Introduction

The diagnosis of ventilator-associated pneumonia (VAP), a complication that often arises in patients receiving mechanical ventilation, is difficult. Clinical suspicion alone is not reliable [1] and may lead to an overestimation of VAP, resulting in unnecessary antibiotic therapy. Results from invasive procedures, such as quantitative culture of bronchoalveolar lavage fluid (BALF), have proven to be useful [2], but these are not available for 24–48 h. A potential candidate as a rapid diagnostic test for VAP is soluble Triggering Receptor Expressed on Myeloid cells-1 (sTREM-1). Testing for sTREM-1 in BALF has promising results with a sensitivity of 98% and a specificity of 90% [3].

TREM-1 is shed from membranes of activated phagocytes and is found in soluble form in body fluids. It is a glycoprotein, expressed as receptor on neutrophils and CD14high monocytes/macrophages [4]. The expression of TREM-1 is strongly up-regulated in tissues infected with extracellular microorganisms [5]. In humans, sTREM-1 in the blood proved to be a good biomarker for sepsis [6]. For pneumonia, however, the value of sTREM-1 in BALF [3, 7–10] has received varying assessments. To evaluate the usefulness of sTREM-1 as rapid diagnostic test for VAP, we tested 240 BALF samples from intensive care unit (ICU) patients.

Materials and methods

Patients

This is a retrospective study that included patients admitted to the ICU of Maastricht University Medical Centre, who underwent BAL while receiving mechanical ventilation. In all patients, mechanical ventilation was initiated on the day of admission. Patients were screened daily for VAP using clinical criteria, and BAL was performed on the day of clinical suspicion. For each patient receiving more than one BAL, the BALs were included in the study when the time period between BALs was ≥2 weeks; when there was <2 weeks between BALs, only the first BAL was included.

Baseline criteria were recorded (Table 1). Time between ICU admission and BAL was also determined. Study protocols were approved by the institutional review board for human studies. Informed consent regarding the use of data for research purposes was obtained from patients or their legal representatives.

Table 1 Characteristics of the study population
Full size table

VAP diagnosis

Clinical suspicion of VAP was according to the definition of Bonten et al. [11], and VAP was considered to be acquired upon manifestation of the disease after ≥48 h of mechanical ventilation. Clinically suspected VAP was considered confirmed VAP when the microbiological results met the following criteria [11, 12]: presence of ≥2% cells containing intracellular organisms (ICO) and/or quantitative culture result of ≥104 colony forming units (cfu) per millilitre of BALF.

Following BAL, antibiotic therapy was started empirically in patients who had not previously received treatment. When patients developed a VAP while receiving antibiotics, the therapy was adjusted. Antibiotic treatment before and during BAL does not influence the predictive value of the percentage ICO in BALF in diagnosing VAP [12]. Thus, due to the combination of quantitative BALF culture and presence of ICO, adjustment of the cfu/ml threshold was unnecessary to diagnose VAP correctly when antibiotics were being used before the BAL.

Two researchers (Guy J. Oudhuis and A. Verbon) retrospectively determined whether patients met the criteria of clinically suspected and confirmed VAP, independently from one another and blind to the sTREM-1 assay results. This yielded two study groups; a confirmed VAP group and a group in which VAP could not be confirmed. A consensus diagnosis was achieved in all cases.

Bronchoalveolar lavage

Bronchoscopy with directed BAL was performed as described previously [12]. Samples were processed within 1 h after the BAL. All aliquots were pooled, except for the first one. The BALF samples were Gram-stained, and quantitative culture was performed as described previously [13]. The urea concentration was determined by means of an enzymatic conductivity rate method using the SYNCHRON LX System (Beckman Coulter BV, Mijdrecht, The Netherlands) as previously described [14]. The remaining BALF was centrifuged at 250 g for 10 min. The supernatant was stored at –80°C in six different aliquots until further processing.

Soluble Triggering Receptor Expressed on Myeloid cells-1 assay

A quantitative sandwich enzyme immunoassay (Quantikine Human TREM-1 Immunoassay; R&D Systems, Minneapolis, MN), was used to detect sTREM-1 in BALF in accordance to the manufacturers’ instructions, as described previously [7].

The measured concentration of sTREM-1 in BALF samples is not identical to that in the epithelial lining fluid (ELF), secondary to dilution of the BALF. To calculate the concentration sTREM-1 in ELF from the concentration of sTREM-1 in BALF, we applied a correction with a dilution factor using the ratio of urea concentration in the serum and BALF [15].

Statistical analysis

The Mann–Whitney U test was used for the numerical data; for sTREM-1 concentrations, we used Student's t test to compare logarithmic means. The Pearson χ2 test was used for categorical data. The analyses were corrected, by clustered logistic regression, for the influence of multiple BALs performed in the same patient. To evaluate the diagnostic value of the sTREM-1 assay, we produced a receiver-operating characteristic (ROC) curve and measured the area under the curve. Statistical significance was defined as a P value of <0.05 in all cases. Analysis was performed with SPSS ver. 14.0 for Windows (SPSS, Chicago, IL).

