Fig. 1): (1) the N-terminal sequence (N-PCT), (2) immature calcitonin, and (3) calcitonin carboxyl-terminus peptide 1 (CCP-1).⁸ The CALC-1 gene, present in chromosome 11, governs PCT production. CALC-1 is translated into pre-PCT, which then moves to an endoplasmic reticulum compartment where its N-terminal sequence is cleaved to generate PCT.^{9 10} PCT then translocates to the Golgi apparatus where it is cleaved to generate mature calcitonin.^{10 11}

Fig. 1.

Procalcitonin structure.

In healthy individuals, PCT is produced by the thyroid C-cells with extremely low concentrations being detected in the blood (∼0.19 fmol/mL), as practically all PCT is converted into calcitonin.^{9 12 13} Increased levels of PCT are observed in medullary thyroid cancer and other neuroendocrine tumors, such as small cell lung cancer and pheocromocytoma.^{9 14} Furthermore, inflammatory and bacterial infectious stimuli, especially sepsis, upregulate the expression of the CALC-1 gene. This induces the production of PCT in not only thyroidal, but also extrathyroidal tissues, such as adrenals, spleen, spinal cord, brain, liver, kidney, pancreas, colon, lungs, and fat.^{9 11 15 16 17 18 19} However, these organs are incapable of biosynthesizing mature calcitonin. It is possible that cytokines block the proteolysis of PCT to calcitonin in the endoplasmic reticulum, but this hypothesis has yet to be confirmed.¹³ This tissue-wide secretion explains the important elevations of PCT with concomitant low serum levels of calcitonin in septic patients.²⁰

Physiological Role of Procalcitonin

It is still not clear if PCT is a surrogate biomarker of bacterial infection and “poor prognosis” or if it actually plays a causative role in the inflammatory response. Multiple animal studies have attempted to address this question. Whang et al showed that the injection of PCT in healthy hamsters did not increase interleukin-1β (IL-1β) or tumor necrosis factor-α (TNF-α) levels (i.e., it did not induce sepsis) and had no effect on mortality.²¹ Based on these findings, the authors hypothesized that PCT acts as a secondary mediator requiring a primary inflammatory insult to exert its toxicity in sepsis.²¹

Dandona et al demonstrated that levels of PCT rise rapidly after the injection of endotoxin.¹² The same findings were observed after the administration of TNF-α, IL-1, IL-2, and IL-6.^{13 22} Studies also showed that the injection of PCT in septic animals further stimulates the release of proinflammatory cytokines (IL-6, IL-1β, and TNF-α) and activates macrophages, leading to the amplification of the initial inflammatory response and increasing mortality.^{23 24 25 26} Similar to IL-8, PCT also induces an increase in the concentration of intracellular calcium in granulocytes.¹⁷ Finally, PCT may be associated with the vasodilation seen in septic shock, as it increased the expression of nitric oxide synthase and, consequently, its production in rat aortic vascular smooth muscle cells prestimulated using lipopolysaccharide and TNF-α.²⁷

Supporting the hypothesis that PCT has immunomodulatory abilities, Nylen et al showed that the neutralization of PCT using a PCT-reactive antiserum significantly reduced the mortality of septic hamsters from 82 to 54% (p = 0.045).²⁸ Furthermore, Wagner et al and Martinez et al also reported improvement of physiologic and metabolic parameters and an increased survival of septic pigs and dogs, respectively, after PCT suppression.^{29 30} The mechanism through which PCT may cause deleterious effects includes the targeting of hematopoietic cells with consequent decrease in phagocytosis, directed neutrophil migration, chemotaxis of monocytes, and microbial activities.^{23 25 31}

Procalcitonin Kinetics

Experimental Studies

Dandona et al described the kinetics of PCT after the injection of Escherichia coli toxin in seven healthy human subjects.¹² PCT was first detected in the participants' plasma 4 hours after the exposure to the toxin. Its blood level peaked at 6 hours. A plateau in the plasma concentration was observed between 8 and 24 hours. Finally, the plasma levels of PCT returned to baseline after 48 to 72 hours. Brunkhorst et al reported the evolution of PCT levels in a patient who accidentally received blood contaminated by Acinetobacter baumannii (a “natural” experiment).³² In this case, a rise in PCT levels was observed 2.5 hours after the exposure, with a peak reached after 13.5 hours and a plateau observed as of 24 hours. Based on the findings of the aforementioned studies, PCT seems to follow first-order elimination kinetics with a half-life of 22.5 hours.³²

Clinical Setting: Infectious Processes

The evolution of PCT demonstrates a rise and fall pattern most consistent with infection resolution in several studies that included adults diagnosed with sepsis and septic shock (Table 1).^{33 34 35 36 37} However, PCT levels were also observed to vary significantly between studies on similar populations, probably due to the moment in the infectious process at which levels are drawn and also due to the use of different assays.

