Adapted ST analysis during labor: relative versus absolute ST events, a case-control study

Abstract

Background

The value of ST analysis of the fetal electrocardiogram during labor to lower asphyxia and cesarean section rates is uncertain. Physiological variation of the electrical heart axis between fetuses may explain false alarms in conventional ST analysis (absolute ST analysis). ST events (alarms) based on relative T/QRS rises (relative ST analysis) correct for this variation and may improve diagnostic accuracy of ST analysis.

Aims

To compare the diagnostic accuracy of absolute and relative ST analysis with regard to fetal acidemia.

Study design

Retrospective case-control study.

Subjects

20 healthy women with an uncomplicated pregnancy monitored with ST analysis during labor: 10 cases (umbilical cord artery pH < 7.05) and 10 controls (pH > 7.20).

Outcome measures

Sensitivity, specificity, positive and negative likelihood ratio.

Results

In 16 of the 20 patients a total of 54 absolute ST events were reported. Two reviewers classified the cardiotocograms; in cases 29% of the absolute ST events were significant, in the controls it was 19%. Relative ST analysis versus absolute ST analysis showed a sensitivity of 90% (55−100%) vs. 70% (35−93%), a specificity of 100% (69–100%) vs. 70% (35−93%), a positive likelihood ratio of infinity vs. 2.3 (0.8−6.5), a negative likelihood ratio of 0.1 (0.0−0.6) vs. 0.4 (0.2−1.2), and diagnostic odds ratio of infinity vs. 5.4 (0.8−36.9). McNemar showed no statistical significant difference between the sensitivity and specificity of the methods.

Conclusions

We observed higher positive and lower negative likelihood ratios for relative ST analysis in comparison to absolute ST analysis. In this small study we found no statistical difference. Relative ST analysis should be studied in a larger study.

Keywords:

• Diagnostic accuracy
• fetal electrocardiography
• fetal monitoring
• ST event
• metabolic acidosis
• T/QRS

Introduction

To prevent neonatal asphyxia, clinicians need a reliable method for intrapartum fetal surveillance to intervene in time when necessary. Cardiotocography (CTG) has high sensitivity, but low specificity [1]. Additional techniques, such as fetal blood sampling or ST analysis, are used to improve specificity. The STAN method® (Neoventa, Mölndal, Sweden), using absolute ST waveform analysis, initially showed promising in terms of lowering asphyxia rates [2]. However, meta-analyses show inconclusive results regarding the effect of absolute ST analysis on cesarean section rates and fetal metabolic acidosis [3]. An explanation may be that the clinical relevance of these absolute ST events depends on CTG classification and that absolute ST events do not always reflect fetal distress.

ST analysis is based on the finding that hypoxia in fetal lambs is related to changes in the ST segment of the fetal electrocardiogram (fECG) [4]. As a result, the STAN® method was developed for fetal surveillance during labor at term in humans. It generates three types of alarms if ST changes occur: episodic T/QRS rise, baseline T/QRS rise and biphasic events. Baseline and episodic events, which are based on T-top rises, are expressed as absolute rises in T/QRS ratio from baseline. Episodic events disappear within 10 min, while baseline events last for more than 10 min. Biphasic events, which are based on a negative slope of the ST segment, are reported to be related to acute stress and absent or exhausted compensation mechanisms [5].

The clinical relevance of these ST events depends on CTG classification. The CTG is classified into four categories according to FIGO guidelines; ST events are defined as significant or not, depending on this CTG classification, following the STAN guidelines [6,7]. Kwee [8] showed that ST events occurred in more than 50% of normal CTG fragments during the first stage of labor and were evenly distributed in normal, intermediary and abnormal fragments in the second stage. These findings emphasize the fact that the STAN method^® strongly depends on correct CTG interpretation, which is known to have high inter- and intraobserver variability [9].

In previous studies we found a physiological explanation for these false positive absolute ST events and presented a solution: relative ST events (relative T/QRS rises from baseline) [10,11]. We demonstrated that normal variations in the electrical heart axis between fetuses result in differences in the fetal vectorcardiogram, which yields different T/QRS baseline values. Moreover, we found that this T/QRS baseline value is positively related to the number of absolute ST events [10]. However, no association was found between T/QRS baseline value and neonatal outcome [12].

