Study population and data collection
After obtaining approval from the Institutional Review Board at Seoul National University Bundang Hospital (B-1510/318-04), electronic medical records from January 2012 to July 2017 were reviewed retrospectively. The requirement for informed consent was waived.
Adult patients aged 20 years or over, who had undergone heart valve or CABG surgery with CPB were included in this retrospective study. Patients who had been preoperatively diagnosed with neuropsychological diseases, such as dementia or Parkinson’s disease, were excluded. Patients with visual disturbances, hearing loss, or reoperation due to postoperative complications were excluded. Also, patients requiring postoperative sedation for any reason were excluded because PD might be unnoticed due to sedation.
Routine general anesthesia practice
On arrival in the operating room, standard monitoring (pulse oximetry, electrocardiogram, and noninvasive arterial pressure), bispectral index (BISTM, Covidien Icn., USA), cerebral oximeter, and invasive arterial monitoring were established. Central vein catheterization was performed after induction of general anesthesia.
During the investigation period, different general anesthetic agents had been administered to patients in our institution. Based on these different main anesthetic agents, patients were divided into two groups: sevo-dex and propofol groups. In the sevo-dex group, anesthesia was induced with intravenous propofol, remifentanil (target effect-site concentration 4 ng/ml), and sevoflurane. During the intraoperative period, anesthesia was maintained with sevoflurane and dexmedetomidine. Dexmedetomidine (PrecedexTM, Hospira Inc., Lake Forest, IL, USA) was diluted with 0.9% saline to make a concentration of 4 μg/ml and administered continuously at 0.5 μg/kg/h. In the propofol group, total intravenous anesthesia was performed as follows: 2% propofol (Fresofol®, Fresenius Kabi, Korea Ltd, Korea) and remifentanil (Ultiva®, GlaxoSmithKline, United Kingdom) diluted to 50 μg/mL were administered using a target-controlled infusion device (Orchestra®, Fresenius vial, France) until the effect-site concentration reached 4 μg/ml and 4 ng/ml, respectively. During surgery, propofol and remifentanil was maintained within 2.0–5.0 μg/ml and 0.5–3.0 ng/ml. Dexmedetomidine was not used in the propofol group. Bispectral index was monitored to maintain suitable anesthetic depth (40– 60) in all patients. Rocuronium was administered as a neuromuscular blocking agent in both groups.
Assessment of postoperative delirium
PD was defined as the positive Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) during the ICU stay. According to the standard ICU practice of our institution, CAM-ICU was examined and recorded twice a day by directed bedside nurses, who had been well trained. The type of PD was determined according to the Richmond Agitation and Sedation Scale (RASS) score. When the RASS level was positive or negative, PD was classified as hyperactive or hypoactive, respectively. When a patient presented with mixed positive and negative RASS scores, they were diagnosed with mixed-type PD.
After patients were transferred to the general ward, PD was assessed until postoperative day 5 by nurses based on the Nursing Delirium Screening Scale per nursing shift. While PD was identified with psychomotor retardation, it was categorized as hypoactive PD. The duration of delirium was defined as the total number of days presenting PD in ICU and general ward.
Other outcome variables
Data were collected and categorized into the following three variable sets: (1) preoperative factors, including age, sex, weight, height, body mass index, American Society of Anesthesiologist (ASA) physical status classification, and preoperative laboratory findings; (2) intraoperative factors, including the urgency of surgery, surgery time, CPB time, volume of estimated blood loss, and amount of red blood cell (RBC) transfusion; and (3) postoperative factors, including the duration of ICU stay, extubation time, postoperative admission period, postoperative complications, amount of RBC transfusion, and postoperative laboratory findings. Pre- and postoperative laboratory tests including hematocrit, platelet counts, electrolytes, creatinine, and albumin, were performed within 1 month before surgery and within 24 h after surgery, respectively. We investigated whether or not patients developed postoperative complications related to the renal and neurological systems.
Data are expressed as median (interquartile range) or number (proportion). All continuous data were assessed for normality using the Shapiro-Wilk test. Incidence was analyzed using the chi-square test or Fisher’s exact test. The Mann-Whitney U test or Wilcoxon signed rank test was performed to compare the numerical data, as appropriate. A binary logistic regression model was used to evaluate the predisposing factors of PD. The dependent variable was the occurrence of PD and the independent variables were the main anesthetic drug, age, sex, BMI, surgery type, ASA class, anesthesia time, CPB time, estimated blood loss, intraoperative and postoperative RBC transfusion amount, and preoperative and postoperative hematocrit, electrolytes, creatinine levels, and albumin level.
Propensity-score matching was performed to reduce the risk of confounder effects between the sevo-dex and propofol groups (Austin 2011). Propensity scores were calculated using a logistic regression model. Covariate included the sex, age, height, weight, BMI, ASA class, urgency of surgery, operation time, preoperative laboratory findings, estimated blood loss, duration of CPB, intra- and postoperative RBC transfused, extubation time, and type of surgery. The dependent variable was the main general anesthetic agent: inhalation with dexmedetomidine or propofol. We performed nearest-neighbor matching (Austin 2011). After propensity-score matching, the paired t-test or Wilcoxon signed-rank test for continuous variables and McNemar or McNemar-Bowker test for categorical variables were performed, as appropriate.
All analyses were carried out using IBM® SPSS® Statistics, version 22.0 (IBM Corporation, NY, USA). P < 0.05 was considered to indicate statistical significance.