Overuse of antibiotics is common in both medical and surgical (perioperative medicine) intensive care units (ICU) leading to the development of antimicrobial resistance and hospital-acquired infections 1 . To decrease hospital-acquired infection incidence, antimicrobial consumption reduction in the surgical ICU is needed but unfortunately controlled studies comparing two different durations of antibiotics are scarce 2 . So, strategies other than randomized controlled studies might be of interest 3 4 . Recently, new strategies to reduce antibiotic duration have included the development of biomarker-directed treatment algorithms 5 6 . Procalcitonin (PCT), the 116-amino acid precursor of calcitonin, is elevated consecutively in several systemic inflammatory conditions and its magnitude correlates well with injury severity and prognosis 7 8 . In the ICU, serial PCT measurement might be used as a surrogate to facilitate the early discontinuation of antimicrobials 5 9 . Indeed, it has been reported that using a PCT plasma threshold from 0.25 ng/ml to 0.50 ng/ml or its decrease of at least 80% compared to its peak 5 9 10 11 allows withholding antibiotics earlier without affecting clinical outcome 12 13 14 15 . Surprisingly, most of the studies that evaluated the interest of PCT to guide antimicrobial duration included mostly medical patients 5 9 , although peritonitis is one of the most habitual reasons to admit a patient in a perioperative situation to the surgical ICU. To the best of our knowledge, few studies focused on PCT in peritonitis, with its very heterogeneous severity criteria 11 16 17 .

As a first step to evaluate PCT as a tool to discontinue antibiotics, the aim of the present study was to assess whether the PCT thresholds of 0.5 ng/ml or its drop from the peak of at least 80%, previously reported in medical patients, could predict the response to initial treatment in surgical patients admitted for an intra-abdominal infection with septic shock. Our hypothesis is that the PCT kinetic may be associated with the patient’s response to the initial treatment.

During the study period, among the 1,692 patients admitted to our ICU, 101 consecutive patients meeting the inclusion criteria were included (Figure 1). Patients who were discharged within two days after ICU admission were not analyzed further (Figure 1). Patients’ demographics are presented in Table 1. Surgical procedure was necessary in 87% of the patients and all were treated with vasopressors and mechanical ventilation within the first 24 hours of ICU stay. Microbiology culture is presented in Table 2. No differences were noted according to the treatment response. Initial treatment failed in 36 patients (Table 3). The main cause of failure was relaparotomy or radiological drainage for additional source control in 17 cases (47%) or death related to the initial infection in 14 cases (39%). Four patients in the treatment success group needed a relaparotomy or a radiological drainage for hematoma (n = 2) or a superinfection suspicion (n = 2) but no infection was diagnosed by this second look. Antibiotic spectrum modification because of treatment failure or nosocomial infection occurred in 14 patients in the treatment failure group (39%) vs. 5 in the treatment success group (8%), P = 0.003. At day 28, treatment response was significantly associated with mortality (Table 3).

Categorical data are expressed as number and percentage. Continuous data are expressed as mean and standard deviation or median and quartiles. Comparisons were made between patients in whom the first-line treatment succeeded and patients in whom the first-line treatment failed. COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; NYHA, New York Heart Association; SAPS II, severity acute physiology score II 23: SOFA, sequential organ failure assessment 24.

Categorical data are expressed as number and percentage.

Continuous data are expressed as mean and standard deviation. Categorical data are expressed as number and percentage. ICU, intensive care unit.

Upon ICU admission, PCT in patients with subsequent failure of initial treatment was double than in those with successful initial treatment, P = 0.04. We assessed whether the PCT thresholds of either 0.5 ng/ml or its relative drop of at least 80% from the peak could predict the treatment response. Among the 101 patients, six patients (five in the success and one in the failure group) had less than three PCT measurements for logistical reasons and could therefore not be analyzed adequately. Almost 50% of cases responded successfully to treatment even if PCT concentration was constantly above 0.5 ng/ml during the ICU stay (Figure 2A). A decrease of at least 80% compared to the peak was not associated with treatment success (Figure 2B). When PCT decreased below 0.5 ng/ml and its drop was equal to or greater than 80% of its peak, 95% did respond successfully to treatment. Interestingly, when PCT decrease remained superior to 0.5 ng/ml and above 80% of its peak, almost 50% of the patients also responded successfully to the initial treatment (Figure 2C). Using the threshold of 0.5 ng/ml, PCT predicted the treatment success with a sensitivity of 48%, specificity of 94%, positive predictive value of 94%, negative predictive value of 52% and accuracy of 65%. Using a drop equal to or greater than 80% of the peak, PCT predicted the treatment success with a sensitivity of 63%, specificity of 43%, positive predictive value of 66%, negative predictive value of 41% and accuracy of 56%. Among the 35 patients admitted for a post-operative intra-abdominal infections, we found similar performance values for PCT without significant differences compared to patients admitted for community-acquired intra-abdominal infection.

However, when the PCT kinetic between admission and day 5 was considered, it was not significantly different according to treatment response (Figure 3).

We then examined the relation between temperature, CRP and SOFA kinetics and the treatment response. Neither temperature nor CRP drops between admission and day 5 were different according to the patient’s response to treatment (Figure 3A and 3B). Interestingly, SOFA score drop was superior in patients who responded successfully to the treatment compared to patients who did not (P <0.001) (Figure 3D).

