Study population
This study was approved by the Research Review Committee of National Taiwan University Hospital (NTUH). The enrolled severe sepsis patients were admitted to our institution between December 10, 2010 and December 10, 2011 and had a highly probable or proven infection and at least three of the following systemic inflammatory response syndrome (SIRS) criteria: body temperature >38°C or <36°C; heart rate >90/min; breathing frequency ≥20/min, PaCO2 < 32 mmHg, or ventilator use; leukocyte count >12000/mm3 or <4000/mm3 or >10% band forms; acute altered mental status; hyperglycemia without a history of diabetes; or blood glucose > 120 mg/dL. Other inclusion criteria were age ≥18 years, admission to the ICU, and at least one organ failure due to sepsis or septic shock. Septic shock was defined as sepsis with hypotension refractory to fluid challenge. Pregnant women, patients who had refused resuscitation, those with known or suspected human immunodeficiency virus infection, and those with known or suspected underlying immune deficiency were excluded. The zero time point was designated as being within 12 hours after the first organ failure due to sepsis. Blood was collected at the zero time point (day 0) and on days 1, 2, and 3.
Flow cytometric analysis
Blood was collected into a Vacutainer tube containing EDTA. One hundred microliters of blood was stained with CD3 PerCP/CD19 APC/CD4 FITC/CD 8PE (Becton Dickinson, San Jose, California, USA) at room temperature for 25 minutes in the dark, lysed with BD lysing solution, washed two times with phosphate buffer saline (PBS) containing 1% heat-inactivated fetal bovine serum, and fixed with PBS containing 0.25% paraformaldehyde. For each test, 20,000 leukocytes were collected and analyzed using a BD FACSCalibur flow cytometer with CellQuest software version 3.2 (Becton Dickinson, San Jose, California, USA).
Cytokine, chemokine, and carbonic anhydrase (CA) IX analysis
Blood was collected in a sodium heparin vacutainer tube. Plasma was collected, aliquot, and stored at −80°C until analysis. The plasma concentrations of MCP-1, IL-6, IL-7, IL-8, and IL-10 were separately analyzed by commercial ELISA kits according to the corresponding manufacturers’ instructions. The MCP-1 kit was obtained from eBioscience (San Diego, California, USA), the IL-7 kit was obtained from BioLegend (San Diego, California, USA), and the other cytokine kits were obtained from Becton Dickinson. Carbonic anhydrase (CA) IX is considered to be a marker of hypoxia [12] and was therefore included in the study. The CA IX kit was obtained from Oncogene Science (Cambridge, Massachusetts, USA).
Statistical Analysis
The patients were divided into the death or survival groups according to outcome. The patients in the survival group were further divided into a mid-term survival group (MTSG, characterized by survival through severe sepsis but death within six months or continued hospitalization for six months) and long-term survival group (LTSG, characterized by recovery from sepsis with survival for more than six months). The groups were evaluated at 4 different time points (Day 0, 1, 2, 3) for ten biomarkers (CA IX, MCP-1, IL-6, IL-7, IL-8, and IL-10, as well as the percentages of CD3+, CD4+, CD8+, and CD19+ lymphocytes). The biomarker levels were LOG10 transformed in all analyses.
The overall comparisons of the groups (the survival group versus the death group and LTSG versus MTSG) and time points for each biomarker were carried out using a mixed model. An appropriate covariance structure specification for each biomarker was chosen from four different covariance structures (unstructured (UN), Huynh-Feldt (HF), compound symmetry (CS), and first order autoregressive (AR1)) based on the smallest values of the Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC) [13]. The analysis of effects was carried out after the selection of a covariance structure. The p-values were calculated using Satterthwaite’s approximation in PRO MIXED. Values were determined to be significantly different when p < 0.05. All statistical analyses were performed in SAS version 9.2 (Cary, North Carolina, USA), and the above models were used.