The study was approved by Institutional Review Board (H- 1305-0450488), Ministry of Drug and Food Safety (20130131631) and registered at http://cris.nih.go.kr. After obtaining written informed consent from the parents or legal guardians, pediatric patients aged 12–48 months who were scheduled to undergo elective urological surgeries, such as penoplasty, orchiopexy, circumcision, and hypospadias repair under general anesthesia were enrolled. Exclusion criteria were the history or presence of disorders of the central or peripheral nervous system, drug allergic reaction to local anesthetics, active sepsis or local skin infection at the injection site, coagulopathy, sacral anomalies, or other complications inappropriate for the study.

On the day of surgery, the patients arrived in the operating theater after appropriate fasting and without premedication. They were monitored (Solar 8000, GE Medical, USA) using lead II of a three-lead ECG, noninvasive blood pressure at 1-min and 2.5-min intervals during and after induction, respectively, peripheral pulse oximetry (SpO₂) on the left big toe, and end-tidal carbon dioxide (ETCO₂). Anesthesia was inducted with 5–6 mg/kg sodium thiopental, and 8 vol% sevoflurane in oxygen was inhaled after loss of consciousness. A neuromuscular blocking agent (0.6 mg/kg rocuronium) was administered to facilitate tracheal intubation. The patients were anesthetized with conventional anesthesia with 2.5–3 vol% sevoflurane in 35% oxygen after intubation, and sevoflurane concentration was decreased to 1 MAC at the end of surgery and fixed at the same concentration while the caudal block procedure was performed. The patient’s body temperature was actively controlled with a warming blanket and air-forced warmer. Five minutes after the caudal block, the patients were awoken from anesthesia by turning off the sevoflurane and recovered from the neuromuscular blockade with 0.25 mg/kg atropine and 0.02 mg/kg neostigmine. The patients were extubated when they were fully awake. The patients were transferred to the post-anesthetic care unit (PACU) and stayed for at least 30 min for stabilization. We also checked the postoperative pain scale (FLACC scale) [18] and emergence delirium scale (PAED scale) [19] in the PACU.

The patients were placed in the right lateral bent knees position after the end of surgery, and the attending anesthesiologist palpated and drew the anatomical landmarks on the body. Heart rate, ECG, blood pressure at 1 min intervals, and SpO₂ were monitored continuously. Subsequently, the sacral hiatus was identified by ultrasonography (Logiq e, GE Healthcare, USA) with a 12-MHz linear probe before puncturing the skin. Next, a 25-gauge B bevel needle was inserted until a marked decrease in resistance was noted, followed by a 1 ml/kg 0.25% ropivacaine injection containing 1 : 200,000 epinephrine. A test dose was injected first, and expansion of the epidural space was noted. During drug injection, the area over the lower lumbar spine corresponding to the needle tip was auscultated for an audible “swoosh” to confirm successful needle positioning. The injectate was also visualized within the caudal canal by using ultrasonography; the block was considered successful with identification of the “swoosh” sound. We maintained the patient’s position at the right lateral bent knees for 5 min after the procedure, which is identical to the position at the beginning of the procedure.

Heart rate, ECG, blood pressure at 1 min interval, and photoplethysmograph (PPG) data from the patient monitor were transferred to a personal computer using an analog-to-digital converter (DA 149, DATAQ Instruments, USA) at a 1,000-Hz frequency.

The PTT was measured as the time interval between the R-wave peak on the ECG and the maximum upslope of the corresponding PPG waveform, which was derived automatically from its first derivative using the CALC package of the Advanced CODAS analysis software (ver. 3.25; WinDaq, DATAQ Instruments, USA) [20]. We filtered the data using the following algorithm to reduce PPG signal artifacts: each pulse wave was compared to the mean pulse wave of 10 subsequent pulse waves ± 1 SD. If 25% or more of the current pulse wave was outside this window, the pulse wave was filtered out. The PTT was calculated for the remaining pulse waves.

