This manuscript adheres to the applicable STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.
The intraoperative data used in this study were obtained from the “Registry Construction of Intraoperative Vital Signs and Clinical Information in Surgical Patients” study (H-1408-101-605, NCT02914444), which was designed to store intraoperative time-synchronized data from multiple anesthesia devices including patient monitors, anesthesia machines, BIS monitors, cardiac output monitors, and target-controlled infusion pumps by use of the ‘Vital Recorder’ (VitalDB team, Seoul, Korea) program. Using this registry, we could obtain complete intraoperative data (BIS-derived values, MAPs, and anesthetic concentrations).
Data collected for this study came from adult patients who underwent surgeries at Seoul National University Hospital between August 2016 and June 2017 under general anesthesia with BIS monitoring (BIS Vista, Covidien, Dublin, Ireland). The surgical procedures performed included abdominal surgeries on the gastrointestinal tract, liver, biliary tract, and pancreas. Data from the following cases were excluded: patients under 18 years old, cases with missing BIS value and MAP data more than 60 seconds, anesthesia times of less than 60 minutes, incomplete data on mortality, and reoperations during the period of analysis.
Vital sign data and clinical information pertaining to the cases were retrospectively analyzed. The data included patient’s diagnosis, age, sex, height, weight, type of operation, type and duration of anesthesia, propofol (Fresofol MCT inj 2%, Fresenius Kabi) concentration, MAC of volatile anesthetics, intraoperative BIS values, and arterial blood pressure. When MAP was less than 20 mmHg or greater than 200 mmHg, and when BIS was 0, these values were regarded as missing values.
To investigate the relationship between the duration of low BIS value maintenance and postoperative outcomes, we estimated the cumulative time in which BIS values were less than 20 or 40 and designated these as "bis20_dur" and "bis40_dur" respectively. To calculate total time of EEG suppression, designated as "eegsup_dur", we used a suppression ratio. The suppression ratio is the percentage of time over the last 63-second period in which the signal is considered to be in the suppressed state. As an example, a suppression ratio of 40 would mean "isoelectric over 40% of the last 63 seconds". After documenting suppression ratios at every second during anesthesia, we estimated the total time during which a patient’s EEG was suppressed by summing each case’s fractional suppression ratios applying a method used previously.(12) Lastly, we divided the sum by 60 to convert seconds to minutes and then by 100 to make percentages absolute numbers. To investigate the effects of short duration of brain suppression on clinical outcomes, we looked at the incidence in which cumulative time of BIS values less than 20 or 40 and EEG suppression lasted more than 5 minutes (bis20_5min, bis40_5min, and eegsup_5min respectively). To evaluate the influence of hypotension, we estimated the total time that MAP was lower than 50 mmHg (map50_dur) considering previous study.(15) We also calculated the cumulative time that MAP was less than 50 mmHg and BIS values were less than 20 or 40 simultaneously (bis20map50_dur, bis40map50_dur).
Potential clinical risk factors of postoperative mortality and delirium were determined in priori by clinical relevance or significance following a previous study.(12) We reviewed electronic medical records to retrieve the variables related to postoperative mortality and delirium. They included American Society of Anesthesiologists (ASA) physical status, past medical histories including the presence of aortic stenosis, congestive heart failure, coronary artery disease, hypertension, peripheral vascular occlusive disease, dysrhythmia, chronic obstructive pulmonary disease, stroke, malignancy, diabetes mellitus, sleep apnea, social history of smoking and drinking, and preoperative laboratory test results including hemoglobin (g/dL) and albumin (g/dL).
Mortality data were obtained from the Korean Ministry of the Interior and Safety using the resident registration number for each patient in February 2018. In this process, every piece of personal information collected was encrypted so as to maintain patient confidentiality. Mortality data were divided into 90-day postoperative mortality and 180-day mortality to compare early-to-intermediate term and intermediate-to-long term outcomes.(16)
Normality of continuous variables was verified with Kolmogorov–Smirnov test. In the univariable analysis, each variable of the data was analyzed by binary logistic regression in ‘enter’ method as an independent variable of postoperative mortality. Variables yielding P-values under 0.2 in the univariable analysis were selected as potential risk factors for multivariable analysis.
We used 2-step multivariable analysis to select more reliable variables considering multicolinearity because some BIS or MAP derived variables had close relationship. In the first step, among the selected risk factors which yielded P-value under 0.2 in univariable analysis, variable related with BIS or MAP was separately included in binary logistic regression with ‘enter’ method after excluding possibly BIS or MAP derived variables. In each binary logistic regression analysis, other potential risk factors not related to MAP or BIS were included. In this step, we removed BIS or MAP derived variables not yielding P-values under 0.05. In second step, the BIS or MAP derived variables yielding P-values under 0.05 in the first step and other potential risk factors not related to BIS or MAP in univariable analysis were included in final multivariable logistic regression analysis in ‘backward LR’ method. Variables remaining in the final logistic regression model were regarded as significant risk factors. The Hosmer-Lemeshow goodness-of-fit test was used to compare the estimate with the observed likelihood of outcomes.
To compare anesthetic concentration and double-low duration between patients with and without adverse outcome, we used Student t test or Mann-Whitney U test, as appropriate. All statistical analyses were performed using SPSS software version 23 (IBM Corp., Armonk, New York, USA).