The clinical presentation and severity of patients having acute pancreatitis varies considerably, from a mild abdominal discomfort to multiple organ failure and death. The potential role of biomarkers in this condition should thus be twofold: 1) a prognostic value, to help define the most appropriate therapeutic approach, by predicting the severity of the disease and accurately select those patients needing close monitoring in the ICU; and 2) a diagnostic value, to help identify those patients having infected pancreatic necrosis, who might need drainage or surgery. Mofidi et al. 1 have recently reviewed the potential role of PCT in answering these two questions, by analysing 12 observational studies [2-9] totalling 956 patients. The threshold PCT value used in these studies to predict the severity of pancreatitis varied from 0.25 to 1.8 mg/L, with an associated combined sensitivity of 0.72 (0.65-0.78), a specificity of 0.86 (0.83-0.89), and an area under the receiver operating curve (AUROC) of 0.87. In the seven studies (n = 264 patients) examining the value of PCT for predicting the presence of infected necrosis 2 3 4 5 7 8 9 , the threshold value varied across studies between 0.48 and 3.5 mg/L, with an associated sensitivity of 0.8 (0.71-0.88), a specificity of 0.91 (0.87-0.94), and an AUROC of 0.91 (Table 1). In these seven studies, PCT levels were confronted to microbiological results obtained from fine-needle biopsy and culture of intra-abdominal collections, taken as the “gold standard.”

^aNumber of patients included in diagnostic studies of infected pancreatic necrosis.

*Directness: studies provide evidence of a direct association between a treatment or a given risk factor and a judgment criterion.

**Consistency: results from studies of similar level of evidence are not contradictory.

Other less commonly measured biomarkers (IL-6, IL-8, sTREM-1, TNF-α) also have been compared to PCT for their ability to help answer the two questions above. These biomarkers provided AUROC comparable to those of PCT, both in terms of prognostic value and of diagnosis of infected necrosis 2 4 5 6 9 10 . Conversely, CRP levels appear less discriminatory for the prediction of infected necrosis 2 . No study has evaluated the value of repeated PCT measurements to predict infection, and no study has evaluated the impact on patients’ outcome of the initiation of antibiotic therapy guided by a biomarker level in patients suspected of infected necrosis.

In summary, we suggest that PCT be measured to help predict infection in patients suspected of infected necrosis during acute pancreatitis; it is however difficult from the available literature to define a precise threshold value (0.5-1.0 mg/L). A PCT value above the threshold might reinforce the clinician’s judgment that a fine-needle aspiration and culture is needed to confirm infection, while a value below this threshold might help deferring this intervention and proceed with watchful waiting. There is insufficient data to recommend initiating antibiotic therapy based on biomarker levels: this decision is based on a careful repeated evaluation of the patient and on the results of fine-needle aspiration material, which currently remains the cornerstone for the decision to initiate or maintain antibiotic therapy.

Antibiotics often are prescribed in excess to patients having a clinical syndrome of community-acquired lower respiratory tract infection (LRTI). Despite the usually viral aetiology of their illness, an estimated 75% of patients with acute bronchitis receive antibiotics 11 ; indeed, clinical presentation does not allow the distinction between bacterial and viral infection, which encourages physicians to err on the “safe side” and prescribe antibiotics. Communication campaigns inciting primary physicians to limit unnecessary prescriptions for LRTI have a moderate impact, which is difficult to maintain over time 12 13 . In this context, the addition of biomarker measurements to the clinical evaluation of such patients may have two main potential effects: improve the diagnostic accuracy, and reassure the patient and the physician that antibiotic therapy is unnecessary.

