Dr S Uccella, Department of Obstetrics and Gynaecology, University of Insubria, Piazza Biroldi, 1 – Varese, 21100, Italy. Email email@example.com
Objective To evaluate whether a standardised algorithm to interpret fetal heart rate (FHR) tracings during the entire length of labour can predict umbilical artery base excess at birth, and to investigate its inter- and intra-observer reproducibility.
Design Retrospective study.
Setting Obstetrics and gynaecology department at a tertiary referral centre in a university hospital.
Population Group 1: 152 consecutive, generally low-risk, labouring women. Group 2: mixed group of 30 women who delivered a fetus with pH < 7.00 and 30 women who delivered a fetus with pH ≥ 7.00.
Methods Intrapartum FHR tracings were retrospectively and blindly evaluated by two independent assessors using an algorithm proposed by Ross and Gala to predict fetal base excess at birth.
Main outcome measures The accuracy in predicting the base excess values of newborns was expressed as the proportion of FHR tracings in which the operator was able to correctly calculate the actual base excess at birth (approximation of ±2 mmol/l). Inter- and intra-observer reproducibility were estimated using the Pearson correlation coefficient.
Results In the group of 152 low-risk labouring women, the two assessors correctly predicted the umbilical artery base excess in 73.1 and 76.3% of cases, respectively. Inter-observer (Pearson correlation coefficient = 0.75) and intra-observer (Pearson correlation coefficient = 0.80 and 0.82 for the first and second assessor, respectively) reproducibility was very good. In the 30 fetuses that were acidemic, the first and second observers correctly predicted base excess values in 23 (76.7%) and 21 (70%) cases, respectively (inter-observer reproducibility, Pearson correlation coefficient = 0.72).
Conclusions The algorithm proposed by Ross and Gala may be a valuable tool to estimate changes in umbilical base excess during active labour, with a high inter- and intra-observer reproducibility.
Electronic fetal heart rate (FHR) monitoring was introduced in the late 1960s with the aim of enabling clinicians to identify hypoxic fetuses potentially at risk of compromise, who may benefit from expedited or immediate delivery,1,2 and it still represents the preferred modality of fetal surveillance during labour worldwide.3,4 Although fetal heart rate monitoring has been a mainstay of intrapartum management,5 the positive predictive value for metabolic acidemia at birth is extremely low, thus leading to unnecessary obstetric interventions without improvements in neurologic morbidity, including cerebral palsy, infant mortality, and other standard measures of neonatal wellbeing.6
Unfortunately, in the past 40 years the obstetric community has been unable to reach a broad consensus on a standardised approach to the management of most FHR monitoring patterns; as a consequence, debates over the correlation between intrapartum asphyxia and abnormal FHR patterns, as well as over the timing of perinatal neurologic injuries, have taken place more often in courtrooms than in the medical literature. The poor reproducibility of visual interpretation of intrapartum FHR tracings represents a well-known weakness of fetal monitoring. Several studies have tried to evaluate inter- and intra-observer reliability of the visual analysis of FHR records, showing high variability and inconsistency.7–11 A decade ago, based on experimental studies on animal models and retrospective analysis of human intrapartum adverse events, Ross and Gala12 examined the correlation between FHR patterns and the magnitude of acidemia at birth, expressed in terms of base excess. The aim of their insightful revision of the existing evidence was to generate an algorithm that permits interpolation of fetal base excess values during the course of labour. By combining the assumed base excess values at the initiation of labour with an assessment of the rate of base excess decrease throughout the period of fetal monitoring (calculated on the basis of the stage of labour plus the type and duration of varying FHR patterns), the proposed model allows a real-time estimation of base excess changes. However, the model proposed by Ross and Gala has not been subjected to scientific validation, and, as such, has never been implemented into clinical practice.
The aim of the present study was to investigate the accuracy of an algorithm for predicting umbilical artery base excess at birth, by retrospectively analysing a series of FHR tracings both in a low-risk population and in the setting of fetal intrapartum acidemia. As a secondary aim, we sought to evaluate the reproducibility of this algorithm, in terms of inter-observer and intra-observer variability.
A two-step retrospective study was performed, in order to determine: (1) the predictive ability of the algorithm proposed by Ross and Gala in a generally low-risk population; and (2) the validity of this model in the setting of fetal asphyxia.