Table 1. Kinetics of PCT in adult and pediatric patients with suspected/proven bacterial infection.

Authors	N	PCT levels (ng/mL) according to length of antibiotic treatment (mean ± SD)
		0 h	6 h	12 h	Day 1	Day 2	Day 3	Day 4	Day 5	Day 8	Day 9	Day 10
Adult patients
Septic shock
Reynolds et al 2012	270	–	–	–	31.7 ± 58.2	25.6 ± 46.4	17.5 ± 31.4	10.9 ± 17.1	8.3 ± 11.9	5.9 ± 8.1	4.4 ± 6.0	3.4 ± 4.2
Sepsis
Miki and Iba 2015	30	10	–	–	10.5	–	2	–	–	–	1	–
Reynolds et al 2012	80	–	–	–	12.8 ± 44.2	12.1 ± 53.7	8.8 ± 52.9	8.0 ± 54.3	6.3 ± 42.8	5.9 ± 38.1	4.7 ± 27.0	2.3 ± 6.6
Charles et al 2009	180	–	–	–	27.2 ± 62.7	27.4 ± 45.1	24.4 ± 58.4	17.3 ± 45.8	–	–	–	–
Castelli et al 2004a	34	–	–	–	1.6	1.2	0.95	0.6	0.5	0.35	0.25	0.25
Ugarte et al 1999	111	2.1	–	–	4.3	4.3	3.7	1.4	–	–	–	–
Nonseptic shock
Reynolds et al 2012	48	–	–	–	11.4 ± 22.4	7.6 ± 13.5	4.9 ± 10.1	4.5 ± 8.2	2.4 ± 3.5	2.7 ± 3.2	1.8 ± 1.6	1.3 ± 1.3
Noninfected patients without shock
Reynolds et al 2012	200	–	–	–	4.0 ± 21.7	4.5 ± 26.1	1.7 ± 4.0	0.9 ± 1.3	0.7 ± 1.0	0.6 ± 0.7	0.6 ± 0.8	0.6 ± 0.6
Ugarte et al 1999a	79	0.8	–	–	1.2	1.6	1.3	0.8	–	–	–	–
Pediatrıc patients
Sepsis
Zurek and Vavrina 2015	62	–	–	–	16.3 ± 28.8	15.6 ± 34.3	18.4 ± 69.0	23.8 ± 87.1	16.2 ± 67.0	–	–	–
Garcia et al 2012	7	3.2 ± 1.9	–	–	6.6 ± 4.6	55.5 ± 90.9	–	–	–	–	–	–
Casado-Flores et al 2003a	80	20.4	33.9	–	18.8	7	–	–	–	–	–	–
Han et al 2003a	78	7.1 (0.9–44.8)	–	–	–	–	2.9 (0.1–32.4)	–	–	–	–	–
Van der Kaay 2002a	64	315	440	275	180	100	–`	–	–	0	–	–
Meningitis
Alkholi et al 2011	20	26.8 ± 12	–	–	–	–	10.8 ± 5	–	–	–	–	–

Abbreviations: PCT, procalcitonin; SD, standard deviation.

^aMedian (interquartile range).

In general, PCT levels are positively associated with the severity of the infectious process and clinical picture (i.e., PCT levels in sepsis are lower than levels observed in septic shock). The meta-analysis of Liu et al showed that high levels of PCT are strongly associated with mortality in adult patients with sepsis, severe sepsis, and septic shock (pooled relative risk [RR], 2.60; 95% confidence interval [CI], 2.05–3.30).³⁸ Furthermore, the absence of PCT clearance during antibiotic treatment was associated with an even higher mortality risk (RR, 3.05; 95% CI, 2.35–3.95).

