Optimal time intervals for vaginal breech births: a case-control study

Introduction

Whilst vaginal breech birth accounts for only 0.3% of all births in the UK¹, it is overrepresented in cerebral palsy litigation costs, accounting for 12% of all claims². Awareness of this increased risk has led to a reliance on caesarean section (CS). Although occurring in only 3–4% of pregnancies, breech presentation is one of the leading causes for a first-time planned CS and associated risks in subsequent pregnancies^3,4. However, a policy of universal 36-week ultrasound scans does not eliminate undiagnosed term breech presentation⁵. As a majority of compensation claims occur following diagnosis late in labour², improving outcomes in these rarely occurring births to reduce the litigation burden remains an important concern. Additionally, some women wish to plan a vaginal breech birth and encounter reluctance from health care professionals who fear they cannot keep the birth safe^6,7.

Studies of vaginal breech birth outcomes frequently seek to identify risk factors associated with the mother or fetus that may increase the risk of a poor outcome, and guidelines often present a set of criteria which should be met before women are offered the option of a vaginal breech birth³. However, even in studies with very large samples, clear associations between commonly accepted risk factors and adverse outcomes are not reliably demonstrated, with the consistent exception of birthweight <10^th centile^8–10.

Although some of the increased risk is explained by underlying conditions which can cause the foetus to present in the breech position, the skill of the practitioner facilitating the vaginal breech birth is understood to have a significant effect on its safety^3,10,11. The components of what constitutes skilled practice, how these are developed and whether they might be modifiable to improve outcomes are less well understood. One of these components is thought to be an understanding of the mechanisms and physiology of a normal breech birth¹². Familiarity with these mechanisms underpins an ability to anticipate and avert complications, a marker of breech experience¹². Careful, evidence-based descriptions of what is ‘normal’ in breech births may therefore help more novice clinicians to anticipate and avert difficulty despite their lack of clinical experience.

The available guidance on timings in late second stage (emergence) in vaginal breech births is inconsistent and largely based on professional opinion. The Royal College of Obstetricians and Gynaecologists (RCOG) 2017 guideline on the Management of Breech Presentation suggests that “intervention to expedite breech birth is required … if there is a delay of more than 5 minutes from delivery of the buttocks to the head, or of more than 3 minutes from the umbilicus to the head”^3(p17). The K2MS Perinatal Training Programme states, “The expected time interval between the birth of the baby’s bottom until the shoulders appear should be approximately 2 minutes.”¹³ A 2009 textbook, Training in Obstetrics and Gynaecology: the essential curriculum, states: “The rule of ‘5’ has traditionally been used – 5 min for each of the three delivery stages: to umbilicus, rest of the body and shoulders, head.”^14(p278)

Recent ability to analyse videos has created an opportunity to base such guidelines on evidence rather than assumption or tradition. Reitter et al.’s recent analysis of a cohort of upright (kneeling) vaginal breech birth videos with good outcomes identified that in over 75% of births, the time interval between birth of the pelvis and head was under 3 minutes¹⁵. However, it is not known whether intervals differ in cases of adverse outcomes, and if so, whether these delays are associated with unpreventable entrapment or avoidable delay in intervention. Further evidence is needed to develop robust guidance.

Based on Reitter et al.’s 2020 analysis, a Physiological Breech Birth Algorithm was developed by Dr Shawn Walker and refined with feedback from professionals attending vaginal breech birth training. The most recent version is presented in Figure 1. Our study aimed to test the ability of the Algorithm to predict neonatal death or intensive care unit (NICU) admission among a retrospective sample of births in a London teaching hospital, based on whether the birth conformed to the guideline time frames in the Algorithm or not. We hoped to further verify or refine the Algorithm for use in a clinical trial.

Figure 1. Physiological Breech Birth Algorithm.

Designed by Shawn Walker RM PhD, June 2022 version. First published in Reitter A, Halliday A, Walker S, 2020, Practical insight into upright breech birth from birth videos: a structured analysis, Birth, 47(2):211–219.

Methods

A single-centre retrospective case control study was conducted. The protocol defined cases as births where neonatal deaths or NICU admissions occurred (primary outcome). Controls were identified as the two vaginal breech births involving no neonatal death or NICU admission, occurring directly prior to the identified case. Two previous births were used to prevent bias on the understanding that an adverse outcome can affect clinical decision-making for subsequent births¹⁶. Any NICU admission was included because this indicates a neonate which requires additional observation, tests and/or intervention. Neonates who are not admitted are deemed as generally well¹⁷. Neonatal admissions are very costly, and reducing avoidable admissions at term has been recognised as a priority¹⁸. Additionally, separation from the baby was considered an important outcome by our Patient and Public Involvement (PPI) Group¹⁹, who also requested more information on the timing of cord clamping.

