This single-center prospective cohort study was conducted at the National Hospital Organization Nagasaki Medical Center, Nagasaki, Japan. This study was approved by our institutional research ethics committee (approval number: 2019059) on 2nd September 2019 and follows the Declaration of Helsinki. The study was registered with UMIN Clinical Trials Registry (Trial registry number: UMIN000037965, registration date: 8th September 2019) before the onset of participant enrollment. Written informed consent was obtained from each participant before study participation. This manuscript adheres to the STROBE guidelines (Supplemental Digital Content 1).
Participant Selection
The inclusion criteria included healthy (American Society of Anesthesiologist physical status Ⅱ) women, between 18 and 40 years of age, with term gestation (≥ 37 weeks), and scheduled cesarean delivery with combined spinal-epidural anesthesia. Exclusion criteria were as follows: unscheduled cesarean delivery, morbid obesity (body mass index [BMI] ≥ 40 kg/m2), preoperative hyperthermia (> 38°C) or preoperative hypothermia (< 36°C), cardiovascular or cerebrovascular disease, hypothyroidism or hyperthyroidism, history of anxiety disorder, difficulty in maintaining the supine position, and contraindication to spinal anesthesia. The study recruitment period was from 17th September 2019 to 9th March 2020.
Study Protocol
No parturients received any premedication. Each parturient was kept off solid food for at least 6 hours, and off clear water for 2 hours before spinal anesthesia. A 20-gauge peripheral intravenous (IV) cannula was inserted at the obstetric ward. Room temperature Ringer’s lactate solution was administered at a flow rate of 80 mL/h, about 2 hours before entering the operating room (OR). All parturients were directly transported from the ward to the OR without preoperative active warming (e.g., wearing socks, using body warming blanket), and the OR temperature was maintained at 27°C.
Each parturient was rapidly infused intravenously with 500 mL of 6% hydroxyethyl starches 130/0.4 (Voluven®; Fresenius Kabi, Tokyo, Japan) for prehydration before spinal anesthesia. Thereafter, Ringer’s lactate solution was infused about 10 mL/kg/h until the end of the surgery. Their infusion fluids in the OR were kept warm preoperatively at 38°C in the heat insulating cabinet. Standard monitoring was performed with an electrocardiogram, automated non-invasive arterial pressure measurement on the right arm, and finger pulse oximetry on the left index finger. Besides, the pulse oximeter probe (Masimo Rainbow SET Pulse CO-Oximeter Radical 7; Masimo Corp., Irvine, CA, USA) was placed on the left second toe for continuous monitoring of toe PI. For core temperature measurement, we attached the 3M™ Bair Hugger™ Temperature Monitoring System (3M Company, St. Paul, MN, USA) over the right temporal region. This Food and Drug Administration-approved Temperature Monitoring System can measure core temperature by heating the skin sensor attached to the forehead and reaching thermal equilibrium between sensor temperature and core temperature [19]. The mean error in measurement accuracy of this device was found to be -0.23°C (95% limits of agreement of ± 0.82°C) compared with pulmonary artery temperature [19].
All parturients received combined spinal-epidural anesthesia in the right lateral decubitus position. After inserting an epidural catheter at the T12-L1 or L1-2 vertebral interspace, spinal anesthesia was performed at the L2-3 or L3-4 vertebral interspace. A 25-gauge Quincke spinal needle was inserted into the subarachnoid space, and 10 mg (2.0 mL) of 0.5% hyperbaric bupivacaine (Marcain®; Aspen Japan, Tokyo, Japan) with 15 µg (0.3 mL) fentanyl (Fentanyl®; Janssen Pharmaceutical K.K., Tokyo, Japan) were administered intrathecally. Following the securing of the epidural catheter, each parturient was returned to the supine position with a 15° left lateral tilt to facilitate left displacement of the uterus. The tilted position was returned to the horizontal supine position after the maternal hemodynamics stabilized. The sensory blockade level was checked after spinal injection using cold ice. If T4 sensory block level was not achieved, 2% lidocaine (Xylocaine® Injection Polyamp 2%; Aspen Japan, Tokyo, Japan) was administered through the epidural catheter in 5 mL increments until it was achieved. To prevent post-spinal hypotension in the parturient, phenylephrine at 0.3 µg/kg/min was started immediately after the induction of spinal anesthesia. Once the systolic blood pressure (SBP) was less than 80 mmHg or there were symptoms consistent with hypotension (e.g., dyspnea, nausea, or vomiting) even without SBP < 80 mmHg, a bolus of 50 to 100 µg phenylephrine or 4 mg ephedrine was administered depending on the patient’s heart rate (HR). When the patient’s HR was less than 60 beats/min without the occurrence of post-spinal hypotension, a bolus of 0.5 mg atropine was given. When the patient’s SBP was stable, the continuous administration of phenylephrine was gradually reduced, and was terminated at the discretion of the anesthesiologist.