An abundant literature is available on PCT-guided initiation of antibiotic therapy in patients suspected clinically of having LRTI, providing a high-level of evidence. To date, 11 randomised, controlled studies using a similar approach have been published and provide consistent results 14 15 16 17 18 19 20 21 22 23 24 25 . All these studies have used a similar algorithm 26 27 28 29 to help decide on the initiation and continuation of antibiotic therapy, with a lower PCT threshold of <0.25 ng/mL to encourage physicians to withhold antibiotic prescription. The absolute risk reduction of antibiotic administration varies between 11% and 72% across these studies compared with “usual care” based on local recommendations and physicians’ judgment and preferences (Table 2). In one study, however, antibiotic prescriptions increased by 6% with PCT-guided therapy 19 . It also should be noted that the 0.25 ng/mL threshold may be less reliable in the elderly, where an 8% false-positive rate has been reported 30 .

*Directness: studies provide evidence of a direct association between a treatment or a given risk factor and a judgment criterion.

**Consistency: results from studies of similar level of evidence are not contradictory.

Among the 14 studies of PCT-guided therapy for LRTI reviewed by the Cochrane Collaboration 31 32 , only 3 enrolled patients with a nosocomial infection (hospital-acquired or ventilator-associated) 33 34 35 , 2 of which evaluated the impact of PCT-guided therapy on the initiation of treatment 33 34 . However, nosocomial acquisition of infection is identifiable only in the study by Bouadma et al. 33 , and only 5% (n = 141) of all patients enrolled fulfilled this criteria; nearly all patients in this subgroup were administered antibiotics (99% in the PCT-guided therapy group and 100% in controls). Repeated measurements might be helpful for initiating antibiotics in this subgroup; however, the few data available on a limited number of patients (n = 89 patients) 36 do not allow making a recommendation in this regard.

In summary, we suggest withholding antibiotic therapy in adult patients suspected of community-acquired LRTI and having a serum PCT level <0.25 ng/mL; if clinical suspicion is high, it is however recommended to repeat the PCT measurement at a 6-h interval and reassess the therapeutic approach, accounting for new clinical findings. In patients having nosocomial LRTI, data are insufficient to recommend tailoring the therapeutic approach based on a single PCT level or even repeated measurements.

Most controlled studies performed in intensive care patients have examined the value of biomarkers to limit the duration of antibiotic therapy, and few have concentrated on its initiation. Although a recent meta-analysis suggests that PCT is helpful for differentiating sepsis from SIRS 63 , the initiation of antibiotic therapy in ICU patients has been assessed in only two randomised open studies testing a PCT-based algorithm (Table 4) 33 34 . In the study by Layios et al. 34 , there was no difference in the rate of initiation of therapy between the control group and the PCT-based group (where antibiotics were strongly discouraged if PCT was lower than 0.25 ng/mL, and strongly encouraged if PCT was higher than 1 ng/mL). In the multicentre study performed by Bouadma et al. 33 and using a similar algorithm, the risk reduction of initiating antibiotic therapy varied between 5% and 13% across centres. However, the small number of patients having CAP (n = 69) and the very low observance of the algorithm for withholding antibiotics when PCT levels were low (6%) in this study do not allow concluding on this point.

*Directness: studies provide evidence of a direct association between a treatment or a given risk factor and a judgment criterion.

**Consistency: results from studies of similar level of evidence are not contradictory.

Relying on changes in PCT levels might be helpful for the initiation of antibiotic therapy in intensive care patients suspected of ICU-acquired infection; however, currently available data (on a total of 207 patients) are insufficient to base a recommendation on these 36 64 . One randomized, controlled study that enrolled 604 ICU patients has tested the diagnostic value of daily measurements of serum PCT levels (using a threshold of 1 ng/mL to rapidly initiate a diagnostic workup and protocolised therapy) 65 . The length of ICU stay and of mechanical ventilation were actually higher in the PCT arm (without difference in 28-day mortality), and time to adequate therapy was not lower (except for patients with bacteremia). Of note, antibiotic consumption was significantly higher in the PCT arm, as well as the total number of days spent in the ICU with three or more antibiotics.

In summary, we do not recommend using a threshold serum PCT value to help in the decision to initiate antibiotic therapy in ICU patients suspected of community-acquired pneumonia. There are insufficient data available to recommend using repeated PCT measurements and serum kinetics for the decision to initiate antibiotic therapy in ICU patients suspected of ICU-acquired infection.

