This research was conducted as a single-center, prospective observational study at a tertiary care teaching hospital from January to March 2024. The aim was to assess the feasibility of using regional blood flow spectral parameters as predictors for ERMF. Ethical approval for the study was granted by the institutional ethics committee (approval number: KY-2024-009-K01), and the study was registered with the China Clinical Trials Registry https://www.chictr.org.cn/ (reference number: ChiCTR2400080507, 31/01/2024). The methodology and procedures were thoroughly explained to all participants, ensuring complete understanding. Following this, written informed consent was obtained from each participant, adhering to ethical standards and research governance.

Participants in this study were parturients who had epidural analgesia during labor at the Women and Children’s Hospital, School of Medicine, Xiamen university. PASS 14.0 software (NCSS LLC, Kaysville, UT, USA) was employed to determine an appropriate sample size, considering the established principles for studies of this nature. The end-diastolic velocity (EDV) was the primary measure of interest. Drawing on previous literature findings and a pilot study [9], the anticipated area under the receiver operating characteristic (AUROC) for predicting ERMF was established at 0.8. Given the local ERMF incidence rate of approximately 20% and an estimated attrition rate of 10%, it was calculated that a minimum of 60 participants would be required [10]. This sample size was necessary to achieve a statistical power of 90%, with a Type I error probability set at 0.05 [11].

Inclusion criteria:

(1) Individuals aged between 25 to 35 years, having a singleton, primiparous women, full-term birth [2].

(2) Had a vaginal delivery.

(3) Consented to participate in the study, verified by the signing of an informed consent document.

Exclusion criteria:

(1) A preexisting severe cardiorespiratory condition.

(2) Any pregnancy complications that could potentially influence autonomic functions, such as gestational diabetes or gestational hypertension.

(3) A documented history of peripheral vascular disease.

(4) The presence of an active infectious disease, fever, or a basal body temperature of 37.5 °C or higher prior to delivery.

(5) Any contraindications to receiving epidural analgesia during labor.

(6) Severe cognitive or psychiatric disorders that could impair study participation or consent.

(7) The use of medications during the ultrasound parameter measurement phase that could affect peripheral vasoconstriction, including nifedipine.

(8) Any other medical or health conditions deemed by the anesthesiologist to render the individual unsuitable for inclusion in this study.

Upon the onset of regular uterine contractions, participants were admitted to the labor room. The anesthesiologist discussed epidural labor analgesia (ELA)’s advantages and potential risks with the participants before the study commenced. The delivery room temperature was regulated to remain between 24–26 °C, with careful attention paid to maintaining the participant’s bilateral lower extremities warm. Peripheral intravenous access was secured, and a Lactated Ringer’s solution was infused at 2–5 mL/kg/h. An experienced attending anesthesiologist performed the epidural insertions. Continuous real-time monitoring of vital signs was done, including noninvasive blood pressure (NIBP), electrocardiogram (ECG), oxygen saturation (SPO₂), and fetal heart rate. For the procedure, participants were positioned in the left lateral decubitus position. The epidural catheter insertion site was selected at the L2/L3 or L3/L4 intervertebral space. The “water resistance disappearing” technique was employed to confirm entry into the epidural space. Upon a successful insertion of the spinal needle, a catheter was inserted into the lumber epidural space to a depth of approximately 3–4 cm. A test dose of 3 mL of 1.5% lidocaine was administered. The participants were then monitored for 3–5 for any signs of local anesthetic toxicity or total spinal anesthesia.

The patient-controlled epidural analgesia pump (DDB-I-A1, Apollo, Nantong, Jiangsu, China) was prepared with a mixture containing 0.08–0.15% ropivacaine and 0.4 µg/mL sufentanil, totaling 100 mL [12]. The analgesic pump was set to deliver an initial bolus of 12–15 mL, followed by a maintenance infusion of 8–10 mL/h, with a patient-controlled bolus of 6 mL and a lockout interval of 20 minutes. A blunt needle was used to ensure the anesthesia coverage extended from the T10 to S4 dermatomes. If a subject’s visual analog scale exceeded 4 after 30 minutes from the initial epidural dose, the analgesia was considered inadequate [13, 14]. In this case, an additional 8 mL of 0.125% ropivacaine was injected, and the participant was excluded from the study. Participants who required vasoactive drugs to manage hemodynamic instability during measurements were also excluded from the study.

