C-reactive protein and procalcitonin as predictors of survival and septic shock in ventilator-associated pneumonia

Study design and population

The study was conducted at our ICU (Sotiria Chest Hospital, Athens, Greece), harbouring a population of mostly medical patients, over a 18-month period (April 2006 to September 2007). All patients consecutively admitted to the ICU suspected of VAP were eligible. The next of kin provided the informed consent for each patient included. The local ethics committee approved the study.

Initially, 54 patients with VAP were screened. Patients with community-acquired pneumonia as a cause of ICU hospitalisation (n = 3), patients with extrapulmonary infection (n = 1) as well as immunocompromised patients (haematological malignancies, HIV, neutropenia <1,000 cells·mL⁻¹, patients who had received chemotherapy within the preceding 45 days; n = 2) were excluded from the study. Patients who died within the first 3 days after VAP diagnosis (n = 3) were also excluded from the study.

Finally, 45 patients ≥18 yrs (34 male and 11 female, mean age±sd 61.5±17.8 yrs) who developed VAP were enrolled in the study. VAP was defined as the occurrence of newly developed lung infiltrates, occurring ≥48 h after initiation of mechanical ventilation and persisting for ≥72 h, plus two of the following three criteria: 1) fever >38.2°C; 2) leukocytosis >12,000 mm⁻³; 3) purulent endotracheal secretions 15.

Additionally, microbiological documentation was necessary with the growth of ≥1×10⁴ colony-forming units (cfu)·mL⁻¹ of a microorganism in bronchoalveolar lavage (BAL), or ≥1×10³ cfu·mL⁻¹ in protected brush specimens, or ≥1 × 10⁶ cfu·mL⁻¹ in endotracheal secretions and/or the isolation of a pathogen from blood cultures 5, 6, 16, 17. Data collected included admission diagnosis (table 1⇓), past medical history and vital signs. In addition, the following were evaluated daily: clinical examination, presence or absence of organ dysfunction(s) and/or infection, temperature, white blood cell count, blood chemistry, arterial oxygen tension/inspiratory oxygen fraction and chest radiographs.

Patients were evaluated daily for evidence of VAP. After the establishment of VAP diagnosis, all patients received empirical antibiotic treatment. The day of VAP clinical diagnosis was defined as D1 and was the same day that empirical antibiotic treatment was started. The following days were accordingly termed as D2, D3, etc. Only the first episode of VAP was evaluated in each patient. None of the patients had recurrent VAP or extrapulmonary infection during the first 10 days of VAP. VAP recurrence was defined as a new VAP episode, that is new clinical and radiological signs compatible with pneumonia, and included persistent infection (the same pathogen responsible for the first episode), relapse (the same pathogen as in the first episode but after the end of antibiotic therapy) and new infection (another pathogen, at any time).

CRP and PCT levels were measured on D1, D4 and D7 in all subjects included in the study. The Acute Physiology and Chronic Health Evaluation (APACHE) II and Sequential Organ Failure Assessment (SOFA) scores were used to assess disease severity 18, 19. The APACHE II score was calculated during the first 24 h of ICU admission. The SOFA score was evaluated on the same days with CRP and PCT measurements (D1, D4, D7). Patient progress was followed until the 28th day after the diagnosis of VAP, when they were considered survivors. Patients who died before D28 were categorised as nonsurvivors.

The evolution of CRP and PCT concentration throughout the course of VAP was analysed comparing survivors and nonsurvivors and also comparing those who developed septic shock or not. Septic shock was defined according to consensus definitions 20.

Blood samples were collected on D1, D4 and D7. The samples were centrifuged at 500×g for 10 min and the plasma was aliquoted and stored at -70°C until analysed in a single batch. Circulating levels of CRP were measured using an immunoturbumetric method with a commercially available kit (Dade Behring, Deerfield, IL, USA). PCT was determined with chemiluscence (Liaison Brahms PCT; Dia Sorin S.P.A., Saluggia, Italy). Normal values were <0.30 mg·dL⁻¹ for CRP and 0.1–0.5 ng·mL⁻¹ for PCT.

