Introduction

Pneumonia is the most frequent infectious complication observed in patients successfully resuscitated from a cardiac arrest (CA). Up to one half of them will further develop a lower respiratory tract infection, mainly due to aspiration, which is associated with an increased mechanical ventilation duration and intensive care unit (ICU) length of stay [1–3]. Moreover, the use of therapeutic hypothermia is potentially associated with an increased incidence of infectious events. However, an accurate early diagnosis remains difficult to perform in this specific setting because of many confounding factors. First, clinical and biological signs of systemic inflammatory response syndrome (SIRS) can be provoked either by an authentic infection or by the “post-resuscitation disease” that mimics sepsis [4, 5]. Second, chest X-ray abnormalities are usual, but mostly related to pulmonary edema and/or atelectasia. Finally, routine use of therapeutic hypothermia deprives clinicians of major infectious criteria, like fever or white blood count (WBC) changes. All together, these confounding factors may explain that the usual clinical diagnostic approach is challenged in face of a common and potentially severe complication. Thus, physicians are dealing with a dilemma: waiting for clinical, radiological, biological evolution and bacteriological results (with the risk of degradation of an already severely ill patient) or treating any potential infection (with the risk of overusing antibiotics).

Reliable sepsis biomarkers could be helpful in this early recognition challenge. Among them, procalcitonin (PCT) has been shown to be superior to C-reactive protein (CRP) and is widely used as a marker of bacterial sepsis [6, 7]. Its kinetics are adapted to the critical care setting, with an early increase after septic injury and a fast decrease after infection control. Early detection in PCT increase could allow time sparing on usual criteria or culture of the endotracheal aspirates [8]. To our knowledge, only one study has investigated the value of PCT for diagnosis of ventilator-associated pneumonia in a small number of survivors of CA who did not receive therapeutic hypothermia [9]. Furthermore, there is a paucity of data about this marker in post-resuscitation shock management, which has recently been acknowledged as a cornerstone after CA resuscitation [10]. Thus, we designed this study to assess the ability of serum PCT measurements to detect early onset pneumonia after CA.

Materials and methods

Study setting and population

All consecutive patients over 18 admitted to our 24-bed medical ICU between July 2006 and March 2008 after a successfully resuscitated CA were studied. We reviewed retrospectively all medical records and data from our prospective ICU database, in which all CA survivor characteristics are registered according to the Utstein style [11]. The following variables were recorded prospectively for each patient: demographic data, clinical parameters, cause of CA, arrest location, no-flow and low-flow period, initial rhythm, Simplified Acute Physiology Score 2 (SAPS 2), hypothermia management, development of infection, major biological parameters and ICU mortality. Post-resuscitation shock was defined as a need for vasopressors (epinephrine or norepinephrine) lasting more than 6 h despite adequate fluid loading. Patient management was strictly standardized (see the Electronic Supplementary Material). Patients who died within the first 24 h, with a known infection prior to CA, with an extra-pulmonary infection developing within 5 days following admission and patients with missing data or incomplete samples were excluded.

Data analysis

All files were retrospectively reviewed by two independent investigators (NM and VL) in order to check the diagnosis of early onset pneumonia [P(+)] or not [P(−)] in the light of clinical, biological, microbiological and radiological data of the first 5 days of ICU stay. Each case of disagreement between prospective diagnosis and retrospective assessment was resolved by consensus between the investigators, with the help of a third expert (AC) if necessary. At the time of this assessment, investigators were blinded to CRP and PCT levels. According to commonly used criteria, early onset pneumonia was defined by the presence of a clinical compatible finding at auscultation and a new pulmonary infiltrate on chest X-ray (persistent for at least 48 h) associated with a positive quantitative culture of the endotracheal aspirates (threshold: 106 CFU/ml). In the absence of a bacteriological sample, the diagnosis was retained when the previous signs were associated with purulent endotracheal aspirates and hypoxemia (PaO2/FiO2 < 200) not explained by pulmonary edema or atelectasia. The main components of SIRS, i.e., body temperature, heart rate and WBC, were not considered since they are affected by hypothermia. Finally, the diagnosis of early onset pneumonia was considered when occurring within 5 days following ICU admission.

Blood sampling

All CRP and PCT dosages were performed in May 2008 by the same investigator (SG) by using blood samples collected at admission, day 1 (D1), day 2 (D2) and day 3 (D3). These samples were initially centrifuged and stored at −80°C within 4 h, as approved by our local institutional review board, as part of a serum collection. Analyses of CRP were performed with a fully automated immunoturbidimetric assay (CRPLX, Modular PP®, Roche Diagnostic, Mannheim, Germany), with a low detection limit of 1 mg/ml. PCT concentrations were quantified with an immunofluorimetric assay (PCT sensitive, Kryptor®, Brahms, Berlin, Germany). Its analytical assay sensitivity was 0.02 ng/ml, according to the manufacturer.