Selection of FHR tracings
A student midwife not involved in the research project, and unaware of the purpose of the study, was asked (under direct medical supervision) to collect intrapartum FHR tracings of consecutive women who delivered at the Obstetric Department of the University of Insubria between January and April 2010, and who fulfilled the following inclusion criteria: (1) term pregnancy (≥37 weeks of gestation); (2) singleton gestation; (3) diagnosis of active labour; (4) electronic FHR monitoring during the entire length of labour; (5) duration of labour ≥20 minutes; and (6) successful umbilical blood sampling at birth, with validated pH and base excess data from both an artery and a vein. The sampling was considered correct when the venous–arterial pH difference was ≥0.02.13,14 We excluded patients who underwent caesarean section before the onset of labour and those who did not have available cardiotocographic tracings for the entire length of labour (e.g. unreadable traces or intermittent auscultation of fetal heart rate during part, or the entire duration, of labour). Women were included independently of the mode of delivery (caesarean section, operative, or uncomplicated vaginal delivery).
A coordinator not involved with the investigative portion of the study identified women who met the same inclusion and exclusion criteria as in group 1, and who delivered a newborn with an umbilical artery pH < 7.00 at birth, by searching the departmental obstetric database, which includes details of all deliveries between 2003 and 2010. To mitigate the potential for review bias arising from knowledge of fetal acidemia, the coordinator selected an equal number of women who delivered a non-acidotic newborn (umbilical artery pH ≥ 7.00) within 2 days of a woman who gave birth to an acidotic newborn, and who also met the same inclusion criteria. Thus, the study cohort for group 2 included a mixed population of acidotic and non-acidotic fetuses, in a 1:1 ratio. FHR tracings were retrieved from the medical charts.
Our institutional protocol for managing labour dictates that continuous electronic fetal monitoring must be used in high-risk pregnancy or in the instance of significant meconium-stained liquor, abnormal fetal heart rate detected by intermittent auscultation, maternal pyrexia, onset of bleeding during labour, and oxytocin use for induction or augmentation of labour. Moreover, whenever one-to-one midwifery care cannot be guaranteed because of a busy labour ward, electronic FHR monitoring is also recommended for low-risk labouring women. In our own labour and delivery room, ancillary tests to rule out acidemia, such as fetal scalp blood sampling, are not used. The active phase of labour is diagnosed if: (1) the cervix is effaced; (2) the cervix is at least 3-cm dilated; and (3) regular uterine contractions occur at least every 5 minutes, lasting a minimum of 20 seconds. Our intrapartum protocol defines FHR patterns according to the nomenclature issued by the Royal College of Obstetricians and Gynaecologists.15,16 FHR is also recorded continuously until birth in the case of operative delivery. When the FHR pattern is non-reassuring, various measures to promote fetal wellbeing are commonly initiated. These include maternal repositioning, reduction of uterine activity (discontinuation of oxytocin or administration of a tocolytic agent), an intravenous fluid bolus, correction of maternal hypotension, and alteration of the second-stage labour pushing effort. If the FHR pattern does not return to normal or evolves into a more serious pattern, consideration is given to the urgent delivery of the fetus.
In the event of an emergency caesarean delivery, the specific logistics, facilities, and staffing at our institution allow a time interval between the end of FHR recording and the initiation of the procedure of just a few minutes.
The algorithm proposed by Ross and Gala (the object of the present study) is not yet used in clinical practice.
Analysis of FHR tracings
All cardiotocographic tracings were retrospectively and independently evaluated by two obstetrics consultants, blinded to patient identity. Both observers were blinded to the clinical data of the patient, the outcome of delivery, the base excess value at birth, and the tracing assessment of the colleague on duty who provided care to the woman during labour. At the time of retrospective evaluation, the only information available to the observers were the length of labour, the duration of the second stage, and the cardiotocographic tracing in its entirety. At the time of writing, both observers have been consultant obstetricians for over 2 years in hospitals handling on average 1800 and 1200 deliveries per year, respectively.