Statistical analysis

Values were expressed as mean±sd or as median (interquartile range 25–75th percentile) in case of a skewed distribution. Comparisons between groups were performed using the Mann–Whitney U-test. Categorical variables were compared with the Chi-squared test. Correlations were performed using Spearman's r-test.

Dichotomized Δ was calculated by the formula: Δ = D4-D1, D7-D4 and D7-D1. Therefore ΔPCT_4-1 = PCT_D4-PCT_D1, ΔPCT_7-4 = PCT_D7-PCT_D4, etc. Δ>0 means increasing values and Δ≤0 means decreasing values. ΔPCT, ΔCRP and ΔSOFA were categorised as increasing or unchanged/decreasing.

A univariate logistic regression analysis was used to define risk factors associated with VAP survival and septic shock development. A multivariate logistic regression analysis model was constructed with either VAP survival or septic shock development as the dependent variables, and CRP and PCT values on D1, D4 and D7, as well as variations in CRP and PCT (ΔCRP_4-1, ΔCRP_7-1, ΔCRP_7-4, ΔPCT_4-1, ΔPCT_7-1, ΔPCT_7-4) as independent variables. To address potential collinearity, models were constructed that included only absolute values or only variations, as well as both absolute values and variations. To control for potential confounding factors, age, sex, APACHE II score and SOFA score were included in the initial model. Results are reported as adjusted OR with 95% CI.

We prospectively validated the previously suggested CRP and PCT kinetics 12 by virtue of the design of our multivariate logistic regression analysis model. Receiver operating characteristic (ROC) curve analysis was performed for PCT and CRP values on the days that these variables differed between survivors and nonsurvivors and between VAP patients who did and did not develop septic shock.

Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR; true-positive rate/false-positive rate or sensitivity/(1-specificity)) and negative likelihood ratio (NLR; false-negative rate/true-negative rate or (1-sensitivity/specificity)) were calculated. Threshold values that gave the best combination of sensitivity and specificity were judged by calculating the Youden's index, i.e. the maximum difference between sensitivity and (1-specificity) 21.

The SPSS statistical package (version 14.0; SPSS, Chicago, IL, USA) was used. A p-value <0.05 was considered significant.

VAP survival prediction

Septic shock development in VAP

During the course of VAP, 22 (48.9%) patients developed septic shock between the second and the fourth days (table 3⇓). The SOFA score was significantly higher in patients with VAP who developed septic shock on all days (table 3⇓). A positive correlation was detected between CRP levels and SOFA score on D1 (r = 0.577, p<0.001) and D7 (r = 0.583, p<0,001). Furthermore, positive correlations were found between PCT levels and SOFA score on D1 and PCT levels and CRP levels on D1 (r = 0.473, p = 0.001 and r = 0.352, p = 0.018, respectively).

CRP and PCT threshold values as predictors of VAP survival

Numerous studies have evaluated the usefulness of CRP 22, 23 and PCT 24, 25 both in the diagnosis and the prognosis of VAP. Póvoa et al. 26 suggested that daily CRP measurements were useful in the identification, as early as D4, of VAP patients with poor outcome. In our study, serum CRP levels on D1, D4 and D7 during the course of VAP did not discriminate survivors from nonsurvivors.

Luyt et al. 11 have suggested that serum PCT levels on D1, D3 and D7 during the course of VAP are strong predictors of unfavourable outcome defined as either death, recurrent VAP or extrapulmonary infection. In our study, we focused on survival as our primary outcome. Although we found that serum PCT levels were significantly higher in nonsurvivors on D1 and D7 than in survivors, similar to previous studies 11, 27, and the AUC was satisfactory (0.79 on D1 and 0.88 on D7, suggesting moderate accuracy), the PLR was always <10 (2.54 on D1 and 5.78 on D7). A PLR of 10 is considered the threshold above which the PLR is considered really important 28, 29. Furthermore, in the multivariate analysis, the PCT levels on either day were eliminated. However, it has to be acknowledged that the PLR of PCT on D7 was 5.78, which can generate moderate shifts in pre-test to post-test probability 28, 29.

