Gone are the days when the quality of our craft was judged solely at the moment a crying baby was placed in her mother's arms. The modern obstetrician is now charged with considering how each clinical decision may impact the reproductive health of our patients and communities in the long term, and nowhere is this more evident than in the recent campaign to prevent the first caesarean delivery. Amongst the modifiable variables that clinicians are asked to reconsider as they manage labour is the duration of the second stage, which some now recommend we allow to proceed for up to 4 hours in nulliparous women with epidurals, as long as descent or rotation have occurred (Spong et al. Obstet Gynecol 2012;120:1181–93). Whereas quality evidence suggests that 80–99% of women with a prolonged second stage will achieve a vaginal delivery (Laughon et al. Obstet Gynecol 2014;124:57–67), the question remains: at what cost? Although other researchers have measured the association between prolonged second stage and adverse neonatal or intrapartum maternal outcomes, Stephansson and colleagues provide answers to a missing piece: the association between prolonged second stage and adverse maternal outcomes in the postpartum setting. The authors levy the strength of a robust Swedish database in which clinical data, including the partograph, are collected prospectively from all maternity units in the region and forwarded to a central database on a daily basis. From a study population of 72 593 women they identify increased risks of any postpartum complication (infection, urinary retention, haematoma, or ruptured sutures), which increase for most strata with every additional hour of the second stage. As with any observational study, we must consider the ways in which bias, often entirely unintentional, may impact the reported associations. Is selection bias, information bias, or confounding present? If not, could the results arise from chance (Schulz et al. The Lancet Handbook of Essential Concepts in Clinical Research, 2006)? With this work, there is little risk of selection or information bias, as deliveries are recorded prospectively and prior to study design, and the only excluded cases are those in which partograph data are not available. There is however concern for residual confounding, as the authors themselves note. For example, chorioamnionitis was not included in the multivariate analysis because of underreporting in the discharge diagnoses, as described by the authors. This is associated with a prolonged second stage, and may also increase the risk of a postpartum diagnosis of infectious morbidity. Finally, could the results arise from chance? Given the robust size of the study population, the authors are able to identity very small but statistically significant differences between most of the subgroups. When interpreting Stephansson's data in the setting of a prolonged second stage, the clinician must consider not only odds ratios but also absolute risks. These data should not be used to discourage vaginal birth after caesarean, for example, as the absolute risk of complications remains low. Stephansson and colleagues should be commended for their contribution to our understanding of maternal morbidity following a prolonged second stage. No conflicts of interest to report.
Read full abstract