During the 10-year study period, there were 32,743 singleton deliveries (excluding elective cesarean sections) with gestational age ≥ 36 weeks of cephalic presentation and normally formed fetuses with intrapartum CTG recorded as per standard care at the John Radcliffe Hospital, Oxford (Figure 1). The selected gestational age of 36 completed weeks is in line with the threshold used in our previous work and in large randomized controlled trials on intrapartum CTG (23–25). We only included cases with at least 20 min of CTG recordings of acceptable signal quality for analysis, taken only at maternity admission or delivery units. The national guidance at the time and still is intermittent auscultation for the intrapartum surveillance of low-risk pregnancies, defined as the absence of any antenatal or intrapartum, fetal, maternal, placental, medical, or obstetric complications or risk factors. Figure 1 shows the study data flow chart. We excluded cases where the CTG recording commenced in the second stage of labor and those with cesarean delivery for indication other than “presumed fetal compromise.” A total of 27,927 deliveries were eligible for inclusion and analysis.

All data were routinely collected as part of standard clinical care and analyzed retrospectively. Outcome groups were based on the mode and timing of cesarean delivery for presumed fetal compromise (Figure 1). Infants delivered by cesarean section within 2 h of CTG monitoring (emergency cesarean delivery, Emergency CS) were compared with their counterparts delivered by cesarean section between 2 and 5 h (urgent cesarean delivery, Urgent CS). The remaining births formed the control group (Others).

Initial assessment was defined as the assessment of fetal wellbeing and clinical risk factors in women presenting to the unit with symptoms and signs of labor.

Sentinel events included placental abruption, uterine rupture, and cord prolapse. Birth trauma included intracranial laceration and hemorrhage due to birth injury; other birth injuries to central nervous system; birth injury to skeleton; and birth injury to peripheral nervous system.

Birth weight deviations were calculated with the adjusted Yudkin's chart percentiles (26) and reported in terms of small for gestational age (birth weights below the 3rd percentile) and large for gestational age (birth weights above the 97th percentile).

Severe compromise was defined as a composite outcome consisting of clinically relevant short-term perinatal outcomes collected at hospital level including stillbirth, neonatal death, seizures, neonatal encephalopathy, and need for intubation or intensive resuscitation followed by neonatal intensive care unit (NICU) admission for ≥48 h (15). Reported birth outcomes include severe compromise, Apgar score <4 at 1 min, Apgar score <7 at 5 min, and/or umbilical cord arterial pH <7.0 or <7.05 (27).

Onset of labor was defined by the attending clinician as per clinical practice (regular uterine contractions associated with cervical dilatation ≥3 cm) and documented in the routinely collected digital maternity notes.

Digital CTG data, sampled at 4 Hz as per the standard output of fetal monitors (typically Philips Avalon) were archived by a central monitoring system (Huntleigh Healthcare Ltd., Cardiff, UK). Computerized CTG features were calculated with the Oxford System (OxSys1.5) algorithms (15). OxSys analyses the CTG recordings in 15 min windows that move forward every 5 min; CTG characteristics are calculated for each window after heuristic noise removal algorithms. Pregnancy and labor data were derived from the Oxford Clinical Maternity Database, detailing maternal demographics, obstetric history, labor, delivery outcomes, and infant characteristics.

We considered only the first available 20–60 min of the CTG traces as part of the initial assessment (28), based on the recommended 20 min duration for CTG around labor onset (22) and additional time for the assessment of fetal behavioral states if they were absent in the first 20 min. Where available, the entire first hour of CTG was analyzed and the results reported. Where not, the longest available CTG segment (≥20 min) was analyzed instead. CTG recordings made after the first hour were not analyzed in this study.

In addition to standard CTG characteristics such as baseline rate, variability, decelerations, and accelerations, OxSys1.5 computes the decelerative capacity (DC), analyzing the entire FHR signal in a moving 15 min window to provide an average measure of downward movements (29). Lower DC values are measured in a normal trace without decelerations, whereas increased DC values are measured in deep, steep-sloped, and/or repetitive decelerations (15). Non-reactive trace is defined as DC <1 bpm and no accelerations as previously reported (15).

Data were analyzed in MATLAB version R2019b (The MathWorks Inc., Massachusetts, USA). Medians with 25 and 75th percentiles were calculated for continuous and ordinal variables. Incidence with percentages were calculated for categorical variables. Univariate analysis of categorical variables was performed with the X² test and Fisher's exact test with relative risks, odds ratios, and 95% confidence intervals as appropriate. The Kruskal-Wallis one-way analysis of variance test was used to compare the medians and interquartile range (IQR) of continuous variables.