CRP and PCT kinetics as predictors of VAP survival

Seligman et al. 12 have suggested that decreasing CRP and PCT values between the onset and the 4th day of VAP could predict survival. In our group of VAP patients, CRP and PCT kinetics between D1, D4 and D7 were not able to predict survival.

The fact that CRP and PCT levels decreased on D7 compared with D1 in most survivors in our study lacks clinically useful prognostic significance. In fact, this result is rather expected, since 7 days after the initiation of antibiotic treatment, VAP has either responded to treatment and progresses towards resolution, or is refractory to it. Decreasing CRP and PCT levels between D1 and D4 would have contained clinically useful predictive information, but this could not be confirmed by our results.

Prediction of septic shock development in VAP

In this study we tested the hypothesis that CRP and PCT levels and kinetics could convey prognostic information for the development of septic shock in patients with VAP. We found that among patients with VAP, serum PCT was significantly higher on D1 and D4 in those who subsequently developed septic shock. Although the AUC suggested moderate accuracy (0.75–0.78), the PLR was again <10 (table 4⇑), implying that the results of the test are not likely to alter clinical decisions 28, 29. Furthermore, in the multivariate analysis, PCT on D1 and D4 could not predict the development of septic shock. Again, the PLR of PCT on D1, 8.36, suggests the potential for useful predictive information.

The only predictor of septic shock development in our study was the SOFA score. Although the SOFA score was significantly higher in VAP patients with septic shock on all days (D1, D4, D7), only the SOFA score on D1 was predictive of septic shock development. To our knowledge, this is the first study showing that the SOFA score on the day that VAP is diagnosed can predict septic shock development.

Critique of methods: limitations

Some limitations of our study should be noted. The number of patients included in the study is rather small, and thus our study was inadequately powered for the multivariate analysis performed. This is a rather common limitation with studies of this type in the field. The inadequate power increases the risk of underfitting (type II error), which could have led to the omission of important predictors from the model 30. However, the performance of the same predictors in the univariate analysis of the ROC curves with the associated likelihood ratios of the best threshold values showed similar results, with moderate predictive performance at best and likelihood ratios always <10, which is the lower limit above which a test can generate large and often conclusive changes from pre-test to post-test probability 29.

Although we made every effort to exclude other causes of systemic infection, measured PCT levels cannot solely be attributed to pulmonary infection. The APACHE II score of our VAP patients was lower than that observed in other studies looking at outcomes. This could have some influence on our results. It should be kept in mind that the majority of our ICU patients are medical critically ill patients.

In our study, the diagnosis of VAP was mainly based upon quantitative endotracheal aspiration cultures, with a minority of patients diagnosed based on the cultures of bronchoscopically obtained material. Although invasive bronchoscopic strategies might be useful in permitting the de-escalation or cessation of unnecessary antimicrobial therapy 31, it is unlikely that the technique we used has significantly influenced our results, given that endotracheal aspirate cultures are associated with similar clinical outcomes when compared with quantitative cultures of BAL 32.

The incidence of septic shock among our VAP patients was relatively high (48.9%). This may partly be attributed to the high number of multidrug-resistant pathogens as causative agents of VAP.

Conclusions: implications

Our findings cannot suggest the routine use of CRP and PCT levels as prognostic markers for the survival or septic shock development in patients with VAP. The PLRs of the PCT on D1 for septic shock development and D7 for survival suggest that the measurement of PCT can provide useful information. However, the context for using this information has to be investigated and our study was not designed to address this issue. For instance, serial serum PCT measurements have been suggested as indicators of the need to change treatment early in patients with VAP, either to intensify treatment when PCT levels remain elevated, or to avoid unnecessary prolonged courses of antibiotics when levels are rapidly decreasing. More studies are needed to address the use of these markers in critically ill patients.