We hypothesize that ST analysis can be improved if relative T/QRS rises (relative ST events) from baseline are used instead of absolute T/QRS rises (absolute ST events). This hypothesis applies to episodic and baseline ST events. Previously, we observed that relative events are more accurately related to fetal metabolic acidosis than absolute events. We found a specificity of 100% for the relative events versus 40% for the absolute events, with a sensitivity of 90% for both methods [11]. However, in the previous study we omitted CTG classification. As a result it was impossible to determine whether the absolute ST events were significant or not. In this study we compare diagnostic accuracy of relative ST events (new method) with significant absolute ST events (current method).

Materials and methods

Design & patients

We performed a case control study using the dataset collected by van Laar et al. [13]. The cases (i.e. fetuses with acidemia) had a cord arterial pH below 7.05, whereas the controls (i.e. fetuses without acidemia) had a cord arterial pH above 7.20. The original study was approved by the medical commission of the Máxima Medical Center and conducted in accordance with the Declaration of Helsinki [13].

Patients were retrospectively, consecutively, selected in two Dutch tertiary care teaching hospitals (University Medical Center Utrecht, UMCU Utrecht and Máxima Medical Center, MMC Veldhoven). Inclusion criteria were healthy women with healthy singleton fetuses of at least 36 weeks of gestation with good quality intrapartum fetal ECG recordings (no ectopic beats or missing data until at least 10 min before birth) and available arterial and venous acid base status from umbilical cord blood. Exclusion criteria were fetal growth restriction, fetal congenital anomalies, and use of medication other than oxytocin or epidural analgesia.

Data acquisition

Fetal ECG measurements were obtained during delivery with a single-helix scalp electrode (Goldtrace^®, Neoventa Medical, Mölndal, Sweden), a maternal skin electrode and a STAN S31 monitor (Neoventa Medical, Mölndal, Sweden). We defined absolute events as episodic, baseline or biphasic ST events that were reported in the Event Log^® by STAN^®. Each up-to-20-min CTG recording preceding an absolute event was selected. Two obstetricians (>10 years experience) and STAN trainers (AK and MP) classified these CTG fragments as normal, intermediary, abnormal or preterminal following the FIGO guidelines [6]. Likewise, they classified each absolute ST event as significant or non-significant based on the STAN guidelines. The reviewers were blinded to postpartum pH values. Discrepancies – regarding ST event classification - were resolved by discussion. A positive test for absolute ST analysis (STAN method^®) was defined as at least one significant ST event per patient.

Our rationale and method of calculating the relative events has been described previously [11]. A positive test for relative ST-analysis (our method) was defined as a relative T/QRS rise from baseline above 70% at least once independent of CTG analysis. We chose this cut off value, since it delivered optimal accuracy in previous analysis [11].

Statistical analysis

Test characteristics and 95% exact confidence intervals were calculated with Episheet © in Microsoft Excel 11 [14]. SPSS Statistics version 22 was used for additional statistical analysis. To compare patient characteristics we used a Mann–Whitney U test for continuous variables and a Fisher’s exact test for discrete variables. Test characteristics were compared using McNemar’s test. The alpha level of .05 indicated statistical significance.

Results

In total 20 patients were included: 10 cases (umbilical artery pH value <7.05) and 10 controls (umbilical artery pH value >7.20). Baseline characteristics are shown in Table 1. Groups were similar in terms of parity, gestational age, and birth weight.

Table 1. Clinical characteristics and outcome parameters.

	pH < 7.05 (n = 10)	pH > 7.20 (n = 10)
Nulliparous	8 (80)	4 (40)
Gestational age – days	283 [8]	278 [11]
Analgesia	5 (50)	6 (60)
Oxytocin	7 (70)	4 (40)
FBS – frequency	1.4 [1.3]	0
Birth weight – gram	3414 [432]	3643 [562]
Apgar score after 1 min	6 [2]	9 [1]
Apgar score after 5 min	8 [1]	10 [0]
Cord arterial pH	6.98 [0.07]	7.26 [0.03]
Cord arterial base excess – mmol/l	−17 [4]	−5 [2]
Cord venous pH	7.08 [0.10]	7.33 [0.04]
Hospital admission	4 (40)	0 (0)
Medium Care	(20)	–
NICU	2 (20)	–
Operative delivery	4 (40)	(0)
Ventouse delivery	4 (40)	–
Cesarean section	0 (0)	–

The values are represented as number (%) or mean [standard deviation]. The values depicted in bold are significant using Fisher exact test in nominal variables and Mann–Whitney U test in continuous variables. FBS: fetal blood sampling; NICU: neonatal intensive care unit.

CTG and absolute ST event classification

In total 54 CTG fragments preceding absolute ST events were classified. AK and MP identically classified 85% of 54 CTG fragments. Discrepancies that led to different ST event classification were resolved by discussion. In the case group 29% of the 38 absolute ST events were significant; seven out of ten cases had at least one significant absolute ST event. In the control group, 19% of the 16 absolute ST events were significant; three out of ten controls had at least one significant ST event (Figure 1).