Intrapartum tracings were interpreted by the two observers with the aim of predicting umbilical artery base excess at birth, using the algorithm proposed by Ross and Gala (Figure 1).12 In brief, the algorithm anticipates that the average fetus enters labour with a base excess of −2 mmol/l; during the active phase of labour, the fetal base excess is expected to decrease by −1 mmol/l every 3–6 hours, with a further decrease of −1 mmol/l per hour of second stage, assuming ‘normal labour stress’ (i.e. a reassuring FHR pattern). Moreover, the algorithm allows the interpolation of changes in base excess with varying FHR patterns. In particular, repetitive, typical, and variable decelerations are expected to decrease the buffer base by −1 mmol/l every 30 minutes, repetitive, late, or atypical variable decelerations are expected to decrease the buffer base by approximately −1 mmol/l every 10 minutes, and severe bradycardia is expected to decrease the buffer base by −1 mmol/l every 2–3 minutes. By applying this algorithm, the two assessors interpolated the probable base excess throughout labour based on the range of base excess reductions for normal first- and second-stage labour, and the duration of FHR monitoring patterns. Institutional Review Board approval was obtained before beginning the study.
The accuracy in the prediction of the level of fetal acidemia was expressed as the proportion of cardiotocographic tracings in which the operator was able to estimate the actual base excess at birth with an arbitrarily chosen acceptable tolerance of ±2 mmol/l. The base deficit of the extracellular fluid from artery cord samples at birth was considered as the reference standard, against which estimated values were compared.
To assess the intra-observer reproducibility of the algorithm in use, 3 months after the initial assessment the student midwife randomly selected 60 FHR tracings among those already evaluated, rearranged them in a random fashion and asked both observers to blindly repeat their readings. Statistical analysis was performed with graphpad prism 5.00 for windows (GraphPad Software, San Diego, CA, USA). P < 0.05 was considered to be statistically significant.
The differences between the base excess calculated by the observers and the actual extracellular fluid base excess measured from the umbilical artery (y-axis) were plotted against the actual base excess value (x-axis) in a Bland and Altman plot, as recommended by Krouwer.17 The bias and 95% limits of agreement for the Bland and Altman comparison were calculated. Analysis of inter-observer and intra-observer agreement was performed using the Pearson product moment correlation coefficient (with 95% confidence intervals). Colton’s benchmarks were used to interpret the correlation coefficient as follows: 0–0.24 indicates little or no relationship; 0.25–0.49 indicates a fair degree of relationship; 0.50–0.74 indicates a moderate to good relationship; and 0.75–1.00 indicates a very good to excellent relationship.
A total of 484 women were delivered at our institution between January and April 2010. One hundred and fifty-two women fulfilled the inclusion criteria and had continuous FHR monitoring available for the entire length of their labour. Continuous FHR monitoring was used in 54 women with uncomplicated pregnancy because one-to-one midwifery care could not be guaranteed, and in 98 women for the following reasons (in cases of multiple reasons, only the main indication is reported): oxytocin administration for induction of labour (n = 64); meconium-stained fluid (n = 21); small-for-gestational-age fetuses (n = 7); and hypertensive disorders of pregnancy (n = 6).
The demographic and clinical characteristics of the 152 women whose FHR tracings were analysed are shown in Table 1. Uncomplicated vaginal delivery occurred in 127 women (83.6%), vacuum delivery occurred in seven women (4.6%), and caesarean delivery occurred in 18 women (11.8%).
Table 1. Demographic and clinical characteristics of patients
Women (n = 152)
Data are expressed as medians (ranges) or absolute numbers (%).
Gestational age (weeks)
Maternal age (years)
Length of first stage of labour (minutes)
Length of second stage of labour (minutes)
Fetal weight (grams)
Five-minute Apgar score
Umbilical artery pH
Umbilical vein pH
Umbilical artery pH < 7.20
Umbilical artery pH < 7.10
Umbilical artery pH < 7.00
Base excess (mmol/l)
−5.2 (–17.3 to 0.6)
Base excess <−8 mmol/l
Base excess <−10 mmol/l
Admissions to neonatal intensive care
The first and second observers correctly predicted the umbilical artery base excess at birth using the algorithm by Ross and Gala for 111 (73.1%) and 116 (76.3%) cases, respectively. No significant differences in predicting the level of fetal acidemia at birth were observed when stratifying for mode of delivery [the two observers correctly predicted the base excess values in 13/18 (72.2%) and 14/18 (77.8%) cases of caesarean sections, respectively; P = 1.00], nor for the following subgroups: uncomplicated pregnancies [correct base-excess assessment in 43/54 (79.6%) versus 44/54 (81.5%) cases, respectively; P = 1.00]; small-for-gestational-age fetuses (71% accuracy for both operators); induction of labour for hypertensive disorders of pregnancy [4/6 (66%) versus 5/6 (83%) cases with correct prediction for assessors 1 and 2, respectively; P = 1.00); induction of labour for other indications [59/85 (68.2%) versus 62/85 (72.9%) cases of correct estimation for assessors 1 and 2, respectively; P = 0.74].
