Optimal Time Intervals for Vaginal Breech Births: A Multi-Site Case-Control Study

Overview

This study uses a case-control design to test the hypotheses that avoidable delay in late second stage and premature cord clamping are associated with admission to the neonatal unit and/or early perinatal death following vaginal breech births. We aim to determine the predictive value of: 1) adherence/non-adherence to the Physiological Breech Birth Algorithm; and 2) premature cord clamping (\<1 minute following birth) for admission to the neonatal unit and/or perinatal death following vaginal breech births. The secondary objectives are to: 1) test all variables for a single-factor association with the primary outcome; and 2) test the predictive values of associated variables using linear regression; in order to 3) explore other factors contributing to adverse outcomes in vaginal breech births.

This study replicates a previous successful pilot study (Spillane et al 2021). The sample size calculation for the original study was based on the hypothesis that among those breech births where a neonatal death or admission to the neonatal intensive care unit (NICU) occurs following the birth, the time between the birth of the fetal pelvis and the birth of the aftercoming head will be greater than three minutes more often than it is among the controls, where no death or NICU admission has occurred. The hypothesis was developed from what is already known from previous research by Reitter, Halliday and Walker (2020). The study found that the time between the birth of the pelvis and the birth of the aftercoming head is more than three minutes in only 25% per cent of breech births with good outcomes. Spillane et al's study hypothesised that this interval would be more than three minutes in 75% of births where death or a NICU admission occurred. Spillane et al's power calculation determined that a sample size of fifteen cases and thirty controls would be required to infer an association between a pelvis to head interval of \>3 minutes and the composite neonatal outcome (death or NICU admission), with a confidence interval of 95% and a power of 80%. The results of that study confirmed that association (p=\<.0005). We have not re-calculated sample size. Our aims in this study are to confirm the results of the original study by replication in multiple different settings and to explore additional confounding variables that may only be apparent in larger data sets. We are particularly interested in the influence of immediate cord clamping on these outcomes, but it was not possible to calculate a sample size based on the original study due to none of the cases having anything other than immediate cord clamping (n=0). Therefore, we are seeking a larger sample size in the hopes of being able to identify an appropriate sample size for future research, and to confirm the results of the previous study. All anonymised data gathered in each site will be combined and analysed as a single data set by the Co-Investigators. The complete anonymised data set, combining data from all sites, will be downloaded and stored within the KCL Sharepoint for analysis. We will first calculate the time to event interval for all variables of interest and report descriptive statistics for all variables, including means, medians and range for continuous variables. Exposures and confounders will also be converted into binary variables, reflecting the cut-offs used in the Physiological Breech Birth Algorithm. These will then be tested against the primary outcome using the non-parametric chi-square, or Fisher's Exact tests where cell frequencies are too small for the chi-square test. Linear and logistic regression analysis will be used to test the predictive values of meeting or exceeding the recommended time limits in the Physiological Breech Birth Algorithm, and of maintaining and intact umbilical cord until the onset of respiration or not. Further linear and logistic regression analyses will be conducted with all variables that show 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.