Given the number of studies examining this question and the high level of evidence generated, investigating this question was limited to examining randomized, controlled studies having tested a strategy based on biomarker measurement(s), to the exclusion of all other study designs. All studies in hospitalised patients used serum PCT level measurements, as there is no study testing the impact of using another biomarker in this specific indication; whereas several studies have tested the value of CRP for initiating antibiotics in pre-hospital care 24 66 67 68 , none examined its potential impact on discontinuation of antibiotics, although several studies are ongoing (see http://www.clinicaltrials.gov/ct2/results?term=c+reactive+protein+and+duration&recr=&rslt=&type=&cond=&intr=&outc=&spons=&lead=&id=&state1=&cntry1=&state2=&cntry2=&state3=&cntry3=&locn=&gndr=&rcv_s=&rcv_e=&lup_s=&lup_e=). Accordingly, only studies using PCT levels are considered below.

To date, 14 trials have examined the clinical impact of PCT-guided antibiotic therapy and its discontinuation 15 16 17 18 19 20 22 23 33 35 69 70 72 , 73 . Nine of these focused on the latter objective; four were conducted in prehospital care or emergency room, whereas the remaining five were conducted in ICU patients. Although specific stopping rules may vary across trials and population enrolled, all studies used a PCT-based algorithm to help decide on stopping antibiotics (Table 5).

CAP community-acquired pneumonia, ICU intensive care unit, PCT procalcitonin, RTI respiratory tract infection, SIRS systemic inflammatory response syndrome, VAP ventilator-associated pneumonia.

In outpatients and emergency room patients (excluding ICU patients), a serum PCT level below 0.25 ng/mL obtained 3 days or more after initiation of antibiotics, or a more than 80% decrease from the peak PCT level, allows stopping therapy.

Five studies have enrolled ICU patients with community- or hospital-acquired infection; four used a similar algorithm and the fifth used a different algorithm. It seems reasonable to recommend using the algorithm tested on the largest number of patients, i.e., as in the ProVAP et Prorata studies 33 69 , where stopping therapy was strongly encouraged when the serum PCT level was <0.5 ng/mL at 3 days or more after initiating antibiotics, or an >80% decrease from the maximal serum PCT value was recorded.

In all studies examining the prognostic and follow-up value of PCT, the site of infection was known, to the exception of a few patients (n = 18) in the PRORATA study (10 and 8 in the PCT arm and control group, respectively). This small number does not allow any conclusion for this subgroup. It should be noted that patients having infective endocarditis, bone and joint infection, acute mediastinitis, intracerebral or intra-abdominal abscess were excluded from the above studies. Therefore, PCT-based algorithms cannot be used in these patients for discontinuing antibiotics.

Therefore, PCT-guided interruption of therapy can be used as indicated above in patients having a clinically documented site of infection to the exception of those sites listed above, which were excluded from clinical trials. When the site of infection is unknown, insufficient data are available to make a recommendation.

Among the nine trials testing the impact of PCT-guided discontinuation of therapy, only the PRORATA trial 33 enrolled immunocompromised patients in the ICU. This trial enrolled patients having HIV infection or AIDS, organ transplant recipients, patients having haematological malignancy or receiving chemotherapy or radiation therapy, immunosuppressive agents or long-term steroids, to the notable exception of bone marrow transplant recipients or those having severe neutropenia (<500 leucocytes/mm³). About a hundred such immunocompromised patients were included (47 and 51 in the PCT-guided therapy group and control group, respectively). In this subgroup, PCT-guided discontinuation of therapy was associated with a significantly reduced duration of therapy (3.6 days; 95% CI, 0.2-7 days), without apparent effect on morbidity or mortality (control vs. PCT, -7.1%; 95% CI -18.7 to 4.5%).

Therefore, PCT-guided algorithms to reduce the duration of antibiotic therapy can be used safely in immunocompromised patients, to the exception of neutropenic patients (<500 neutrophils/mm³) or bone marrow transplant recipients, which were excluded from trials and in whom PCT-guided therapy cannot be recommended.