Analysis of this routinely collected cohort of over 27,000 term births showed that an important proportion of fetuses with severe compromise or at risk of severe compromise had significantly different features of the first-hour CTG and presence of clinical risk factors such as thick meconium, small for gestational age, and maternal fever ≥38C. We considered three exclusive groups: infants delivered by CS for “presumed fetal compromise” within 2 h of starting the CTG (Emergency CS); between 2 and 5 h of starting the CTG (Urgent CS); and the rest of deliveries (Others). This data stratification allowed us to account for any effects the emergency or urgent cesarean sections had on labor outcomes, as well as to obtain estimates for the proportion of severely compromised infants for whom potential early signs were present. Severely compromised infants in these groups had higher rates of baseline FHR ≥150 bpm, non-reactive initial CTG, reduced short-term and long-term variability, less accelerations, and decelerative capacity amongst fetuses with severe compromise. Prolonged decelerations (≥3 min) were also more common in the first hour CTG of infants born with severe compromise. None of the severely compromised infants in the Urgent CS had a first hour baseline ≥160 bpm and only a small proportion had STV <3 msec. In this group, non-reactive initial trace and DC <1 bpm were stronger risk factors in about a third of severely compromised infants. There was a significantly higher proportion of babies with arterial cord pH <7.00 in the severely compromised babies with Urgent CS (Table 2), with the length of stay in NICU longest for the severe compromised babies in the Emergency CS group, suggesting that the components of the “severe compromise” varied between the groups. Umbilical artery acidemia was absent in the majority of severely compromised infants in Emegency CS and Others. The rate of decelerations in the initial CTG was not associated with severe compromise, but our data was “censored” by clinical practice.

In addition, we demonstrated that computerized analysis of the CTG adds value by permitting the analysis of large cohorts to support clinical guidelines.

Our data suggest that clinical guidelines for CTG interpretation for the initial trace may need to differ from those applied to the rest of labor. Perhaps there should be less emphasis on the presence of decelerations per se, unless they are prolonged (≥3 min). Instead, there should be a lower threshold for the definition of FHR “tachycardia” at this time (i.e., 150 bpm), an assessment for a non-reactive trace (i.e., lack of cyclicity or reactivity). Our data also confirm the relevance of and provide evidence for STV and LTV, as well as clinical risk factors such as thick meconium, small for gestational age, and others.

These findings suggest that a consistent assessment of the CTG together with known clinical risk factors around the onset of labor has the potential to identify early a small but important proportion of babies at risk of severe compromise. Our data shows that it is likely that some of the cases were “missed” in the group of Others and recognized late in the Urgent CS group. Some of the CTG characteristics important for the first-hour CTG assessment, such as reduced variability, decelerative capacity, and the non-reactive trace, are not typically detectable with intermittent auscultation (20).

The primary outcome of “severe compromise” was broadly defined to include cases which may not have their origins in hypoxic-ischaemic insults. Approximately half of small for gestational age fetuses attributable to placental insufficiency are missed in the antenatal period, and about 10% of them sustain severe compromise (30). Other vulnerabilities, including intrauterine infection, cerebral hemorrhage, and maternal hyperglycaemia (13, 18, 19, 31), may also alter fetal physiological adaptations to reduced oxygen during labor (18, 19) in a manner that is difficult to detect with conventional CTG (13, 18, 31). Therefore, in principle, clinical factors and fetal physiology should be considered for accurate interpretation of CTG in the initial assessment of women in labor. Our findings confirm the importance of both first-hour CTG and clinical characteristics.

Our findings are relevant for fetuses for whom cesarean delivery before or in early labor may avoid the more dramatic emergency deliveries and the attendant maternal and fetal risks (30). Furthermore, better risk assessments with CTG in the initial assessment could enhance individualized care, informed decision-making, and improve maternal engagement during labor (17, 32).

Using computerized methods, we studied a large, detailed, and complete, high-quality maternity database. Specific markers of brain development were not collected at the hospital level and therefore unavailable for our data analysis of routinely collected cohort data. We included all eligible infants resulting in nearly equal proportion of babies with CTG starting before and after the onset of labor, 46 vs. 54% respectively. Therefore, our findings are relevant to both periods, pre and early labor, largely around onset of labor where women are assessed in a maternity unit triage room or bays. This is particularly important because the definition of onset of labor and CTG initiation varies and depends, amongst other things, on the time the woman presents to the hospital.

Clinical practice has inevitably changed since the timespan of this dataset and the data for the subsequent 10-year period (2011–2020) were not fully available yet at the time of analysis, but their analysis will follow the presented format in this study to allow comparison.

A limitation of this study was the lack of information for the cervical dilatation at the time of initial CTG recording as well as the fact that our routinely collected data relate to a high-risk obstetric population, as per clinical CTG use in the UK.

We propose to build on the various univariate analyses presented in the current study to undertake multivariate analyses in future studies using larger datasets.