Figure 1. Significant absolute ST events per patient.

Figure 2 shows the distribution of absolute and relative ST events toward the time of delivery. The median time between the first event and birth was 14 min (ranging from 2 min after birth to 4 h and 38 min before birth) for significant absolute events and 4 h and 8 min (ranging from 11 min to 8 h and 7 min before birth) for relative events.

Figure 2. ST events and time-till-delivery. Pink bar, duration of measurement; navy bar, relative ST event (T/QRS rise from baseline >70%); blue dot (empty), non significant absolute ST events; Blue dot (filled): significant absolute ST event.

Test characteristics

Test characteristics of absolute and relative ST analysis are shown in Table 2. Absolute ST analysis showed a sensitivity and a specificity of 70%. Relative ST analysis showed a sensitivity of 90% and a specificity of 100%. Both positive and negative likelihood ratios of absolute ST analysis indicated a small diagnostic effect, while the negative likelihood ratio of relative ST analysis indicated a moderate diagnostic effect. The positive likelihood ratio, the diagnostic odds ratio, and their confidence intervals could not be calculated for relative ST analysis as the specificity was 100%. These variables approximate to infinity.

Table 2. Test characteristics.

	Sensitivity (95%CI)	Specificity (95%CI)	LR+ (95%CI)	LR- (95%CI)	OR (95%CI)
Absolute ST analysis	70% (35−93)	70% (35−93)	2.3 (0.8−6.5)	0.4 (0.2−1.2)	5.4 (0.8−36.9)
Relative ST analysis	90% (55−100)	100% (69−100)	∞	0.1 (0.0−0.6)	∞

95%CI: 95% confidence intervals; LR+: positive likelihood ratio; LR-: negative likelihood ratio; OR: diagnostic odds ratio; ∞: infinity.

McNemar’s test did not reveal significant differences between absolute and relative ST analysis regarding the true positives (p = .63) or the true negatives (p = .25).

Discussion

In this pilot study, we compared the diagnostic accuracy of absolute ST events (STAN method) and relative ST events with regard to fetal acidemia. We found a non-significant trend toward better test characteristics in relative ST analysis compared with absolute ST analysis. As confidence intervals of the positive likelihood ratio and diagnostic odds ratio of the relative method could not be determined, these variables could not be compared statistically to the absolute method. It seems that relative ST analysis is more promising for better diagnostic accuracy than absolute ST analysis, as these test characteristics approximate to infinity.

In a previous study, we found that the specificity of absolute ST-analysis was 40% in contrast to 70% in the current study [11]. The difference between these studies can be explained because CTG classification and biphasic ST events were not taken into consideration in the previous study. By using CTG classification many false positive absolute ST events can be ignored. It emphasizes how the STAN method^® depends on correct CTG classification, which is a major drawback of absolute ST analysis. CTG interpretation, which is known to be subjective, has low-to-moderate inter- and intraobserver agreement [9,15,16]. The new method has shown better specificity as a stand-alone objective measurement of fetal well-being than the old one, as it does not depend on CTG interpretation.

Moreover, the value of a deteriorating CTG without ST events should not be underestimated. Another earlier described pitfall is that clinicians do not assess CTG continuously and feel “safe” as long as no ST event arises when the STAN method^® is used in clinical practice. This way, slow CTG alterations (including baseline elevation or diminished variability) could be overlooked, resulting in adverse neonatal outcome [15].

In three healthy control patients (pH > 7.20) significant absolute ST events emerged. The time gap between event and delivery was less than 20 min. One event emerged within one minute after delivery. All these cases ended as a spontaneous vaginal delivery; no interventions that could improve postpartum pH, such as vacuum extraction or tocolytic agent, were used. This is why we can confirm that these cases are truly false positives of absolute ST analysis.

Biphasic ST events did not influence diagnostic accuracy in the current study, as none were significant. So far, we did not include biphasic events in relative ST analysis, as they are not influenced by T/QRS baseline and electrical heart axis. Besides, Becker et al. (17) showed that significant biphasic events do not have additional predictive value for fetal distress.

Remarkably, the sensitivity of absolute ST events with CTG information (70%) was lower than without CTG information (90%). These false negatives could be explained by CTG classification difficulties, even though experts classified the fragments. Suboptimal or abnormal CTG fragments could be classified as normal or suboptimal, which would change ST events to be classified as significant or non-significant.