The present study reports that neither PCT threshold of 0.5 ng/ml nor its decrease of at least 80% from its peak value could accurately predict the treatment response in a subpopulation of intra-abdominal cases with septic shock. Although 95% of patients in whom PCT decreased below 0.5 ng/ml responded successfully to treatment, 50% of the patients in whom PCT remained superior to 0.5 ng/ml also responded positively to the treatment making this threshold of 0.5 ng/ml a specific but not a sensitive biomarker. Decrease of at least 80% from the peak was not associated with treatment response. In perioperative medicine, PCT may be used to tailor antibiotic therapy using either an absolute cutoff (for example 0.5 ng/ml) or a significant drop compared to the peak value (for example 80%) 12 . In the present study, we examined whether those thresholds were associated with initial treatment response in perioperative critically ill patients admitted for an intra-abdominal infection with shock.

PCT was initially used as a diagnostic biomarker. However, high interindividual differences, failure of a single measurement to accurately identify infection 27 29 and false positive cases have been reported 30 . Recent meta-analyses and reviews concluded that PCT cannot reliably differentiate infectious from noninfectious causes of inflammation in critically ill patients 30 . Moreover, PCT threshold to differentiate bacterial infection vs. inflammation is commonly higher in perioperative medicine than in medical patients 31 . Another way to consider PCT interest in critically ill patients is to use it as a guide to discontinue the antibiotic therapy earlier. Studies that focused on PCT as a tool to withdraw the antibiotics earlier used both an absolute value of PCT (from 2 to 0.25 ng/ml) and a significant drop from the peak value (either a significant reduction of 25 to 35% in three to five days or a reduction of up to 90% compared to the peak) 5 7 9 10 11 32 . In the present study, we studied, as recently suggested 9 12 whether an absolute drop down to at least 0.5 ng/ml or a significant reduction of at least 80% compared to the peak value was associated with the success of the initial treatment for septic shock related to an intra-abdominal infection. We reported that a PCT threshold of 0.5 ng/ml was specific but neither sensitive nor accurate and that a PCT decrease of at least 80% from its peak was not associated with the patient’s response to treatment (Figures 2, 3). Duration of antibiotics in this subpopulation could not be recommended based on PCT level whatever the threshold used.

This finding contrasts with studies on ventilator-associated pneumonia prognosis 8 and with the main study that focused on the PCT kinetic as a prognostic biomarker in nonselected critically ill patients. However, those studies focused on pneumonia or on medical patients, including less than 20% of patients needing surgery although higher PCT ranges might be observed in surgical patients 31 .

Studies focusing on the PCT static threshold value (for example 0.5 ng/ml) to predict the patient’s outcome in severe intra-abdominal infections are sparse. In a study combining 246 cases of sepsis, severe sepsis or septic shock secondary to peritonitis, PCT was associated with patient’s survival in the ICU 17 . Conversely, another study reported that PCT of 16 ng/ml discriminated survivors vs. nonsurvivors with a positive predictive value of only 30% in secondary peritonitis 16 . To the best of our knowledge, the present study is the first to focus on a homogenous group with all patients presenting septic shock related to an intra-abdominal infection. We report that, upon admission, the PCT absolute value might be an indicator of treatment failure (Figure 3) but a PCT decrease of at least 80% from the peak or its kinetic from admission to day 5 could not accurately predict treatment response (Figure 2B). One study reported that PCT but not APACHE II day 2/day 1 ratio was associated with the treatment response in a case mix with an unknown ratio of shock 21 . Another single center study prospectively included 37 patients (including 21 pneumonia cases) with a PCT peak value above 10 ng/ml 33 . The authors reported that the PCT kinetic between day 0 and day 5 was not better than the SOFA kinetic to predict ICU mortality. We focused on initial treatment response rather than on mortality but our results are consistent with that study, highlighting that organ failure score (SOFA score in the present study) might be as helpful as other biomarkers to assess response to treatment in critically ill patients.

The present study has some limits. First, it is a single-center observational study. Second, to distinguish treatment failure from treatment success, we considered that failure of treatment could be either deaths related to infection or infectious surgical complications. Although it may appear cumbersome to cumulate death with other complications, these criteria have been extensively used in studies on antibiotic evaluation. Patients with a length of stay in the ICU of less than two days were not analyzed to avoid fulminant septic shock. Third, we did not perform receiver operating characteristic curves as the aim of the present study was to assess whether published thresholds could be used in intra-abdominal infection with shock rather than to determine new thresholds in another population of interest. Finally, PCT could remain persistently above 0.5 ng/ml because of other reasons than abdominal infection but treatment response was assessed based solely on the abdominal infection course.

In this cohort of 101 perioperative cases with septic shock consecutive to an intra abdominal infection, we report that, PCT threshold of 0.5 ng/ml or its decrease of at least 80% from its peak are not accurate markers to predict patient’s response to the initial treatment. Before using PCT to discontinue antibiotic therapy in surgical patients, further studies evaluating specifically PCT in surgical septic shock are needed.

• Procalcitonin decrease below 0.5 ng/ml was specific but neither sensitive nor accurate to predict treatment response in 101 septic shock patients secondary to an intra abdominal infection

• Procalcitonin decrease of at least 80% from its peak failed to accurately predict treatment response in this population

CRP: C-reactive proteine; ICU: Intensive care unit; IDSA: Infectious diseases society of America; PCT: Procalcitonin; SAPS II: Simplified acute physiology score II; SOFA: Sepsis-related organ failure assessment.

Boris Jung has received speaking fees from Merck, not in relation with the present study. Samir Jaber has received research grants and speaking fees from Maquet, Draeger, Hamilton Medical, Fisher Paykel and Abbott, not in relation with the present study. The other co-authors declare they have no competing interests.

BJ, SJ and GC designed the study, analyzed the results and wrote the manuscript. MN and ZH collected data and analyzed the results. HJP analyzed the results and helped in the interpretation of the microbiological data. FP made corrections to the manuscript and surgical thoughts. NM analyzed the results and performed the statistical analysis. All authors read and approved the final manuscript.