Results

Patients

Between January 2001 and October 2006, 361 BALs were performed, of which 59 were excluded from our retrospective analysis since they did not meet the quality criteria [12], and 62 were excluded because they were performed within 48 h after ICU admission. The remaining 240 BALs, performed in 207 patients were included in our study.

Seven patients underwent three BALs, 19 patients underwent two BALs and 181 patients had one BAL. The presence of VAP was confirmed in 97 cases (40.4%) and not confirmed in 143 cases (59.6%).

Confirmed VAP cases did not differ significantly from unconfirmed cases, and both groups did not differ significantly in terms of in-hospital mortality and ICU mortality (Table 1). The number of patients with chronic obstructive pulmonary disease (COPD) was low and did not differ between both study groups (Table 1).

Soluble Triggering Receptor Expressed on Myeloid cells-1 levels

The mean sTREM-1 concentration was significantly higher in the BALF of patients with confirmed VAP [1849 pg/ml, 95% confidence interval (CI) 1515–2256 pg/ml] than in those without confirmed VAP (1424 pg/ml, 95% CI 1218–1664 pg/ml; P = 0.04; Fig. 1a).

Fig. 1
figure 1

a Mean soluble Triggering Receptor Expressed on Myeloid cells-1 (sTREM-1) levels in bronchoalveolar lavage fluid (BALF) samples of confirmed (n = 97) and non-confirmed ventilator-associated pneumonia (VAP) (n = 143) cases. Vertical bars 95% confidence intervals of the mean. b Receiver-operating characteristic (ROC) curve analysis of sTREM-1 in BALF samples of patients with and without confirmed VAP

Full size image

Mean sTREM-1 levels did not differ between patients with COPD (1570 pg/ml, 95% CI 1081–2279 pg/ml) and those without (1617 pg/ml, 95% CI 1404–1863 pg/ml) (P = 0.93).

No significant difference in mean sTREM-1 level was observed between medical patients (1676 pg/ml, 95% CI 1352–2077 pg/ml) and surgical patients (2233 pg/ml, 95% CI 1769–2817 pg/ml) (P = 0.08), while the number of confirmed VAP cases did not differ between both groups (P = 0.17).

Diagnostic value of sTREM-1

A ROC curve analysis showed an area under the curve of 0.58 (95% CI 0.50–0.65; P = 0.04; Fig. 1b). Therefore, the sTREM-1 levels in our hands were not discriminative for VAP. Choosing a sensitivity of 95% resulted in a positive predictive value (PPV) of 41% and a negative predictive value (NPV) of 62% in our population. A specificity of 95% led to a PPV of 67% and a NPV of 62%.

Discussion

In this study, sTREM-1 levels were not discriminative for VAP. The differences between the results of our study and those of five other trials in terms of sTREM-1 and pneumonia are highlighted in Table 2. These may explain the differences in results. A correction for the dilution of BAL has been found to allow a better representation of the actual concentration of sTREM-1 in the alveoli [15]. However, this does not appear to explain the major differences in outcome, since using the concentration of sTREM-1 without correction for dilution did not result in a significantly better ROC analysis in our study (data not shown).

Table 2 Characteristics of the six clinical studies regarding sTREM-1 and pneumonia
Full size table

Samples from medical and surgical patients were included in this study. Our results show that the sTREM-1 levels tended to be higher in surgical patients, which is in accordance with the results of other studies in which surgery was reported to increase sTREM-1 expression levels in peripheral blood monocytes independent of infection [16]. We therefore conclude that in terms of diagnosing VAP in an every day ICU setting, the measurement of sTREM-1 levels in BALF does not seem to be discriminative for VAP.

The levels of sTREM-1 are not only increased by infecting micro-organisms [17]; cancer cells have also been shown to be able to directly up-regulate TREM-1 expression in patients’ macrophages [18]. A number of inflammatory disorders, such as pancreatitis, are also known to increase levels of sTREM-1 [19]. In contrast, TREM-1 is not up-regulated in samples from patients with other inflammatory disorders, such as psoriasis, ulcerative colitis or vasculitis caused by immune complexes [5].

The strength of this study is that it was performed in a large, well-defined group of both medical and surgical ICU patients. One limitation may be that the analysis was performed retrospectively. Only patients that underwent BAL ≥48 h after being admitted to the ICU were included in the study. Therefore, these patients stayed longer in a ICU than the general ICU population. A second limitation may be that VAP is still difficult to diagnose. However, we used the widely accepted modified criteria of the Centers for Disease Control [11] to diagnose VAP and confirmed the diagnosis by both staining for ICO and culturing BALF samples [12].

Conclusion

In conclusion, the results of this study imply that the sTREM-1 assay on BALF samples as a rapid diagnostic test for VAP may not be discriminative.