Statistical analysis

Continuous variables were expressed as medians (with interquartile range), and categorical variables were reported as count and proportions, unless specified otherwise. Statistical significance of differences within and between groups was evaluated as follows: continuous variables with the t test or Wilcoxon rank sum test, as appropriate; categorical variables with the χ² test or the Fischer’s exact test, as appropriate. Receiver-operating characteristic (ROC) curves were performed to assess the ability of PCT concentrations to identify early onset pneumonia. Statistical significance was defined as p < 0.05. Analyses were performed using Stata 7.0 software (Stata Corp., College Station, TX).

Results

During the 20-month study period, 245 patients were admitted after being successfully resuscitated from a CA. Forty-eight died within the first 24 h, and 54 other patients with missing data or incomplete samples were excluded. Furthermore, 11 patients presented an active infection at the moment of CA or experienced an extra-pulmonary infection within 5 days following admission. Consequently, we included 132 consecutive patients.

The patient population had a median age of 60 (49–70) years, and 92 (69%) were men. CA occurred in the hospital for eight patients (6%). Most were cardiac etiologies (80 patients, 61%), followed by respiratory (31 patients, 23%) and neurological (6 patients, 4%) failures; the etiology was uncertain for 15 patients. The initial rhythm was asystole or pulseless electrical activity for 75 patients (57%) and ventricular tachycardia/fibrillation for 57 patients (43%). Hypothermia was instituted in all but one patient. The ICU mortality rate reached 55% (73 patients).

Early onset pneumonia was retrospectively established in 86 patients (65% of this population), with microbiological documentation in 72 patients. Responsible microorganisms are listed in the Electronic Supplementary Material. Antibiotherapy was initiated in 115 patients during the first 5 days of ICU stay: 98% of the patients were P(+), and 67% of the patients were retrospectively considered to be P(−) (p < 0.001). Characteristics of P(+) and P(−) patients are summarized in Table 1. Higher SAPS 2 and an increase in cardiac etiology in P(+) patients were the only significant differences.

Table 1 Baseline admission characteristics and outcomes of patients admitted for successfully resuscitated cardiac arrest, classified according to presence [P(+)] or absence [P(−)] of early onset pneumonia
Full size table

Median serum PCT values in P(+) and P(−) patients were respectively (Fig. 1): 0.38 ng/ml (0.12–2.56) versus 0.18 (0.11–0.81) at admission (p = 0.051), 4.58 (0.77–21.86) versus 1.03 (0.45–4.68) at D1 (p = 0.017), 3.76 (0.82–25.6) versus 0.73 (0.4–4.4) at D2 (p = 0.002) and 3.76 (0.82–25.64) versus 0.73 (0.42–4.4) at D3 (p = 0.046). The PCT peak value was significantly higher in P(+) patients [5.9 (1.2–34.3) versus 1.6 (0.65–6.1) in P(−) patients (p < 0.001)]. The areas under the ROC curves of PCT for identification of early onset pneumonia were 0.59 (95% CI 0.49–0.69) at admission, 0.64 (95% CI 0.54–0.74) at D1, 0.68 (95% CI 0.57–0.79) at D2 and 0.63 (95% CI 0.49–0.76) at D3 (Fig. 2). Considering the kinetics of this biomarker, we also assessed the diagnostic value of the delta between PCT level measured at admission and PCT levels measured at D1 and D2. The areas under the curves of the delta were respectively 0.62 and 0.5 for admission D1 and for D1–D2. Predictive values of PCT to diagnose pneumonia were poor, despite determination of a threshold of 0.5 and 5 ng/ml (Table 2).

Fig. 1
figure 1

Serum PCT levels on admission, D1, D2 and D3 in patients experiencing early onset pneumonia [P(+), grey boxes] or not [P(−), white boxes]. PCT levels are expressed in box plot forms (median, interquartile range). Values are expressed in ng/ml according to a logarithmic scale

Full size image
Fig. 2
figure 2

ROC curves comparing the ability of PCT concentrations to identify early onset pneumonia at admission, D1, D2 and D3

Full size image
Table 2 Predictive values of PCT to diagnose early onset pneumonia according to two different thresholds in all patients and in patients with post-resuscitation shock
Full size table

CRP levels and WBC values were not significantly different between the patients with and without pneumonia (Table 3). Areas under the ROC curves of CRP were poorly informative, with a value of 0.55 (95% CI 0.41–0.66) at admission, 0.51 (95% CI 0.39–0.64) at D1, 0.50 (95% CI 0.37–0.64) at D2 and 0.50 (95% CI 0.37–0.63) at D3 (not shown).