Baseline information on subjects was documented, including weight, height, gestational age, duration of labor, hemorrhage, side effects, oxytocin dosage, basal body temperature, and any other medications taken during the study. Blood flow spectrum images were acquired while subjects were in the left or right lateral position to avoid the influence of the expanded uterus compressing blood vessels on ultrasound data. Ultrasound measurements of the right lower extremity were uniformly performed on all participants using a PWD US system (MX7, Mindray, Shenzhen, Guangdong, China) with a 3–12 MHz high-frequency linear array transducer probe to reduce variability.

Subjects underwent ultrasound measurements in the left lateral recumbent position. The anterior tibial artery (ATA) was scanned just above the ankle joint on the anteromedial aspect of the tibia, while the posterior tibial artery (PTA) was scanned at the medial malleolus. Mark the probe position at the measurement point with a black marker pen after the first measurement to ensure that all subsequent measurements were made at the same position. Colour Doppler images of the longitudinal axis of the artery were obtained by keeping the probe vertical and tilting it slightly to the left or right, ensuring that the adventitia and posterior intima were visible at the maximum diameter of the vessel. Adjust the probe position so that the blood flow is as close as possible to the direction of the sound beam [15]. No pressure should be exerted on the probe that could cause deformation of the vessel. After optimising the B-mode US image, the PWD US mode was activated. The angle between the PWD scan line and the vessel should be $\le$60°. The placement of the sampling volume was done under the guidance of color Doppler and placed in the centre of the vessel where the colour was brightest. Adjust the size of the sampling volume, usually to 1/3 to 1/2 of the inner diameter of the vessel. The scale and baseline were optimized for the PWD scan. Peak systolic velocity (PSV) and EDV were recorded using automated tracing (Supplementary Fig. 1).

Data was captured one minute prior to induction of epidural analgesia (PSV₀) and then at five-minute intervals for thirty minutes post-induction (PSV₅–PSV₃₀). The difference relative to baseline in blood flow parameters in the ATA and PTA was calculated ($\mathrm{\Delta }$PSV = PSV₃₀ – PSV₀; $\mathrm{\Delta }$EDV = EDV₃₀ – EDV₀). These measurements were performed by a certified physician trained in ultrasonography at the Second Affiliated Hospital, Zhejiang University School of Medicine. Throughout labor, the participants’ temperatures were measured every hour using an infrared ear thermometer, continuing until two hours after the conclusion of the second stage of labor. Following UK guidelines, ERMF was defined as a maternal temperature reading of 38 °C on one occasion or two consecutive readings of 37.5 °C spaced one hour apart [16].

Statistical analyses were performed utilizing SPSS software, version 22.0 (IBM Corp., Armonk, NY, USA). Graphical representations were generated with GraphPad Prism version 8.0 for macOS (GraphPad Software, San Diego, CA, USA) and R version 3.6.1 (R Core Team, Vienna, Austria). The normality of data distributions was evaluated using the Kolmogorov-Smirnov test. Data adhering to a normal distribution were expressed as the mean $\pm$ standard deviation (SD). For comparisons between two independent groups, the t-test was applied. Generalized estimating equation was used to assess differences both between and within groups across multiple time points. For continuous data that deviated from normal distribution, the median and interquartile range (IQR) were used as descriptive statistics, and the Mann-Whitney U test was used for group comparisons. Categorical data were expressed as counts or percentages, and the $\chi$² test or Fisher’s exact test was used when deemed suitable for between-group comparison. A p-value 0.05 indicated statistical significance. The correlation between two variables was analyzed using Spearman’s rank correlation coefficient. Receiver operating characteristic (ROC) curves were plotted to ascertain the sensitivity, specificity, and threshold values of lower extremity arterial ultrasound flow spectral parameters for the prediction of ERMF. The Youden index on the ROC curves identified the optimal threshold value. The Youden index is calculated as sensitivity + specificity – 1. The maximum value of the Youden index indicates the greatest ability to predict the ERMF population; the corresponding diagnostic threshold is considered the optimal cut-off value. The area under the curve (AUC) and 95% confidence interval (CI) were calculated and reported to provide a complete statistical assessment.