The kinetics of PCT in children with different severe bacterial infections has also been described, but most authors limited the monitoring of PCT to the first 24 to 72 hours after initiation of treatment (Table 1).^{39 40 41 42 43 44 45 46} Consequently, most pediatric studies do not provide the full portrait of PCT evolution during bacterial infections. One exception is the Van der Kaay et al study, which included 64 children with meningococcemia. PCT was measured on days 0, 2, and 8 of treatment; on day 0, its level was 300 to 350 ng/mL and decreased to virtually zero on day 8.⁴⁴ In contrast, Zurek and Vavrina did not observe a similar fall over the first 5 days of antibiotic therapy in 62 septic children, with mean PCT levels remaining between 15.6 ± 34.3 and 23.8 ± 87.1 ng/mL.⁴⁶ As no information regarding the clinical response to antibiotic treatment was provided, it is possible that the persistent elevation in PCT was due to antibiotic treatment failure. Another possible explanation is that the 5-day follow-up period was not long enough to capture the fall in the PCT levels. From these studies, we can conclude that PCT levels are higher in children with severe bacterial infections, but we cannot yet properly describe its trajectory once antibiotics are initiated and define the thresholds associated with treatment success.

Clinical Setting: Inflammatory Processes

The presence of high PCT levels is not perfectly specific to bacterial infections. PCT levels are also elevated in the presence of systemic inflammatory response syndrome (SIRS) caused by major surgery/trauma, severe burns, cardiogenic shock, and exposure to cardiopulmonary bypass (CPB). Similarly to what is observed in patients with bacterial infections, PCT levels in inflammatory conditions present a rise and fall pattern in adults and children with good clinical evolution.^{35 36 45 47 48} Furthermore, PCT levels tend to increase proportionally to the severity of the clinical picture, but, when compared with levels observed in sepsis and septic shock, PCT levels for SIRS (related or not to CPB exposure) and nonseptic shock are significantly lower.

Difference in Kinetics between Procalcitonin and Other Infection Markers

The kinetics of PCT and those of other infection markers have been compared in adult and pediatric patients. Dandona et al and Brunkhorst et al observed that a rise in PCT levels occurs more rapidly after the insult than an increase in C-reactive protein (CRP), but slower than TNF-α and IL-6 (Table 2).^{12 32} The same pattern is observed regarding the return to baseline levels, with TNF-α and IL-6 being the first markers to reach normal levels and CRP being the last. Moreover, Castelli et al compared the levels of PCT and CRP in 34 adult patients with sepsis from ICU admission to day 8 of evolution. They showed that, while CRP levels peaked on day 3 of treatment and remained high until day 8, PCT levels peaked on day 1 and declined thereafter.³³

Table 2. Kinetics of PCT compared with other biomarkers.

Biomarker	Start to rise (h)	Peak (h)	Plateau (h)	Return to baseline (h)
PCT	2.5 to 4	6 to 13.5	8 to 24	48 to 72
CRP	12 to 24	30	20 to 72	72 to 168
IL-6	1 to 1.5	3	1.5 to 14	–
TNF-α	1	2 to 2.5	–	6 to 9

Abbreviations: CRP, C-reactive protein; IL-6, interleukin 6; PCT, procalcitonin; TNF-α, tumor necrosis factor-α.

As previously mentioned, studies comparing the evolution of PCT and other infection markers in pediatric patients mainly focused on the first 24 to 72 hours after diagnosis. In 20 children treated for bacterial meningitis, Alkholi et al demonstrated a fall in PCT from day 1 to day 3 for PCT (26.8 ± 12 vs. 10.8 ± 5.3 ng/mL; p < 0.001) and white blood cell count (18.4 ± 9.1 vs. 12.6 ± 5.8 mm³; p < 0.05), while CRP remained stable (24.4 ± 12.1 vs. 28.6 ± 12 mg/dL; p > 0.05).³⁹ Van der Kaay et al measured PCT, CRP, and IL-6 levels in 64 children treated for meningococcemia and reported that both PCT and IL-6 peaked at PICU admission with levels falling within 6 hours. In contrast, CRP peaked at 24 hours followed by a subsequent decline.⁴⁴ In this study, levels were also measured at day 8, showing a return to baseline for all markers.