ECG data were reviewed by visual inspection. Segments with signal loss or significant noise, or atrial or ventricular extrasystole(s), were discarded. R-R intervals were measured and tested for the presence of outliers (heart rate > 200 beats/min), and any outliers found were removed. The HRV was calculated automatically with LabChart (ver. 7; AD Instruments, USA) at 1-min intervals. Time domain indices that utilize NN interval (normal-to-normal interval, equivalent to R-R interval), such as standard deviation of normal-to-normal intervals (SDNN) (measures total variability that arises from both periodic and random sources) and root mean square of successive differences (RMSSD) (independent of long-term trends and predominantly reflects vagal tone), were obtained. The frequency domain method is a kind of HRV analysis that assigns bands of frequency (ultralow, very low, low, and high) and then count the number of NN intervals that match each band. This method yields information regarding the amount of the overall variance in heart rate resulting from periodic oscillations of heart rate at various frequencies. High frequency (HF: 0.15–0.4 Hz) is mediated parasympathetically and primarily represents respiratory variation. Low frequency (LF: 0.04–0.15 Hz) is modulated by both the sympathetic and parasympathetic nervous systems and is strongly affected by the oscillatory rhythm of the baroreceptor system. Very low frequency (VLF: 0.0033–0.04 Hz) and ultralow frequency power (ULF: 1.15 × 10⁻⁵–0.0033 Hz) may represent the influence of thermoregulatory, peripheral vasomotor, or renin–angiotensin systems [21]. The LF/HF ratio has been used as an index of the sympathetic/parasympathetic balance. Cardiac autonomic function was assessed by frequency-domain HRV analysis in the range of 0–1 Hz on normal R-R intervals. A fast Fourier transformation was used to perform power spectral density analysis based on the Welch method.

As a non-linear method, approximate entropy (ApEn) of the R-R interval was calculated and taken into analysis. ApEn is a derivative that describes the regularity of pattern in data sets. It is defined as the logarithm of likelihood wherein similar short segments of data, two consecutive data for example, remain similar when the next datum (the third in this case) is added. This can be regarded as a measure of predictability. A higher ApEn was associated with a lower predictability [22]. The ApEn of the R-R interval in the ECG was calculated using MATLAB (ver. 2016b, MathWorks Inc., USA) at 1-min intervals from 1 min before to 5 min after the block. The MATLAB code for calculating ApEn originated from the following web site: https://kr.mathworks.com/matlabcentral/fileexchange/26546-approximate-entropy. The mean heart rate for each 1-min period was also calculated and taken into analysis.

Changes in PTT, blood pressure, heart rate, and HRV parameters from the baseline in the supine position to 5 min after the caudal block were the primary outcome of this study. The secondary outcomes were ApEn of the R-R interval, postoperative pain, and emergence agitation scales. All of the mentioned data were obtained for 1 min before the block as baseline, and for 5 min after the block at 1-min intervals. The FLACC score was measured upon arrival, 15 min, and 30 min after arrival at the PACU. The PAED score was measured upon arrival and at discharge from the PACU.

Values are shown as the mean ± SD (normally distributed data). Analyses of PTT, HRV parameters, ApEn of the R-R interval, and FLACC scores were performed using one-way repeated measures analysis of variance (RMANOVA). Either RMANOVA or Friedmann RMANOVA on ranks was selected after the normality test. Sphericity assumption was evaluated via the Mauchly sphericity test. The Greenhouse–Geisser correction was used when sphericity was not violated. The Wilk lambda, a multivariate test statistics, was used to obtain the F ratio when sphericity was violated. The PAED scores upon arrival and at discharge from the PACU for each individual were compared using paired t-test. Statistical analyses were performed with the SPSS software (ver. 22.0; IBM Corp., USA). P values < 0.05 were considered statistically significant.

We planned a one-way RMANOVA for a single group as a method for PTT and HRV analyses. Each measurement was analyzed six times (baseline, each 1-min period after caudal block for 5 min). Because no previous similar study was published, the effect size f was set at medium (0.25). At the level of α = 0.05, power = 0.8, the required total sample size was 19. The expected dropout rate, including block failure, was 30%, and the number of candidates was set to 27. The calculations were performed via G*Power (ver. 3.1.9.2; Universistät Kiel, Germany).