When the evaluations of the two assessors were considered jointly (n = 304), a total of 77 (25.3%) incorrect estimates of umbilical artery base excess at birth were identified. Among these 77 cases, an overestimation of base excess (interpolated base excess value lower than the actual value: for example, calculated base excess of −8 mmol/l and actual base excess of −5 mmol/l) was observed in 53 (68.8%) cases, and an underestimation of base excess (i.e. interpolated value higher than the actual value) was observed in 24 (31.2%) cases. Overall, the assessors underestimated the real level of acidosis in 24 cases out of the total number of 304 assessments (7.8%). When an underestimation of acidosis occurred, the median error (difference between the interpolated base excess and the real base excess) was −3.2 mmol/l (range from −2.1 to −7.6; interquartile range from −2.9 to −4.3). In eight (33.3%) of these 24 cases the length of labour was ≤1 hour, and in five (20.8%) cases a small-for-gestational-age fetus was delivered.
In this study group one (0.7%) newborn had a base excess of <−12 mmol/l at birth, and 15 (9.9%) had a base excess of <−8 mmol/l. In the subgroup of fetuses with a base excess of <−8 mmol/l, the first and second observer correctly predicted the actual base excess in 11 (73.3%) and 13 (86.7%) cases, respectively. Even when the base excess was not correctly predicted (with an error >±2 mmol/l), the two assessors correctly identified a status of at least moderate acidosis (i.e. interpolated base excess <−8 mmol/l) in 13 out of 15 (86.7%) and 14 out of 15 (93.3%) cases, respectively.
The inter-observer reliability of the algorithm proposed by Ross and Gala in the prediction of base excess at birth was very good (Pearson correlation coefficient = 0.75; 95% CI = 0.71–0.83). When 60 randomly selected FHR tracings were blindly resubmitted to the two assessors 3 months later, the estimated intra-observer reproducibility was excellent for both observers (Pearson correlation coefficient = 0.80 and 0.82; 95% CI = 0.74–0.84 and 0.75–0.87; for the first and the second assessor, respectively).
Fifty-eight (0.42%) newborns with umbilical artery pH < 7.00 at birth were identified out of 13 715 deliveries in the period between 2003 and 2010. Of these, 28 (48.3%) were not included for the following reasons: in 12 (20.6%) cases caesarean section was performed before the onset of labour, and in 16 (27.6%) cases FHR monitoring was not continuously available during the entire length of labour. A total of 30 FHR tracings from fetuses with acidosis at birth were included for the final analysis, together with an equal number of traces of newborns with umbilical artery pH ≥ 7.00. The characteristics of the cases included in this second part of the study are shown in Table 2. With an approximation of ±2 mmol/l, the first and second operators correctly predicted base excess values at birth in 23 (76.7%) and 21 (70%) newborns with pH < 7.00, respectively, and in 21 (70%) and 22 (73.3%) newborns with pH ≥ 7.00, respectively. The first and second operators correctly identified the presence of at least moderate acidosis (i.e. base excess <−8 mmol/l) in 29 (96.7%) and 29 (96.7%) cases, respectively, and the presence of a severe grade of acidosis in 18 (81.8%) and 17 (77.2%) out of the 22 newborns with umbilical artery base excess <−12 mmol/l, respectively. The inter-observer reproducibility in the setting of fetal acidosis (umbilical artery pH < 7.00) was good (Pearson correlation coefficient = 0.72; 95% CI = 0.54–0.86).
Table 2. Characteristics of cases with pH <7.00 (i.e. all women who delivered an infant with an umbilical artery pH <7.00 at birth) and of controls (i.e. an equivalent number of randomly selected women with an umbilical artery pH ≥ 7.00)
pH <7.00 (n = 30)
pH ≥7.00 (n = 30)
Data are expressed as median (range) or absolute number (%).