Procalcitonin Use in the Critical Care Setting

Use of Procalcitonin for the Diagnosis of Severe Bacterial Infections/Sepsis

PCT has been proposed as a biomarker capable of discriminating between bacterial infections and other inflammatory processes (including viral infection) based on the significantly higher increase in its levels in the presence of bacterial stimulus. Although two cutoffs are commonly used (≥0.5 and ≥2.0 ng/mL),^{8 49 50} other values have been proposed to adjust for the timing of measurement (if early in the infection, PCT levels may not have reached their peak) and different patient populations (e.g., patients in the emergency room vs. in the ICU). Importantly, studies have suggested that the use of anti-inflammatories, such as steroids, does not meaningfully affect PCT levels in the context of pneumonia and sepsis.^{51 52}

Tang et al performed a meta-analysis to estimate the accuracy of PCT to diagnose sepsis in critically ill adult patients.⁵³ The pooled results of 14 phase II diagnostic studies (i.e., studies evaluating the discriminatory ability of a test) showed positive and negative likelihood ratios of 3.03 (95% CI, 2.51–3.65) and 0.43 (95% CI, 0.37–0.48), respectively, and an area under the receiver operating characteristics curve (AUC ROC) of 0.73 (95% CI, 0.69–0.77). Wacker et al tried to answer the same question with a meta-analysis that included 26 adult and 4 pediatric studies, with PCT cutoffs ranging from 0.28 to 5.79 ng/mL.⁵⁴ The pooled results showed a sensitivity of 0.77 (95% CI, 0.72–0.81), specificity of 0.79 (95% CI, 0.74–0.84), and an AUC ROC of 0.85 (95% CI, 0.81–0.88). Importantly, despite the substantial heterogeneity between studies, the proportion of heterogeneity caused by different cutoffs was small (0.05) and the AUC ROCs for the pediatric and adult studies were identical (0.85 [95% CI, 0.82–0.88] vs. 0.85 [95% CI, 0.81–0.88]).

Several meta-analyses were performed to estimate the usefulness of PCT for the diagnosis of different severe bacterial infections in pediatric patient populations (Table 3).^{50 55 56 57 58 59} The results for sepsis demonstrate that PCT is a moderately useful diagnostic test, with sensitivity between 0.59 and 0.89, and specificities between 0.74 and 0.81.^{55 57 59} The exception is meningococcal disease, for which an excellent AUC ROC was obtained (0.95; 95% CI, 0.93–0.97).⁵⁵ PCT was found to be a good diagnostic test for severe bacterial infections, with AUC ROC between 0.84 and 0.85.^{50 56 58}

Table 3. Meta-analyses evaluating the use of PCT to diagnose severe bacterial infections in pediatric patients.

Authors	Subgroup	N (articles)	PCT cutoff range (ng/mL)	Pooled results (95% CI)
				Sens	Sp	LR (+)	LR(–)	AUC ROC	DOR	RR
Sepsis
Bell et al 2015	Meningococcal disease	6	0.2–2.0	0.89 (0.76–0.96)	0.74 (0.40–0.92)	–	–	0.95 (0.93–0.97)	–	–
Lin et al 2012	Neutropenic patients	10	0.12–7.0	0.59 (0.42–0.74)	0.76 (0.64–0.85)	1.97 (1.46–2.66)	–	–	4.82 (2.10–11)	–
Simon et al 2004	Adult, children, and neonates. Also other SBI	12	0.5–6.1	0.88 (0.80–0.93)	0.81 (0.67–0.90)	3.58 (2.99–4.28)	0.18 (0.15–0.23)	–	–	–
Severe bacterial infection
England et al 2014	<91 d old	7	0.12–0.5	–	–	–	–	–	–	3.97 (3.41–4.62
Yo et al 2012	7 d to 36 mo old	8	0.5–2.0	0.83 (0.70–0.91)	0.69 (0.59–0.85)	2.69 (1.87–3.87)	0.25 (0.15–0.40)	0.84 (0.80–0.87)	–	–
Jones et al 2007	–	7	0.5–2.0	–	–	–	–	0.85 (0.63–0.95)	–	–
Viral versus bacterial infection
Simon et al 2004	Adult, children, and neonates. Also other SBI	12	0.5–6.1	0.92 (0.86–0.95)	0.73 (0.42–0.91)	6.05 (4.67–7.82)	0.10 (0.06–0.15)	–	–	–

Abbreviations: 95% CI, 95% confidence interval; AUC ROC, area under the receiver operating curve; DOR, diagnostic odds ratio; LR(+), positive likelihood ratio; LR(–), negative likelihood ratio; PCT, procalcitonin; RR, relative risk; SBI, severe bacterial infection; sens, sensitivity; sp, specificity.

