This was a prospective observational cohort study. The Institutional Review Board of Guizhou Provincial People’s Hospital approved the study (#2014099). Verbal and written informed consents were obtained from all participants before surgery.
Patients
Patients who met the following inclusion criteria from December 20, 2017, to December 30, 2018, were recruited for this study: 1) age ≥ 18 years, 2) elective cardiac surgery, and 3) CPB requirement. We excluded patients younger than 18 years or scheduled for surgeries that were emergent, off-pump, or that required deep hypothermic circulatory arrest.
Vascular cannulation
Before anesthesia induction, the following vascular cannulations were performed under local anesthesia. A 20G catheter was inserted into a radial artery at the wrist for continuous arterial blood pressure monitoring and blood sampling. A 20-cm-long 7F triple-lumen central venous catheter was inserted into the superior vena cava through the internal jugular vein for central venous pressure monitoring and blood sampling. A 20-cm-long 16G single-lumen catheter was inserted into the right internal jugular vein retrogradely until encountering resistance, and it was then slowly retracted until the blood was aspirated for jugular venous blood sampling.[20] The position of the catheter for internal jugular venous blood sampling was confirmed by fluoroscopy.
Anesthetic care
The anesthesia team administered anxiolytics to patients at their discretion before surgery. In the operating room, patients were monitored using electrocardiography, pulse oximetry, and invasive arterial blood pressure. Following pre-oxygenation, anesthesia was induced with midazolam (0.1 mg/kg), etomidate (0.2–0.5 mg/kg) or propofol (1–2 mg/kg), and sufentanil (0.2–0.5 mcg/kg). Endotracheal intubation was facilitated by muscle relaxation using vecuronium (0.1 mg/kg). Anesthesia was maintained using sevoflurane, with intermittent boluses of sufentanil for analgesia and vecuronium for muscle relaxation. Patients were mechanically ventilated with a tidal volume of 6–8 ml/kg, a respiratory rate of 8–12 times per minute, and an inspired oxygen fraction of ~70%. A nasopharyngeal probe was used for temperature monitoring.
Surgery and CPB
Surgery was performed through midline sternotomy. Heparin was administered to maintain an activated clotting time > 480 seconds. All patients underwent non-pulsatile CPB with a membrane oxygenator. The temperature during CPB was maintained at 30°C. Patients were rewarmed to a target nasopharyngeal temperature of 37°C before the termination of CPB. The target mean arterial pressure during CPB was 60-80 mmHg. Inotropic and vasopressor options included dobutamine, epinephrine, and norepinephrine, which were given as intravenous infusions alone or in combination as needed.
Blood gas analysis
We analyzed alpha-stat arterial, jugular venous, and central venous blood gases (GEM premier 3000, Instrumentation Laboratory, Bedford, MA, USA) using blood samples from the radial artery, internal jugular vein, and central vein, respectively. Blood samples were harvested at the following time points: 1) before anesthesia induction (pre), 2) after anesthesia induction (post), 3) during CPB with body temperature at ~30°C (30°C), 4) during CPB with the patient rewarmed to ~37°C (37°C), and 5) at the end of surgery (end).
Outcome measures
The primary outcome measures included lengths of mechanical ventilation (LMV), intensive care unit stay (LICU), and hospital stay (LOH). The secondary outcome measure was the incidence of major organ morbidity and mortality (MOMM) defined by the Society of Thoracic Surgeons.[21] MOMM is a composite measure of any of the following occurrences: 1) operative mortality defined as death from any cause; 2) stroke defined as a new-onset central nervous system deficit persisting longer than 72 hours; 3) renal failure defined as a new requirement for dialysis or an increase in serum creatinine to more than 2.0 mg/dl and doubling the most recent preoperative measurement; 4) prolonged ventilation defined as a need for mechanical ventilation longer than 24 hours; 5) deep sternal wound infection; and 6) reoperation for any reason.
Data collection
We collected patients’ demographic data including age, gender, weight, New York Heart Association classification, and American Society of Anesthesiologists physical status scores. We also obtained data on comorbidities including diabetes, dyslipidemia, hypertension, stroke, and myocardial infarction. And, we collected data including preoperative ejection fraction, medications (angiotensin converting enzyme inhibitors, beta-blockers, nitroglycerin, digoxin, and diuretics), procedural details (CPB time, aortic cross-clamp time, surgical time, and type of surgery), and anesthetic details (agents and drugs used, blood products and fluid administered). The results of blood gas analysis of different types (arterial, jugular venous, and central venous blood) at different pre-determined time points were recorded. We prospectively collected outcomes including LMV, LICU, LOH, and MOMM. All patients were followed up with, and medical records were reviewed on a daily basis until discharge.
Statistical analysis
In a multiple linear regression analysis of a dependent outcome, a sample size of 190 achieves 80% power to detect an R2 of 0.03 attributed to 1 independent variable (i.e., a blood gas parameter) when using an F-test with a significance level (alpha) of 0.05 and adjusting for an additional 5 independent variables (covariates) with an R2 of 0.20 (0.4). The value of R2 = 0.03 is equivalent to a global Cohen’s f2 effect size of 0.03 (= R2/(1- R2)), and f2 values of 0.02, 0.15, and 0.35 represent small, medium, and large effect sizes, respectively.
We checked continuous variables for normality of their distribution. If normally distributed, we expressed them as the mean ± SD. Otherwise, we reported medians [interquartile ranges]. Categorical variables were expressed as actual numbers and percentages.
We log-transformed three outcomes (LMV, LICU, and LOH) before the linear regression analyses. We first conducted a simple (i.e., univariate) regression analysis to look for the association of a single factor (e.g., patient demographics, blood gas parameters, and other clinical covariates) with each dependent outcome (i.e., LMV, LICU, and LOH), separately. To avoid fitting too many models, we entered only those factors found significantly associated (i.e., p < 0.05) with all three outcomes in the multiple linear regression analysis to identify independent risk factors. Because we aimed to assess the association of a blood gas parameter with the outcome after adjusting for important covariates, if one of the blood gas parameters was found to be statistically significant in the univariate analysis, we kept it in the multivariable model, in which we implemented a backward selection procedure to select for other risk factors, with an entry and exit significance level of 0.10. We determined the collinearity between factors by variance inflation factors (< 10). We calculated the regression coefficient (beta) and its standard error to quantify the effect sizes for the final factors that remained in the multiple regression model.
Next, we dichotomized LMV, LICU, or LOH into binary outcomes (unfavorable, for values > median; favorable, for values ≤ median). We compared patient demographics, blood gas parameters, and other clinical covariates between two groups stratified by three new binary outcomes and MOMM by univariate logistic regression analyses. As a measure of effect size, we reported the odds ratio (OR) with a 95% confidence interval (CI) for each factor.
We performed all statistical analyses using the IBM SPSS 23.0 package (Chicago, IL, USA) and the SAS software version 9.4 (Cary, NC, USA). We considered two-sided p-values lower than 0.05 as statistically significant.