Gestational age (weeks)
Maternal age (years)
Length of labour (minutes)
Length of second stage of labour (minutes)
Fetal weight (grams)
Five-minute Apgar score
Base excess (mmol/l)
−13.1 (−10 to −24)
−4.1 (−0.5 to −9.5)
Intrauterine growth restriction
Induction of labour
Caesarean section for non-reassuring FHR
Admissions to neonatal intensive care
When considering only fetuses with a base excess at birth of <−8 mmol/l (n = 45) from both groups in this study (1 and 2), the two observers correctly predicted the base excess (±2 mmol/l) on the same subject in a total of 30 cases (67%).
Figure 2 describes the level of disagreement between base excess estimates and actual measurements according to the magnitude of the actual base excess value at birth in the entire study cohort (groups 1 and 2).
The present validation study shows that the use of Ross and Gala’s algorithm for the interpretation of FHR tracings during labour has an accuracy of ≥70% in predicting umbilical artery base excess at birth (with a tolerance of ±2 mmol/l) both in a low-risk obstetrical population (group 1) and in the setting of fetal acidemia (group 2). Moreover, the proposed model for the calculation of fetal base excess showed a high inter-observer reproducibility as well as excellent intra-observer reliability.
The criteria published by Ross in 200212 were developed, analysing short segments of FHR tracings in the imminence of delivery or using ovine models of fetal hypoxaemia. The novelty of the algorithm lies in its attempt to objectively quantify the loss of buffer base during labour, according to different FHR trace abnormalities and their relative duration. As a preliminary test of the potential use of this algorithm in clinical practice we sought to retrospectively evaluate the application of these criteria throughout the entire duration of labour, with the aim of predicting the level of fetal acidemia at birth.
Umbilical artery base excess is the most direct measure of fetal metabolic acidosis, and threshold levels of base excess (e.g. −12 mmol/l) have been associated with an increased risk of neonatal central neurological injury, as well as adverse respiratory outcomes.19,20 Therefore, the possibility of extrapolating a rate of base excess decrease and a time at which a threshold level is reached may aid in prospective case treatment, and in the retrospective interpretation of FHR tracings, particularly in the setting of medicolegal liability. Validated tools for the determination of base excess during labour may assist the obstetrician in evaluating the margin of safety before obstetric intervention to deliver the fetus as soon as possible, in future clinical practice. Moreover, modalities to time hypoxic/ischaemic injury (i.e. to identify the time at which a threshold base excess level has been reached) may be of value in determining the opportunities and/or responsibilities for the prevention of adverse neonatal outcomes when cases are reviewed retrospectively in litigation or clinical audit.
In the obstetric literature there has been little examination of the rate of base excess loss in the presence of particular FHR patterns, and few attempts to deliver recommendations regarding the speed of clinical reactions to certain FHR abnormalities to minimise fetal acidemia. A case–control study published by Low et al. in 1999 revealed that the absence of variability and/or the presence of late prolonged deceleration ‘for at least 1 hour’ are more frequently associated with a base excess of <−12 mmol/l at birth. However, the authors were not able to predict the actual level of acidemia, and less than one in five fetuses exhibiting these FHR patterns had a base excess of <−12 mmol/l at birth.20
Although we acknowledge that the availability of results of fetal scalp blood sampling would have provided valuable information on the performance of the algorithm, a real-time determination of fetal base excess changes throughout labour is impractical because of limitations of repeated fetal scalp blood sampling. Moreover, often base deficit cannot be obtained from scalp blood because of the large sample needed and the more sophisticated equipment required for this measurement, compared with the determination of pH.21 Lastly, obvious interventions that are prompted by the suspicion of fetal compromise prevent investigators from gathering important information regarding fetal loss of buffer base in response to hypoxemia. Further ancillary testing has recently been proposed for patterns in which it is believed that the risk of acidemia is uncertain, such as fetal electrocardiogram analysis.7–22 While awaiting a definite agreement on the acceptability of these ancillary techniques, and their widespread introduction, we believe that research on simple tools for the standardised management of FHR patterns during labour is crucial.
