INTRODUCTION

Since its initial implementation in 1968, electronic fetal monitoring via cardiotocogram (CTG) has aimed to facilitate early detection of fetal hypoxia, thereby permitting timely interventions to prevent adverse neonatal outcomes. Despite its widespread adoption, the predictive value of CTG features remains limited, with studies indicating a high negative predictive value but a low positive predictive value for fetal hypoxic injury¹. Moreover, the subjectivity inherent in interpreting CTG patterns leads to significant interobserver variability, which poses challenges to consistent clinical management. Previous guidelines, notably those issued by The International Federation of Gynecology and Obstetrics (FIGO), have undergone multiple revisions, with the most recent being in 2015, aiming to standardize interpretation criteria based on fetal physiological parameters². Despite these efforts, variability persists, especially regarding deceleration interpretation, accelerations, and overall categorization, which may contribute to diagnostic inaccuracies and inconsistent clinical responses³. The overall assessment of a CTG as normal, suspicious, or pathological, and making the right decision at birth, remains one of the greatest challenges. Increasingly, it is argued that it would be appropriate to define individual CTG parameters that have both good interobserver agreement and the best prognostic value in assessing the fetal condition. One example of such a parameter could be the total width of decelerations in a 30-minute CTG interval, given that it is the duration of deceleration that has a very good prognostic value in predicting neonatal acidosis^4,5. The proposed parameter is likely to be easy to estimate and the possibility of misinterpretation is low. If this objectively easy-to-measure parameter also has a good negative and positive predictive value in assessing the fetal status at the time of investigation, it could be an ideal CTG parameter.

It has been reported that in the presence of meconium-stained amniotic fluid, even after emergency cesarean section, a high proportion of newborns are born in a serious condition^6,7. This shows that the current interpretation of the CTG makes it difficult to predict the actual fetal condition and to make a timely decision. For this purpose, we wished to assess if the evaluation of individual CTG parameters could more accurately predict fetal condition. The aim of this study was to assess interobserver agreement among investigators in assessing fetal status during labor using selected CTG parameters, according to the FIGO 2015 CTG assessment guidelines². We evaluated whether the assessment of individual CTG parameters is more objective and has better inter-investigator agreement than assigning the entire CTG to separate categories (normal, questionable, and pathological). The second hypothesis is that the clinical experience of investigators positively affects the interpretation of CTG parameters. Addressing these hypotheses may serve as the starting point for a future study on the assessment of fetal status in meconium-contaminated amniotic fluid.

METHODS

This multicenter interobserver agreement study involved two leading Lithuanian maternity institutions, with ethical approval obtained from the local ethics committee. The study took place in the Hospital of Lithuanian University of Health Science Kauno Klinikos (tertiary maternity unit) and Klaipeda University Hospital (maternity unit, together with a midwife-led unit), both having about 3000 deliveries per year. Study period was April 2024. Delivery records from term singleton pregnancies presenting with meconium-stained amniotic fluid and without major fetal anomalies or complicating factors were included. A total of 84 CTGs (50 and 34 from different hospitals), each with a minimum duration of 120 minutes during the active labor phase, were randomly selected for detailed analysis. These recordings were subdivided into four equal segments of 30 minutes each for comparative evaluation. Only recordings of adequate quality with no artifacts were considered. CTG with poor quality tracings, mostly related to the pushing stage of birth and the different pushing positions used, were not considered.

Six clinicians, each blinded to clinical outcomes, independently assessed the CTGs on standardized parameters outlined by the FIGO 2015 guidelines² – specifically, baseline fetal heart rate, variability (categorized as normal, reduced, or absent), presence and nature of decelerations, their morphology and type, total deceleration duration, and bradycardia. The observers, comprising both highly experienced practitioners (>10 years, three doctors, obstetricians, and gynecologists, including those from both hospitals) and less experienced trainees (three young doctors, residents with <3 years’ experience from the teaching university hospital), received prior standardized training on the guidelines. The evaluators had no time limit for the assessments, and in each hospital they analyzed the same cardiotocographic traces.

The manuscript was prepared in accordance with the GRASS to ensure transparent and comprehensive reporting. Data analysis involved calculating Cohen’s and Fleiss’ kappa statistics to determine interobserver agreement, complemented by ICC for interval data such as deceleration duration. Kappa values were interpreted according to established categories (weak <0.20; poor 0.21–0.40; moderate 0.41–0.60; substantial 0.61–0.80; almost perfect ≥0.80). PA was interpreted using conventional thresholds, with >80% agreement considered acceptable.

RESULTS

The data are summarized in Table 1. A total of 663 variables were analyzed to determine interobserver reliability. The findings demonstrated that most CTG parameters – specifically, variability, bradycardia, presence or absence of decelerations, reduced variability accompanied by tachycardia, and the total deceleration duration – exhibited good to moderate levels of agreement. These parameters yielded kappa coefficients within the range of moderate to very good concordance (kappa: 0.67–1). Conversely, assessments pertaining to accelerations showed only moderate or poor agreement (kappa: 0.29–0.47). The evaluation of deceleration type and morphology (shape) showed the lowest interobserver concordance, with kappa values between 0.12 and 0.23, indicative of poor or weak agreement.