Table 3 C-reactive protein (CRP) and white blood count (WBC) in the presence [P(+)] or absence [P(−)] of early onset pneumonia
Full size table

A post-resuscitation shock was observed in 66 patients and in a similar proportion between P(+) and P(−) patients. PCT levels were higher in patients with shock than in vasopressor-free patients at admission (p = 0.09), at D1 (p < 0.001), at D2 (p < 0.001) and at D3 (p = 0.03) (Table 4). In the subgroup of patients with shock, positive and negative predictive values of PCT were low whatever the cut-off value (Table 2).

Table 4 Serum PCT levels in the subgroups of patients with or without post-resuscitation shock and renal replacement therapy requirements
Full size table

We also examined PCT concentrations in patients requiring renal replacement therapy (n = 50), which was performed with intermittent hemodialysis. They exhibited significantly higher PCT levels from admission to D3 (Table 4). Those requiring dialysis were strongly linked with the development of post-resuscitation shock; 70% (n = 46) of patients with shock required dialysis, and 92% (n = 46) of patients receiving renal replacement therapy were vasopressor-dependent (p < 0.0001).

Discussion

To our knowledge, this is the first study designed to evaluate the potential usefulness of PCT determination for diagnosis of early onset pneumonia in successfully resuscitated CA patient undergoing therapeutic hypothermia. We found that PCT was significantly higher in patients with confirmed pneumonia in the early post-resuscitation period. However, sensitivity, specificity, and positive and negative predictive values of PCT were quite poor and barely superior to CRP, despite attempts to remove potential flaws in this setting. PCT cut-off values for the diagnosis of infection in critically ill patients are still to be determined, ranging from 0.44 ng/ml [12] to 9.7 ng/ml [13], according to the medical or surgical ICU setting, as well as the severity of the septic insult. As a result, we calculated the diagnostic value according to a low (0.5 ng/ml) or high (5 ng/ml) threshold. This failed to improve diagnostic utility. PCT kinetics have been proposed to circumvent the drawback of assigning a specific threshold [12, 14]. However, in our study, the use of PCT kinetics between admission, D1 and D2 did not significantly improve the diagnostic accuracy of this biomarker. As renal dysfunction affects many of survivors of CA and is known to interfere with the diagnostic accuracy of PCT [15], we investigated this specific point. Renal replacement therapy was used as a surrogate of acute kidney injury [16] and was associated with significantly higher PCT concentrations. However, patients requiring renal therapy were also likely to present post-resuscitation shock, thus making it difficult to distinguish the respective role of acute kidney injury, hemodialysis and shock on PCT levels.

Our results contrast with the recent increasing use of PCT, which is widely reported as a useful biomarker of infection in critically ill patients in many indications: to diagnose infection [7, 17], to guide the antibiotic treatment decision [18, 19] or duration [20]. Moreover, PCT was also proposed to predict outcome by examining the absolute values [21], or kinetics [22].

The high rate of antibiotic initiation in our cohort, especially in patients without confirmed pneumonia, is noteworthy. Early overuse of antibiotics may be explained by clinical, biological and radiological confounding factors in survivors of CA treated with therapeutic hypothermia, added to the preoccupation of missing a septic insult in such severe patients. This reinforces the need for complementary tools to improve diagnosis of early onset pneumonia and justifies our study using PCT in this context. However, to our knowledge, only four studies have investigated the interest of PCT after CA, and none of them focused on the PCT value in aspiration pneumonia. Fries et al. [23] studied PCT concentrations in the first 3 days in 23 patients, showing that they were higher in patients with impaired neurological outcome. Remarkably, neither infection nor SIRS was diagnosed in their population. Post-resuscitation shock was also not a clearly-defined entity. These authors recently proposed an investigation of inflammatory biomarkers, including PCT, after CA treated with hypothermia [24]. They draw the same conclusion on the relationship between a high PCT level and impaired neurological evolution. Unfortunately, a high infection rate may have influenced the marker measurements. In line with our study, Oppert et al. [9] reported that an elevated PCT value greater than 1 ng/ml diagnosed pneumonia with a sensitivity of 100% and a specificity of 75% during the first week. Interestingly, elevation of PCT preceded the pneumonia diagnosis with a median of 2 days, but the study was performed in only 28 patients, extended to late-onset pneumonia with a debatable definition of pneumonia and without daily measurement of this marker or exclusion of extrapulmonary infections. Consequently, results of these three studies do not allow any firm conclusion about the diagnostic value of PCT after CA. Finally, Adib-Conquy et al. [25] compared PCT values in patients experiencing cardiac surgery, CA or sepsis. In 54 CA survivors without infection, they showed that PCT levels in patients who died from refractory post-resuscitation shock overlapped those measured in septic patients. Moreover, PCT was higher in patients who ultimately died from neurological failure than in survivors. Even if the authors did not report PCT levels in patients with infection or with post-resuscitation shock, post-CA patients exhibited high levels of PCT (median concentrations of 0.43, 2.19 and 2.4 ng/ml at admission, D1 and D2). Notably, these values are slightly above the concentrations observed in our P(−) cohort. This could be explained by the fact that all patients reaching ICU were included (even those who died in the early ICU period), preventing from getting the time to rule out infectious complications. The authors concluded that PCT was more likely a marker of severe inflammation than infection.