To calculate our sample size, based on the work of Reitter et al.,¹⁵ we hypothesised that the rate of exposure to a pelvis-to-head interval >3 minutes would be 25% among controls and 75% among cases. Using a case:control ratio of 1:2, we determined that 15 independent cases and 30 controls were required to infer an association between a pelvis-to-head interval >3 minutes and the composite neonatal outcome with a confidence interval of 95% and a power of 80%. We began seeking cases from the year 2020 and worked backward until the specified sample size was achieved.

The study was conducted within the maternity unit at a London District General Hospital which serves a large population of 176,313 people. Two thirds are of white British ethnicity and one third from Black, Asian and Minority Ethnic (BAME) backgrounds. The community the hospital serves is thought of as affluent, with good employment rates, particularly employment in high-end jobs. The hospital itself serves a wider community than the borough it is situated within and has 5000 births per year. It has a level two NICU situated within the maternity unit.

During the time period of the study (2012-2020), the hospital’s local guidance was based on current RCOG guidance^3,20. All staff received annual mandatory training in obstetric emergencies, including a brief session on vaginal breech delivery. The Algorithm was first developed in 2017 and was not in general use at the site until 2021. A physiological breech birth training day was provided at the site in January 2018, which was attended by 39 members of staff. None of the authors were employed by the Trust, until 2020. The sample reflects a standard practice environment at the time, with mixed experience levels, and some staff having exposure to physiological breech birth theory and practice.

Our sample of 15 cases was achieved within the window of 2012 and 2020. These involved NICU admissions as no neonatal deaths occurred among the sample. A total of 71 term vaginal breech births were identified from routine electronic health records during this period. From this, we selected our 30 controls as the two vaginal breech births in the sample occurring immediately prior to each case. The Medical Record Numbers were sent to the Health Records Department for the complete files to be retrieved. Data were extracted by the lead researcher from the intrapartum care records and recorded anonymously in a Microsoft Excel spreadsheet.

A structured data collection tool was developed based on Reitter et al.¹³ The data collection tool consisted of information usually recorded in the notes during a breech birth and included: lead professional, type of breech, position, epidural, fetal monitoring, meconium, what emerged first, time each part of the breech born, documented manoeuvres used, time performed and information related to the condition of the neonate at birth. Data points were included if the information was clearly documented or could be reliably extrapolated. Some examples of this include: where the pelvis and head were born in the same minute, the umbilicus can reliably be assumed to have been born in the same minute as well; classification of rumping included any of the definitions used, both buttocks visible, anus visible, +3 station. Where data points could not be reliably discerned, they were not included in our analysis, and this is reflected in the denominators reported in the tables. Sometimes this information was extracted from risk reviews conducted following adverse outcomes, which recreated a timeline of events in detail based on notes and interviews with those in attendance.

First, we calculated the time to event interval for variables of interest. We then reported descriptive statistics for all variables, including means, medians, absolute and interquartile ranges for continuous variables. Exposures and confounders were converted into binary variables, reflecting the guidance used in the Algorithm. These were then tested for association with the primary outcome using the non-parametric chi-square, or Fisher’s Exact tests where cell frequencies were too small for the chi-square test and odds ratios.

Logistic regression analysis was used to test the predictive values of meeting or exceeding the recommended time limits in the Physiological Breech Birth Algorithm. Logistic regression analyses were conducted with all variables that showed an association with the composite neonatal outcome to determine their predictive value, and additional variables to explore their potential as confounding factors for investigation in future studies. Finally, a Receiver Operating Characteristics (ROC) curve analysis was conducted to compare the sensitivity and specificity of the 7-5-3 minute time limit guidance. All statistical analyses were performed using IBM SPSS version 26.

This research was unfunded and was conducted as part of Spillane’s role as a Consultant Midwife with a remit for supporting vaginal breech births within the Trust. Spillane is also a Principal Investigator and Walker is the Chief Investigator for the OptiBreech Care Trial, an NIHR-funded feasibility study (NIHR300582, ISRCTN14521381) currently using the Physiological Breech Birth Algorithm. Two of the researchers have had a long-term involvement with the OptiBreech PPI group, who have experience of breech pregnancy and childbirth. Whilst the PPI group did not directly consider this study, their input into other aspects of the OptiBreech Project, including the prioritisation of outcomes, influenced our choices about variables of interest¹⁹.