A strength of this study is that two independent experienced obstetricians (>10 years experience) and STAN trainers classified the CTG fragments. They strongly agreed on the CTG classification. The test characteristics of the STAN method would probably be worse in clinical practice, as less experienced doctors and midwives will usually interpret the CTG in daily routine. Westerhuis has shown lower agreement on CTG classification among less experienced doctors and experts, than within a group of experts [16].

We observed a large difference in the time-to-delivery interval from the first event between both methods. Significant absolute ST events probably led to clinical decisions to end delivery, which would explain the short median of 14 min. As this is a retrospective study, relative ST events were not available at the time of measurement, so they could not influence the clinical course. Relative ST events emerged earlier, which implies that relative ST analysis could predict metabolic acidosis at an earlier stage.

The physiological rationale for relative ST analysis

Each contraction of the heart is initiated by electrical potential propagation. The direction and the amplitude of the electrical potential change during the heart cycle; these factors depend on the direction of propagation (as different parts of the heart depolarize after each other) and the thickness of the muscular wall. The direction in which the electrical potential has maximal amplitude is called the electrical heart axis. If the fetal ECG signal is measured with a single electrode (scalp electrode), the amplitudes of the various ECG waves reflect the projection of the electrical potential in the direction of that – single – electrode. If the electrode is aligned in the same direction as the electrical heart axis, the largest ECG signal (largest amplitude) is measured. If the alignment differs, this influences the ECG amplitude. Since the various waves in the ECG are not all parallel with the electrical heart axis, different waves are influenced to a different extent, causing changes in the ECG waveform. As a result, the alignment between the electrical heart axis and scalp electrode can influence ST events, regardless of fetal condition.

The current STAN method^® normalizes the signal to address this problem; the T-top amplitude is normalized against the QRS complex as a T/QRS ratio. This method ignores the fact that the propagation direction in the depolarization phase (QRS) is not parallel (or antiparallel) to that in repolarization phase (T-top).

If the electrode alignment is almost perpendicular to the propagation during repolarization, a small T- wave arises. Hypoxia-induced T-wave changes will be small; they will not be detected as a substantial change. This leads to the false-negative absence of ST events. On the contrary, optimal alignment between the electrode and electrical heart axis leads to large amplitudes and thus larger T/QRS rises. The threshold to raise a ST event is exceeded more easily, even in normally oxygenated fetuses, which leads to false-positive ST events.

If events were raised based on a relative change from baseline instead of an absolute change, it would compensate for these inter-patient differences in electrical heart axis. Both false-positive and false-negative events will then be reduced.

Limitations

The limitations of this exploratory study are the small sample size and the contrasting groups (no complete spectrum of pH values). We chose this setup, as no previous research has been conducted on this subject yet. This way we could explore the new method as a proof of principle.

We found no significant differences in diagnostic accuracy. This could be explained by the small sample size.

Furthermore, relatively short CTG fragments preceding each ST event were available for CTG classification. We chose to present only 20-min fragments to the reviewers, since the STAN method^® requires this minimum and the main CTG criteria can be determined (basal heart rate, accelerations, deceleration depth and duration, and variability). Besides, we wanted to ascertain optimal uniformity in the CTG length preceding each ST event.

Implications for further research

Relative ST analysis should be studied in a larger population with a complete spectrum of pH values. In such a large sample a new cut off for relative ST rises might be determined; consequently this new cutoff should be validated in an external dataset.

Registration of fECG was obtained with scalp electrodes in this study, whereas noninvasive abdominal registration might be used in future studies. If multiple abdominal electrodes are used, this will address the problem of electrical heart axis variation more specifically. Then the true electrical heart axis could be determined for each individual fetus. That way we could use an adjusted algorithm to calculate relative T/QRS rises, based on the individual angle between the electrode alignment and the fetal electrical heart axis. This would be a more specific solution than the general solution proposed in this study (relative T/QRS rises). This may improve the diagnostic accuracy of ST analysis even more. However, transabdominal fECG acquisition will pose new challenges, such as lower signal quality [17].

In conclusion, relative ST events are promising to improve fetal monitoring during labor. This retrospective study showed a non-significant trend toward better diagnostic accuracy in relative ST analysis compared to absolute ST analysis (the current STAN method). More research should be conducted to determine the clinical value of this new method.

Acknowledgements

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Disclosure statement

RV is a shareholder in Nemo Healthcare BV, the Netherlands. SO initiated the scientific research from which Nemo Healthcare originated, there is no financial relationship between Nemo Healthcare and SO. No potential conflict of interest was reported by the author(s).