Finally, in their systematic review, Simon et al demonstrated the ability of PCT to discriminate between bacterial and viral infections to be excellent, with positive and negative likelihood ratios of 6.05 (95% CI, 4.67–7.82) and 0.10 (95% CI, 0.06–0.15).⁵⁷ After the publication of this meta-analysis in 2004, Laham et al performed a retrospective cohort study to evaluate the usefulness of PCT for the diagnosis of bacterial coinfection in 40 PICU patients with bronchiolitis.⁶⁰ Using a cutoff of 1.5 ng/mL, they obtained a sensitivity of 0.80, a specificity of 1.0, and an AUC ROC of 0.88.

Utility of Procalcitonin to Guide Antibiotic Use

The rise in PCT levels observed in patients with bacterial infections and their decline in the ones who responded to antibiotic therapy led to the hypothesis that this biomarker could be used to tailor antibiotic use and duration. In the last decade, several randomized controlled trials (RCTs) evaluated the benefit of using PCT-based algorithms to drive medical decisions regarding antibiotic use (initiation and/or withdrawal) in adult critical care patients with suspected/proven severe bacterial infections, including sepsis, septic shock, and ventilator-associated pneumonia (Table 4).^{61 62 63 64 65 66 67 68} Importantly, all aforementioned studies tested different PCT algorithms, with cutoffs for antibiotic initiation ranging from ≥0.5 to >1.0 ng/mL. Regarding antibiotic discontinuation, suggested cutoffs included absolute PCT values ranging from <1.0 to <0.1 ng/mL and PCT level reductions between ≥65 and ≥90% of initial value (baseline).

Table 4. Evidence regarding the use of PCT to tailor initiation/duration of antibiotics in adult critical care patients.