Approval was obtained through the Health Research Authority (IRAS 294936, 21/HRA/0562) and the Trust’s Research and Development department. This was a retrospective study using data that was anonymised by a member of the clinical care team; therefore, explicit consent was neither required nor sought.

Results

The Physiological Breech Birth Algorithm reported in Reitter et al.¹⁵ proposes three key interval limits: rumping(+3 station)-to-birth within 7 mins, pelvis-to-birth within 5 mins and umbilicus-to-birth within 3 mins. Our single-factor correlation tests showed that, in each of these categories, exceeding these limits was associated with NICU admission (Table 1). When tested together in a logistic regression, the percentage accuracy (PAC) was 86.8%. The combination had a positive predictive value of 80.0% and sensitivity of 66.7%, and a negative predictive value of 85.7% and specificity of 92.3%. The most contributory factors predicting the primary outcome were an umbilicus-to-birth interval >3 minutes (aOR:9.508 [95% CI:1.390-65.046], p=0.022) and, to a lesser extent, a rumping-to-birth interval >7 minutes (aOR:6.282 [95% CI:0.940-41.990], p=0.058). The ROC curve is presented in Figure 2.

Figure 2. Predictive value of Physiological Breech Birth Algorithm’s 7-5-3 time limits.

There was a statistically significant association between a pelvis-to-head interval of >3 minutes, the interval we used to calculate our sample size, and NICU admission following the birth (p=<0.005). However, this result was highly confounded with an umbilicus-to-birth interval >3 minutes.

As expected, there was an association between use of manual interventions to assist the birth and the composite outcome of interest (p=<.005). Manoeuvres were used in 13/30 controls and 15/15 cases. The intervals between the birth of the pelvis and the first manoeuvre used to assist the arms or head were twice as long in cases (mean 5.83, median 4, range 1–19 minutes) compared to controls (mean 2.45, median 2, range 0–6 minutes) (Table 2). Where an episiotomy was performed, the interval between rumping and episiotomy was also longer in cases (mean 5.67, median 5, range 0–18 minutes) compared to controls (mean 1.75, median 1, range 0–5 minutes).

In both cases and controls, the mean and median reported times spent on manoeuvres to release the head were <1 minute. In only one case did the reported time exceed 2 minutes. In this case, 7 minutes were spent trying to release the head. However, intervals between the birth of the arms and initiation of manoeuvres to release the head were longer in cases (mean 2.42, median 2, range 1–5 minutes) than in controls (mean 1.20, median 1, range 0–2 minutes). In the case where head manoeuvres required 7 minutes, interventions were not attempted until 2 minutes after the arms were born and 13 minutes after the pelvis was born.

Similarly, in both cases and controls, the mean and median reported times spent on manoeuvres to release the arms were <1 minute. In one case, this took 3 minutes, and in one case it required 5 minutes. The intervals between the birth of the umbilicus and initiation of manoeuvres to release the arms were also longer in cases (mean 2.5, median 2, range 0–6 minutes) than in controls (mean .88, median 1, range 0–2 minutes).

In this sample, lengths of the first stage of labour were shorter for cases compared to controls. In contrast, the intervals between diagnosis of second stage of labour and birth were longer for cases compared to controls, as were the intervals between the onset of expulsive pushing and birth. Neither of these were significantly associated with NICU admission, either in single-factor analysis or logistic regression. Although 5-minute Apgar scores were lower for cases (mean 6.14, median 6) compared to controls (mean 9.76, median 10), arterial pH and base excess results did not differ.

All admissions to the neonatal unit occurred following neonatal resuscitation, and this was documented as the reason for admission. An unexpected finding was that 28/44 (64%) of neonates experienced immediate umbilical cord clamping (UCC) following their breech births (Table 1). This included 100% of the babies admitted to the neonatal unit, although in general the arterial cord blood gas results were marginally better among cases than controls (Table 2). Mean arterial pH was 7.14 among controls vs 7.21 among cases, and mean base excess (BE) was -7.11 among controls compared to -6.65 among cases.

The data that support the findings of this study are openly available on Figshare, reference number 15134427.

Discussion

Conclusion

In this research, we have confirmed our hypothesis that an interval greater than 3 minutes between the birth of the fetal pelvis and the birth of the head is associated with neonatal admission or death. We have also demonstrated that births taking longer than the maximum parameters described in the Physiological Breech Birth Algorithm are predictive of neonatal admission or death. Questions remain about how often and at what point delay is associated with severe adverse outcomes, whether earlier intervention causes more harm than it prevents, and the role of umbilical cord clamp timing in mitigating some of the effects of hypoxia in vaginal breech births. We aim to explore these questions further in larger samples and a prospective study, which is currently on-going.