A prospective study was performed of all patients consecutively admitted to a 12-bed medical ICU of Rabat University Hospital between February and May 2006. Patients who died or were discharged within 24 hours after admission were excluded from the study. Rabat University Hospital is the referral venue for habitants in Western-North Morocco. The 12-bed medical ICU admits approximately 550 patients annually with an average age of 40 years. Surgery patients, coronary patients, neonates and burn patients are treated in specialized units. The study protocol was approved by the hospital ethics committee. Informed consent was not demanded because this observational study did not require any deviation from routine medical practice.

At the time of ICU admission, for each patient we evaluated their age, gender, principal diagnosis, and vital signs (body temperature, heart rate, respiratory rate, systolic and diastolic arterial pressure, and urine rate). The Mc Cabe index 20, the Acute Physiology and Chronic Health Evaluation II score 21 and the Sequential Organ Failure Assessment score 22 were also recorded on admission. The white blood cell count, the eosinophil cell count and the C-reactive protein (CRP) level were only systematically recorded on admission to the ICU and not daily during the entire ICU stay.

Blood samples were obtained by venipuncture on admission, and subsequently each morning at 07:00 hours. The clinical practice in the unit follows the recommendations of the task force of the American College of Critical Care Medicine of the Society of Critical Care Medicine 23: blood cultures were taken if a patient's body temperature exceeded 38.3°C, if a patient had clinical signs of severe sepsis, or if there was a need for vasopressor therapy for suspected septic shock. The samples for blood cultures were taken from two different sites, most commonly through intravascular devices (arterial cannula, central vein catheter or pulmonary arterial catheter.

Other cultures including urine, cerebrospinal fluid, and respiratory secretions were obtained according to the clinical circumstance and before antibiotics were given. Empirical antibiotic treatment was based on the presumptive diagnosis and received on the day of bacteriological cultures. When bacteriological results became available, the antibiotics were changed according to the pathogen isolated and the antimicrobial susceptibility test results.

According to the Criteria of the American College of Chest Physicians/Society of Critical Care Medicine 2, patients were classified as having SIRS, sepsis, severe sepsis, or septic shock at the time of admission. SIRS is defined by two or more of the following criteria: body temperature >38°C or <36°C, heart rate >90 beats/min, respiratory rate >20/min or PaCO₂ < 32 Torr, and white blood cell count >12,000 cells/mm³, <4,000 cells/mm³, or >10% immature forms. Sepsis is a SIRS associated with the presence of an infectious process. Severe sepsis is a sepsis associated with organ dysfunction, hypoperfusion, or hypotension (systolic blood pressure <90 mmHg or a reduction ≥ 40 mmHg from baseline). Septic shock is a subset of severe sepsis and is defined as a persisting sepsis-induced hypotension despite adequate fluid resuscitation.

Infection was diagnosed by textbook standard criteria 24 and was categorized according to the following: culture\microscopy of a pathogen from a clinical focus; positive urine dip test in the presence of dysuria symptoms; clinical lower respiratory tract symptoms and radiographic pulmonary abnormalities that are at least segmental and not due to pre-existing or other known causes; infection documented with another imaging technique; lumbar puncture when meningitis was suspected; obvious clinical infection (erysipelas); and identification of a pathogen by serology or by PCR.

Importantly, two investigators retrospectively reviewed all medical records pertaining to each patient and independently classified the diagnosis as SIRS, sepsis, severe sepsis, or septic shock at the time of admission on the basis of the review of the complete patient charts, results of microbiologic cultures, and radiographs. Both intensivists were blinded to the eosinophil cell count and CRP levels. Concordance among the two independent investigators was excellent and the reliability was high (κ = 0.94).

We assessed the value of eosinopenia as marker of sepsis by comparing the eosinophil cell count between noninfected patients (negative, SIRS) and infected patients (sepsis, severe sepsis, and septic shock), and between SIRS patients and infected patients on the day of admission to the ICU.

Blood samples were collected in microtubes containing ethylenediamine tetraacetic acid anticoagulant. The white blood cell count and the eosinophil cell count were performed by the Coulter (Gen·S) hematology analyzer (Beckman Coulter, Fullerton, CA, USA). To determine the CRP level, blood samples were drawn into green-top vacutainer tubes containing lithium-heparin as anticoagulant. Plasma CRP was also measured by immunoturbidimetry using the analyzer Cobas Integra (Roche Diagnostics, Mannheim, Germany). The limits of detection were 0.071 mg/dl.

Data are presented as the mean ± standard deviation for variables with a normal distribution, and as the median and interquartile range for variables with skewed distributions. Parametric or nonparametric tests were used for continuous variables as appropriate after the normality of the distribution was tested by the Kolmogorov-Smirnov test with Lilliefors correction. Statistical differences between groups were evaluated by the chi-square test for categorical variables. Comparison of group differences for continuous variables was carried out by one-way analysis of variance or the Kruskal-Wallis test. Bonferroni's post hoc test was used to locate the significance. The Spearman rank correlation coefficient (r) was calculated to describe the quantitative relationships between the eosinophil count and clinical or biological features.