Authors	N	Infection type	Antibiotic duration (mean ± SD), d	Mortality (%)	ICU LOS (mean ± SD; d)	Hospital LOS (mean ± SD), d
Individual RCTs
Shehabi et al 2014	400	Suspected severe bacterial infection	9 (6–20) versus 11 (6–22)a HR, 1.01 (95% CI, 0.80–1.26)	90-d mortality, 35 (18) versus 31 (16)a (p = 0.60)	6 (3–9.5) versus 6 (4–10)a (p = 0.87)	15 (9–29) versus 17 (10–32)a (p = 0.19)
Annane et al 2013	62	Suspected severe sepsis and septic shock	5 (2–5) versus 5 (3–5)a (p = 0.52)	Hospital mortality, 23 versus 33% RR, 0.68 (95% CI, 0.30–1.55)	22 (8–42) versus 23 (10–60)a (p = 0.58)	27 (9–49) versus 33 (11–69)a (p = 0.22)
Jensen et al 2011	1,200	Suspected severe bacterial infection	6 (3–11) versus 4 (3–10)a	28-d mortality, RR, 0.98 (95% CI, 0.83–1.16)	6 (3–12) versus 5 (3–11)a (p = 0.004)	–
Bouadma et al 2010	630	Suspected severe bacterial infections	10.3 ± 7.7 versus 13.3 ± 7.6 (p < 0.0001)	28-d mortality, 21.2 versus 20.4% MD, 0.8% (95% CI, −4.6 to 6.2)	15.9 ± 16.1 versus 14.4 ± 14.1 MD, 1.5 (95% CI, − 0.9 to 3.9)	26.1 ± 19.3 versus 26.4 ± 18.3 MD, −0.3 (95% CI, −3.2 to 2.7)
Hochreiter et al 2009	110	Suspected/confirmed severe bacterial infection	5.9 ± 1.7 versus 7.9 ± 0.5 (p < 0.001)	Hospital mortality, 21.4 versus 23%	15.5 ± 12.5 versus 17.7 ± 10.1 (p = 0.046)	–
Schroeder et al 2009	27	Suspected severe sepsis	6.6 ± 1.1 versus 8.3 ± 0.7 (p < 0.001)	Hospital mortality, 26.3 versus 26.4% (p > 0.05)	16.4 ± 8.3 versus 16.7 ± 5.6 (p > 0.05)	–
Stolz et al 2009	101	Suspected VAP	10 (6–16) versus 15 (10–23)a (p 0.038)	28-d mortality, 16 versus 24% (p = 0.33)	–	26 (7–21) versus 26 (16.8–22.3)a
Nobre et al 2008	79	Severe sepsis or septic shock	6 (2–33) versus 9.5 (3–34)b MD, 2.6 (95% CI, − 0.3 to 5.5)	28-d mortality, 20.5 versus 20% (p = 0.82) RR, 0.9 (95% CI, 0.9–1.3)	7.7 ± 5.7 versus 12.3 ± 9.7 (p = 0.02) MD, 4.6 (95% CI, 1.0–8.2)	20.9 ± 17 versus 28.1 ± 20 (p = 0.85) MD, 2.5 (95% CI, −1.5 to 6.5)
Meta-analyses
Soni et al 2013	5 RCTs	Suspected severe bacterial infection	MD, −2.05 (95% CI, −2.59 to 1.52)	Short-term mortality (hospital + 28-d) MD, 0.00 (95% CI, −0.06 to 0.05)	MD, 0.33 (95% CI, −1.88 to 2.53)	–
Prkno et al 2013	7 RCTs	Severe sepsis and septic shock	HR, 1.27 (95% CI, 1.01–1.53)	28-d mortality RR, 1.02 (95% CI 0.85–1.23)	HR, 1.05 (95% CI, 0.82–1.29)	HR, 1.03 (95% CI, 0.82–1.24)
Schuetz et al 2012	5 RCTs	Acute respiratory infection	MD, −3.17 (95% CI, − 4.28 to −2.06)	30-d mortality OR, 0.79 (95% CI, 0.53–1.17)	MD, 1.01 (95% CI, −1.26 to 3.28)	MD, −1.36 (95% CI, −4.55 to 1.77)
Matthaiou et al 2012	7 RCTs	Severe bacterial infection	MD, −3.15 (95% CI, −4.35 to −1.95)	28-d mortality OR, 0.95 (95% CI, 0.79–1.16)	MD, −0.36 (95% CI, −1.97 to 1.26)	MD, −0.12 (95% CI, −1.09 to 0.85)
Heyland et al 2011	5 RCTs	Severe bacterial infection	MD, −2.14 (95% CI, −2.51 to −1.78)	Hospital mortality OR, 1.06 (95% CI, 0.86–1.30)	–	–
Kopterides et al 2010	5 RCTs	Severe bacterial infection	MD, −2.36 (95% CI, −3.11 to −1.61)	28-d mortality OR, 0.93 (95% CI, 0.69–1.26)	MD, −0.57 (95% CI, −1.90 to 0.76)	MD, −0.13 (95% CI, −1.10 to 0.84)

Abbreviations: 95% CI, 95% confidence interval; HR, hazard ratio; ICU, intensive care unit; LOS, length of stay; MD, mean difference; OR, odds ratio; RCTs, randomized controlled trials; RR, relative risk; SD, standard deviation; VAP, ventilator-associated pneumonia.

^aMedian and interquartile range (IQR).

^bMedian and range.

Since 2010, six meta-analyses assessing the effect of using PCT algorithms to tailor antibiotic use in adult critical care have been published (Table 4).^{69 70 71 72 73 74} They all showed a reduction in antibiotic treatment duration, with pooled mean differences from −2.05 days (95% CI, −2.59 to −1.52) to −3.17 days (95% CI, −4.35 to −1.95), with no difference in mortality and ICU/hospital length of stay between study groups. Regarding antibiotic treatment duration, it is important to mention that three RCTs failed to demonstrate a reduction in antibiotic use associated with PCT algorithms.^{63 66 67} Their results are discussed in depth in the next paragraphs. Importantly, the results of two of these trials were released in 2013 and 2014, which precluded their inclusion in the most recent meta-analyses (published in 2013).^{66 67}