Another interesting aspect of our study emerges in the graphical representation of the values calculated by the two assessors plotted against the actual base excess values at birth (Bland and Altman plot; Figure 2): a visual interpretation of the graph suggests a tendency towards an overestimation of the level of acidemia for the less acidotic fetuses, and an underestimation for the more acidotic ones. Nevertheless, it is of key importance to underline that in the population of fetuses delivered with pH < 7 (group 2), the ability to discover a state of at least moderate acidemia (i.e. a base excess of <−8 mmol/l) was extremely high for both the assesors involved (96.7%). This means that the algorithm appears to perform well in correctly identifying fetuses that would benefit from prompt delivery.
Our findings of inter-observer reliability of the algorithm proposed by Ross and Gala appear encouraging, particularly if compared with the available literature on the interpretation of FHR tracings. In 2008 Chauhan et al.23 evaluated inter-clinician agreement for the interpretation of cardiotocographic tracings during labour: the authors report a k-correlation coefficient of −0.12 and 0.15 for the classification as reassuring or non-reassuring tracings in early labour and imminence of delivery, respectively. An acceptable inter-observer agreement was only found in the recognition of recurrent severe variable decelerations and tachycardia, with a k-coefficient in the imminence of delivery of 0.43 and 0.59, respectively.
It should be acknowledged that the proposed model for the calculation of fetal base excess relies on a number of simplifying clinical and physiologic assumptions that have not always been widely and rigorously tested on humans, including the starting base excess values. For instance, the algorithm assumes that during a prolonged latent phase of labour the fetus undergoes no real change in base deficit. Similarly, the model assumes that a post-term fetus or one with intrauterine growth restriction begins labour with the same level of acidosis as an appropriately grown term fetus. Moreover, no correction is made for the presence of tachycardia or decreased/absent variability of FHR. Finally, the algorithm does not take into account base excess changes in the presence of common obstetrics intervention such as the induction or augmentation of labour.
We recognise that the retrospective nature of the present study represents a possible source of bias. Unquestionably, the assessment of FHR tracings is less challenging when it is removed from the anxiety and emotional stress that the labour ward often fuels.24,25 Moreover, the wide use of intermittent auscultation and the consequent unavailability of continuous FHR monitoring in all cases prevented us to generalise our results to every unselected labouring woman. A prospective validation is therefore crucial before firm conclusions on the proposed algorithm can be made. In particular, it is extremely important to explore the false-positive and false-negative rates, in order to verify how many unjustified caesarean sections for suspected (but not confirmed) acidemia would be performed, as well as how many obstetrical interventions would be accomplished too late, possibly comparing the performance of the algorithm with the simple visual interpretation of tracings. Hopefully, with an algorithm that is simple to apply, and that may help in recognising the magnitude of fetal acidosis, obstetricians will be best prepared to manage labour in a standardised fashion for an optimal outcome.
It should be obvious that until it is fully validated, we do not encourage the implementation of the algorithm in everyday clinical practice. Further studies on larger populations should be performed to verify whether the algorithm maintains accuracy and reliability in a real delivery room setting.
In conclusion, the present study suggests that the algorithm proposed by Ross and Gala may be a valuable tool in estimating the level of fetal base excess throughout labour, with high accuracy and promising inter- and intra-observer reproducibility. This is a preliminary study, and the algorithm will ultimately need to be subjected to appropriate testing in larger, prospective cohorts before being implemented in the decision-making processes of the labour ward.
Disclosure of interests
The authors declare that there are no conflicts of interest.
Contribution to authorship
SU and AC collected the electronic FHR monitoring tracings. SU and GFC performed the analysis of the electronic FHR monitoring tracings. SU, AC, GFC, GB, MA, JC, and FG contributed to the conception and design of the study. SU, AC, and FG performed statistical analysis and interpreted the results obtained. SU, AC, GFC, and FG wrote the article. SU, AC, MA, and FG critically revised the article.
Details of ethics approval
Retrospective analysis of our obstetrical database for research purposes was approved by the Ethical Committee of our hospital on 17 June 2008. The present research received further approval by our Institutional Review Board on 15 January 2010.
No funding sources supported this investigation.
We are grateful to Viola Rusca and Matteo Vidoni, who helped with the collection of all the FHR tracings.