The best cutoff value was chosen using Younden's index. Receiver operating characteristic curves and the respective areas under the curves were calculated for eosinophils and CRP. The sensitivity, specificity, and positive and negative likelihood ratios (with 95% confidence intervals (CIs)) were calculated at the best cutoff value. A multiple logistic regression was performed to explore the association between the eosinophil cell count, CRP levels, and infection, controlling for the potential confounders (age, Acute Physiology and Chronic Health Evaluation II score, Mc Cabe index, and Sequential Organ Failure Assessment score). Results are presented as the odds ratio and 95% CI.

A two-tailed P value <0.05 was considered significant. Statistical analyses were carried out using SPSS for Windows, version 13.0 (SPSS, Inc., Chicago, IL, USA).

During the study period, 198 patients were admitted to the ICU (Figure 1), and 21 patients were excluded because of death (n = 12) or discharge within 24 hours (n = 9). The remaining 177 patients were enrolled into the study, having a mean age of 42 ± 19 years. Mortality during the ICU stay occurred in 58 out of 177 patients (33%). At the time of admission, 120/177 patients (68%) had an infection. The sites of infections and clinical characteristics of the study patients are presented in Table 1.

Patients were classified as follows (Figure 1): negative group, 21% (n = 37); SIRS group, 11% (n = 20); sepsis group, 23% (n = 41); severe sepsis group, 31% (n = 55); and septic shock group, 14% (n = 24). Diagnoses in the negative group were acute poisoning (n = 30), scorpion envenomation (n = 3), acute ischemic stroke (n = 2), and hypercalcemia (n = 2). SIRS was caused by acute exacerbation of chronic obstructive pulmonary disease (n = 6), acute asthma (n = 4), diabetic ketoacidosis (n = 4), acute poisoning (n = 3), cardiogenic shock (n = 2), and gastrointestinal hemorrhage (n = 1).

Infections were microbiologically documented in 70 of 120 patients (58.3%); 60% had Gram-positive infection and 40% had Gram-negative infection. The major sources of infection were the respiratory tract (60%) and the urinary tract (21%).

The comparison of the eosinophil cell count and CRP levels among the different groups showed significant differences (Kruskal-Wallis test, P < 0.001) (Figure 2). There were no differences in the leucocyte count between the different groups (one-way analysis of variance, P = 0.095).

Concerning the comparison between the noninfected and infected groups, the median (interquartile range) eosinophil count was 109 (102 to 121) in noninfected patients and was 13 (8 to 28) in infected patients (P < 0.001). The median (interquartile range) CRP was 42 (18 to 79) and 108 (58 to 198) in the noninfected and infected groups, respectively (P < 0.001). Eosinophils had a higher discriminative value than the CRP level, with an area under the receiver operating characteristic curve of 0.89 (95% CI, 0.83 to 0.94) versus 0.77 (95% CI, 0.70 to 0.84) for CRP (P = 0.010) (Figure 3). At a cutoff value of 50 cells/mm³, eosinophils yielded a sensitivity of 80% (95% CI, 71% to 86%), a specificity of 91% (95% CI, 79% to 96%), a positive likelihood ratio of 9.12 (95% CI, 3.9 to 21), and a negative likelihood ratio of 0.21 (95% CI, 0.15 to 0.31) (Table 2). In multivariate logistic regression, the eosinophil cell count (adjusted odds ratio per 10-cell decrease, 1.09; 95% CI, 1.04 to 1.16; P = 0.002; frequency of significance in 1,000 bootstrap samples, 100%) and the CRP level (adjusted odds ratio per 1-point increase, 1.01; 95% CI, 1.00 to 1.01; P= 0.019; frequency of significance in 1,000 bootstrap samples, 98%) were found to be independent predictors of infection.

Concerning the comparisons between the SIRS and the infected groups (Figure 4), the median (interquartile range) eosinophil cell count was 121 (64 to 121) in SIRS patients and was 13 (8 to 28) in infected patients (P < 0.001). The median (interquartile range) CRP level was 59 (17 to 85) and 108 (58 to 198) in the SIRS and infected groups, respectively (P < 0.001). The area under the receiver operating characteristic curve was 0.84 (95% CI, 0.74 to 0.94) for eosinophils versus 0.77 (95% CI, 0.67 to 0.87) for CRP (Figure 5). The comparison of the areas under the receiver operating characteristic curves between eosinophils and CRP was not significant (P = 0.175). At a cutoff value of 40 cells/mm³, eosinophils yielded a sensitivity of 80% (95% CI, 71% to 86%), a specificity of 80% (95% CI, 55% to 93%), a positive likelihood ratio of 4 (95% CI, 1.65 to 9.65), and a negative likelihood ratio of 0.25 (95% CI, 0.17 to 0.36) (Table 2). In multivariate logistic regression, only the eosinophil cell count (adjusted odds ratio per 10-cell decrease, 1.07; 95% CI, 1.01 to 1.14; P = 0.019; frequency of significance in 1,000 bootstrap samples, 90%) was found to be an independent predictor of infection.

There were correlations between the eosinophil cell count and CRP level (r = -0.312, P < 0.001), between the eosinophil cell count and systolic blood pressure (r = 0.162, P = 0.030), between the eosinophil cell count and respiratory rate (r = -0.195, P = 0.011), between the eosinophil cell count and heart rate (r = -0.335, P < 0.001), and between the eosinophil cell count and Acute Physiology and Chronic Health Evaluation II score (r = -0.265, P < 0.001). No correlation was found between eosinophils and leucocytes or other clinical or biological features.