Out of 103 parturients initially enrolled in the study, 11 were excluded for the following reasons: gestational diabetes mellitus, gestational hypertension, the use of contraindicated medications, and acute infectious diseases. During labor, 19 parturients were removed from the study: three due to inadequate epidural analgesia, seven proceeded to cesarean section, eight presented with suboptimal arterial visualization, and one was temporarily withdrawn from the study. Finally, data were collected from 73 participants. Among these, 13 (17.8%) developed a fever and were classified as the febrile group, while 60 (82.2%) remained afebrile (Fig. 1). Table 1 presents a summary of the primary demographic data and obstetric characteristics. The baseline demographics showed no significant differences between the febrile and afebrile groups (p $>$ 0.05). However, the duration of both the total labor process and the second stage of labor was significantly longer in the febrile group compared to the afebrile group (p 0.05).

The baseline spectral parameters of ultrasound blood flow did not differ significantly between the febrile and afebrile patients (p $>$ 0.05) (Fig. 2). Following the administration of ELA, there was a significant increase in the blood flow spectral parameters—PSV and EDV—for both the ATA and the PTA at 5 minutes post-anesthesia. This elevation continued progressively, with statistically significant deviations from the baseline observed at all subsequent time points (p 0.05). Moreover, at 10 minutes post-ELA, the PSV values for ATA and PTA were significantly higher in the febrile group than those in the afebrile group (p 0.05). Furthermore, the EDV for ATA and PTA in the febrile group displayed significant elevations at 15 minutes post-anesthesia when contrasted with the afebrile group (p 0.01). Multivariate logistic regression comparisons of $\mathrm{\Delta }$PSV and $\mathrm{\Delta }$EDV between groups after adjustment for total labour duration and whether or not oxytocin was used before ELA showed that significant differences could still be found. These results are consistent with unadjusted comparisons (Supplementary Table 1).

Post-epidural labor analgesia, there was a tendency for core body temperature to increase incrementally, reaching a peak towards the end of the second stage of labor. Subsequently, the temperature began to fall as the labor process concluded. Similarly, PSV and EDV of both lower extremity arteries showed a gradual rise following analgesia. To investigate the relationship between these variables, we conducted a Spearman correlation analysis focusing on PSV₃₀ and EDV₃₀ concerning T_peak, as presented in Table 2. The analysis revealed a positive correlation between PSV₃₀ and EDV₃₀ with T_peak, as shown in Fig. 3.

Using the changes in blood flow parameters post-analgesia, ROC curves were generated for the rate of change in the ATA and PTA 30 minutes after analgesia to predict ERMF (Fig. 4). Utilizing the Youden index (sensitivity + specificity – 1), we identified the optimal cut-off value and calculated the AUC, and the corresponding sensitivity and specificity for the ERMF prediction. The detailed statistics are shown in Table 3. The results demonstrate that $\mathrm{\Delta }$EDV has a superior predictive value compared to $\mathrm{\Delta }$PSV. The AUC values of the ROC curves for predicting ERMF after ELA using $\mathrm{\Delta }$EDV of ATA and PTA were 0.733 (sensitivity 69.2%, specificity 75.0%) and 0.713 (sensitivity 53.8%, specificity 88.3%), respectively. The corresponding optimal cut-off values were 29.94 and 33.10, respectively.