Interestingly, we also found that patients with post-resuscitation shock had higher PCT levels, suggesting that PCT could be a marker of the systemic response following CA that correlates with clinical severity. SIRS, which is commonly observed in these patients, could explain the lack of predictive value and specificity of PCT after CA. This hypothesis is supported by similar observations performed in other settings known to provide SIRS. For instance, Geppert et al. [26] reported than high concentrations of PCT were also encountered in cardiogenic shock, suggesting that hypoperfusion provoked by myocardial and endothelial dysfunctions could be a supplementary cause of elevated PCT. In fact, SIRS occurs very early in the course of resuscitated CA and with a variable intensity [5]. Post-resuscitation disease is also characterized by immunologic, endothelial and cardiovascular dysfunctions sharing many features with sepsis. The main hypothesis is related to the activation of endotoxin and cytokine pathways [27]. Endotoxin, which is a strong trigger for PCT synthesis [28], is released as a consequence of splanchnic hypoperfusion and bloodstream translocation in nearly half of post-CA patients. Furthermore, high levels of pro-inflammatory cytokines may also participate in the non-infectious increase in PCT level [27, 29]. As suggested by two recent meta-analyses, PCT cannot accurately differentiate sepsis from other non-infectious causes of SIRS and precludes the recommendation for its routine use as a tool for septic screening, especially following CA [30, 31]. The pro-inflammatory, non-infectious pattern of PCT elevation prevents its use in this setting.

Even if performed in a large cohort of post-CA patients, some limitations of our study deserve careful consideration. First, we considered a retrospective single-institution cohort. However, all analyzed data were prospectively collected, and the monocentric design led to a homogeneous strategy. Second, diagnosis of early onset pneumonia lacks a consensus strategy, which may over- or underestimate the true incidence. We should admit that our findings regarding PCT levels might have been different if diagnosis had relied on other criteria. Nevertheless, we employed a robust methodology to minimize this risk as all files were reviewed by two separate investigators who were unaware of the biomarker levels. The CPIS score, widely used in the assessment of pneumonia in mechanically ventilated patients, cannot be used in survivors of CA. This is because it takes into account the temperature and WBC [32], which are both artificially modified by therapeutic hypothermia. Furthermore, its lack of diagnostic accuracy has been emphasized [33, 34]. The routine use of quantitative culture of endotracheal aspirates in our patients can also be discussed. However, the best microbiological tool is still a matter of debate, and its discussion is beyond the scope of this manuscript. Third, we did not include patients who died within the first 24 h as refractory shock is the main cause of early death in this setting [35]. We can speculate that PCT, as a marker of the inflammatory response, would have been very high in this population whatever the infectious status and would have reinforced our results. Moreover, it increased our ability to confirm or exclude infection, whatever the source. Fourth, the utility of PCT was evaluated as a single parameter. We did not determine if the combination of PCT level with clinical, biological and radiological data would improve the diagnostic value. In the same way, one may argue that comparing the PCT levels of the suspected (n = 115) with the non-suspected patients (n = 17) would have more closely reflected daily use of this biomarker; realistically, this would have more dramatically impaired its real accuracy by mixing confirmed and refuted pneumonias. Finally, this study was performed in patients undergoing therapeutic hypothermia. The relationship between inflammatory cytokines or biomarkers and the effect of therapeutic hypothermia is controversial [24, 36]. In our cohort, there was no significant difference in hypothermia use between the P(+) and P(−) groups, so that it did not prevent comparison. Fries et al. [24] demonstrated that hypothermia blunted the PCT elevation in patients with bad neurological outcome, whereas it did not affect levels in patients with good recovery. Interpretation of this finding is limited by the small number of patients and by the observational study design, making it difficult to precise the impact of hypothermia on PCT concentrations. As a result, our conclusions cannot be extrapolated to patients not undergoing therapeutic hypothermia. Moreover, as implementation of therapeutic hypothermia is still low [37] despite compelling evidence [10], further studies should involve patients undergoing therapeutic hypothermia or not to reflect more real-life practices.

To conclude, in patients successfully resuscitated from CA, serum PCT levels measured during the first 3 days were significantly higher in patients with early onset pneumonia. However, this biomarker was associated with a poor sensitivity, specificity, and positive and negative predictive values. At this time, PCT levels should not be used to assess early onset pneumonia after CA undergoing therapeutic hypothermia. Post-resuscitation shock, which is present in a large subset of these patients, could contribute to the PCT elevation and may affect its diagnostic accuracy.