In 2008, Nobre et al published the first RCT evaluating a PCT-based algorithm to tailor antibiotic use in adult ICUs.⁶³ In the intention-to-treat analysis, they observed a clinically relevant, but not statistically significant, mean difference in the antibiotic treatment duration between groups which was favorable to PCT (median of 6 days [range 2–33] vs. 9.5 days [range 3–34]; p = 0.15). Study results were significantly impacted by the fact that stopping rules in the PCT group were overruled by treating physicians in 19% of patients. In their per-protocol analysis, a clinically and statistically significant difference between groups was calculated (median of 6 days [range 4–16] vs. 12.5 days [range 8–16]; p = 0.0002). Furthermore, the probability of having antibiotics stopped earlier was estimated to be twofold higher in the PCT group (hazard ratio [HR], 1.9; 95% CI, 1.2–3.2).

In their 2013 publication, Annane et al evaluated the benefit of using a PCT-based algorithm in 62 adults with suspected septic shock, but without a clear source of infection or positive cultures at study enrolment.⁶⁶ Due to the low incidence of eligible patients, this trial was prematurely stopped. Data up to this point showed no difference between groups regarding antibiotic use on day 5 after randomization (median of 5 days [interquartile range, IQR, 2–5] vs. 5 days [IQR 3–5]; p = 0.52). The authors proposed that the low number of patients whose PCT levels were <0.25 ng/mL (cutoff to not initiate or to stop antibiotics) throughout the study probably explained the lack of difference between groups. Moreover, the noncompliance with the PCT algorithm ranged from 17 to 37% at the three study time points (6 hours, 3 days, and 5 days). Finally, it is likely that the 5-day interval used was not long enough for PCT levels to decrease below the <0.25 ng/mL threshold in patients who were more severely ill, making it impossible for the proposed algorithm to show clinical benefit.

In 2014, Shehabi et al also reported the findings of their RCT, which failed to show a reduction in antibiotic treatment duration for adult ICU patients with suspected sepsis with the use of a PCT algorithm (HR, 1.01; 95% CI, 0.80–1.26).⁶⁷ Possible reasons for this negative result include a much lower PCT cutoff to stop antibiotics (0.1 ng/mL instead of 0.25–0.5 ng/mL) and the use of the algorithm only while patients were in the ICU (i.e., ward medical teams had no access to PCT levels, thus having to use their clinical judgment to make decisions about treatment duration for patients in both arms of the study).

Currently, only two RCTs evaluated the use of PCT to tailor antibiotic use in children and neither included critically ill pediatric patients. The open-label, parallel RCT of Esposito et al included 310 children (1 month to 18 years old) admitted to a pediatric ward for community-acquired pneumonia.⁷⁵ The use of an algorithm that encouraged to not start and/or to stop antibiotics when PCT level was <0.25 ng/mL led to a 5-day decrease in antimicrobial duration (5.37 days vs. 10.96 days; p < 0.05), as well as a reduction in antibiotic-related adverse events (3.9 vs. 25.2%; p < 0.05) and infection recurrence (0.007 vs. 0.04%; p-value not calculated). In the RCT by Baer et al, a cut-off of <0.25 ng/mL was used to decide upon both antibiotic initiation and termination for children with lower respiratory tract infection.⁷⁶ They also observed a reduction in antibiotic use (mean difference, −1.8 days; 95% CI, −3.1 to −0.5) with no increase in serious adverse events.

Conclusion

PCT is clinically useful in critically ill patients. Specifically for the pediatric population, PCT is a moderately helpful diagnostic test, except for the diagnosis of meningococcal disease, where it may be highly accurate. Regarding its use to tailor antibiotic therapy, pediatric studies suggest that the use of PCT-based algorithms may safely decrease antibiotic treatment duration in patients with acute lower respiratory tract infection. In addition, adult studies including severely ill patients have showed that PCT-guided therapy leads to a decrease of antibiotic exposure without compromising infection cure and patient safety. At this point, the same cannot be said for critically ill children, as no studies have yet been performed to determine the utility of PCT-driven antibiotic therapy in this patient population. Before such studies can be performed, it is imperative to determine PCT kinetics during antibiotic treatment in PICU patients with severe bacterial infections and to define the PCT thresholds associated with infection cure.