The overall classification of CTGs into categories– normal, suspicious, or pathological – demonstrated limited reproducibility; specifically, Cohen’s kappa for the T1–T2 group was 0.44, reflecting moderate agreement, whereas the other evaluator group (T3–T6) exhibited only minimal agreement (kappa: 0.19). When stratified according to the Lithuanian national guidelines, the assessment of CTG categories (normal, suspicious, or pathological) resulted in kappa values of approximately 0.48, 0.42, and 0, respectively, for T1–T2, and only weak or poor agreement ranging from 0.22 to 0 among T3–T6.

DISCUSSION

Application of the 2015 FIGO CTG scoring criteria² yielded moderate to high interobserver agreement for parameters such as baseline fetal heart rate, variability, presence or absence of decelerations, and total deceleration duration, corroborating previous findings by Rei et al.⁸. These features are comparatively straightforward to interpret objectively following standardized training, thereby minimizing subjective variability. In contrast, assessments of accelerations demonstrated lower interobserver concordance (kappa: 0.29–0.47) in our cohort compared to prior studies (kappa: approximately 0.72)^9,10, potentially attributable to recent literature highlighting interpretative ambiguities – for instance, the differentiation of true versus pseudo-accelerations – confusing clinical assessment⁸. The work by Al Fahdi et al.¹⁰ has underscored that misinterpretation of accelerations may impact reliability, which could explain the lower kappa values observed. Recent evidence¹¹ further supports this, reporting acceleratory agreement as low as 0.09 (poor).

Furthermore, the agreement concerning the classification of deceleration types (early, late, variable) was similarly low across studies, with coefficients ranging from 0.19 to 0.38^8,9,12. Our findings align with this trend, yielding kappa values of only 0.12–0.23, indicating substantial interobserver variability.

The overall CTG scoring – differentiating between normal, suspicious, and pathological – demonstrated moderate agreement between experienced observers (T1–T2; kappa: 0.44), consistent with prior research, such as that of Rei et al.⁸ who reported kappa=0.60. However, less experienced examiners (T3–T6) only achieved poor to weak concordance (kappa: 0.19). This discrepancy underscores the necessity of continuous education, ongoing training initiatives, and long-term mentorship to enhance the consistency and accuracy of CTG interpretation.

Notably, the classification of CTGs into categories (normal, suspicious, and pathological) was particularly unreliable. Kappa values between T1 and T2 ranged from 0 to 0.48 for this parameter: normal CTG 0.48 (moderate agreement), suspicious CTG 0.42 (moderate agreement) and pathological CTG 0 (poor agreement). The results were significantly worse for T3, T4, T5 and T6, with poor to weak agreement in the assignment of CTGs to individual categories, ranging from 0 to 0.22 (normal 0.22, suspicious 0.17 and pathological 0). The results are also surprising in that they contradict the notion that normal and pathological CTGs are the easiest to detect, while the category of suspicious CTG is more difficult to identify. International studies indicate the best interobserver agreement in the normal CTG group (kappa: 0.67–0.71) and in the pathological CTG group (kappa: 0.57), while suspicious CTG had a lower agreement (0.36–0.52)^8,13. In our study, all observers had trouble reaching a consensus on the pathological CTG (kappa: 0). Given that the identification of a pathological CTG is critical for urgent decision-making, these findings raise concerns regarding the current reliability of categorical assessments.

An innovative aspect of this study was the emphasis on the objective quantification of deceleration duration. The calculated kappa values, ranging from 0.48 to 0.80, demonstrate that estimation of total deceleration duration is one of the most reliably interpreted parameters of CTG with minimal influence from clinician experience and enhances the case for its routine and objective measurement in clinical practice. Cahill et al.⁴ have shown very convincingly in their study that the duration of deceleration has the best prognostic value in predicting neonatal acidemia compared with other CTG parameters⁴. Although decelerations are one of the most readily identifiable CTG features with the best prognostic value, it should be noted that high agreement is achieved by assessing whether decelerations are present and how much time the fetus spends in deceleration, but not by assessing the types of decelerations. The debate about the value of different decelerations arises from the work of Xodo and Londero¹⁴ who refuted the correlation of distinct types of decelerations with distinct physiological mechanisms for their occurrence. The mentioned authors argue that there is no better prognosis for so-called ‘physiological’ decelerations, and that all decelerations are ‘bad’ and have a negative impact on the fetal condition. The fetal deterioration and the development of acidemia are not determined by the type of deceleration (variable, early, late), but by its duration and the residual time of the basal rate.

CONCLUSIONS

The overall evaluative reliability of CTG interpretation demonstrates only moderate to low interobserver concordance. Key CTG parameters – including baseline fetal heart rate, variability, presence or absence of decelerations, and the duration of decelerations – are features that can be reliably quantified and interpreted with high reproducibility following standardized training, independent of clinical experience.

Given the limited agreement observed in classifying CTGs into categorical designations (i.e. normal, suspicious, or pathological), it would be indicated to discuss whether it is worth continuing to rely on such CTG categorical stratification. Instead, clinical decision-making should primarily be guided by more objective, reproducible features – such as baseline rate, variability, bradycardia, presence or absence of decelerations, and the cumulative deceleration area within a defined interval (e.g. 30 min) –which demonstrate higher interobserver reliability and prognostic validity.