Since the start of the surgery, the patient’s upper body was warmed using a 3M™ Bair Hugger™ multi-position upper body warming blanket (Model 622; 3M Company, St. Paul, MN, USA) attached to a 3M™ Bair Hugger™ warming unit (Model 675; 3M Company, St. Paul, MN, USA) set to 38°C. The OR temperature was changed from 27°C to 24°C after placing the newborn baby in the infant incubator.
Measurements
Parturients’ characteristics and baseline parameters were obtained from electronic medical and anesthetic records. Toe PI and core temperature were recorded at one-minute intervals from entering the OR until the end of the surgery. Preoperative toe PI was defined as the average PI value measured for three minutes in the horizontal supine position, immediately before right lateral decubitus repositioning. All parturients were instructed to remain motionless and rested during the preoperative toe PI measurement. To evaluate redistributive temperature decrease after spinal anesthesia, we investigated the maximum core temperature decrease in the perioperative period. The primary outcome was the maximum core temperature decrease. In this study, the perioperative period was defined as the time from entering the OR until the end of the surgery, and the maximum core temperature decrease was defined as the difference between the core temperature at OR admission and the minimum intraoperative core temperature. Moreover, we evaluated shivering severity and thermal comfort when leaving the OR. Shivering severity was assessed using the Bedside Shivering Assessment Scale: 0 = no shivering, 1 = shivering localized to the core and neck, 2 = shivering including the upper extremities, 3 = total body shivering [20]. Thermal comfort was measured using the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) scale, which is a seven-point Likert scale: Hot (+ 3), Warm (+ 2), Slightly warm (+ 1), Neutral “just right” (0), Slightly cool (-1), Cool (-2), and Cold (-3) [21].
Surgery time was defined as the time between the start of the surgery and the end of the wound closure. The total volume of intraoperative IV fluids, the total dose of cardiovascular drug administration, and estimated blood loss were also recorded.
Statistical Analysis
General characteristics of the parturients and the surgical data are presented as median [interquartile range], or the number of patients (%). Parturients’ body temperature, cold-related outcomes, and parturients’ toe PI data are presented as mean ± standard deviation (SD), median [interquartile range], or the number of patients (%). The Shapiro–Wilk test was used to determine normality. The sample size was determined solely according to the number of parturients hospitalized during the study period since these parturients did not receive any intervention.
To analyze whether there is a relationship between preoperative toe PI and decreased maximal core temperature, firstly, correlation analysis was performed. Secondly, a segmented regression model (SRM) and a generalized additive model (GAM) were used. We suspected that the relationship between preoperative toe PI and decreased maximal core temperature was nonlinear from the scatter plot; thus, we conducted analyses using SRM and GAM.
SRM (also called change-point regression) is a practical analysis if we expect to have several slopes between dependent and independent variables different from a simple linear regression. These slopes quantify the change in the relationship between the two variables. Points where the slope changes are called “change-points.” The change-point can be interpreted as a critical, safe, or threshold value beyond or below which desired effects occur and is important in decision making [22]. In this study, we considered that two slopes exist between the dependent and independent variables from the scatter plot. We pre-determined four expected preoperative toe PI change-points based on quartile ranges (first, second, and third quartiles) and mean, created four SRMs, and evaluated the Akaike's Information Criterion (AIC) of each model to determine the best fit model. GAM provides a modeling approach that combines powerful statistical methods with interpretability, smooth functions, and flexibility. Although generalized linear models, such as simple regression analysis, can only express linear relationships, GAM can also express non-linear relationships while maintaining interpretability and flexibility using multiple smoothing functions (smoothers). To find the smoother that best fits the data, the choice of smoothing parameters—i.e., the parameters that control the smoothness of the predictive functions—is key as in SRM [23]. We created several models to find the optimal parameters in GAM. The results of each model were evaluated using root mean squared error (RMSE) to determine the model that best fit the data. SRM and GAM included covariates associated with maternal hypothermia, which are BMI and core temperature at OR admission [24].
Statistical significance was defined as P value < 0.05. All statistical analyses were performed with R version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).