4.1 Principal findings

In this cohort study, longer active first stage of labour duration was linearly associated with longer second stage of labour duration until approximately 12 h of active first stage of labour duration. After 12 h, a non-linear trend is seen, demonstrated by a plateau in the second-stage duration. In addition, longer active first stage of labour duration was associated with increased likelihood of operative vaginal delivery or caesarean delivery.

4.2 Strengths of the study

The strengths of this study include use of a large population-based database that contains detailed information on date-time for regular and painful contractions, cervical effacement, dilation, rupture of membranes and all routine longitudinal cervical examinations and interventions recorded in the partograph. This enabled us to customise the start of active labour for each woman, meaning that women identified by this hierarchy were clinically cared for in compliance with standardised guidelines for active labour. Further, the start of the exposure was independently identified and not by default identical to admission time (a commonly used method in studies on labour duration/progression) which we consider being a strength. To put it differently, by using the hierarchy we could also include women with spontaneous start admitted in early labour to the hospital in the study cohort, restricting their contributing time to active phase (i.e. excluding time between admission and start of exposure). Another strength was the use of more precise and advanced statistical methods compared with previous studies to investigate associations between first- and second-stage durations. Modelling active first-stage duration either as a categorical or a continuous variable also enhanced understanding of its relationship to second-stage duration and multivariable regression models allowed us to control for potential confounders.

4.3 Limitations of the study

Despite our efforts to identify covariates using DAG, analyses might be limited by residual confounding from unrecognised variables, and variables without accessible information, as foetal position during labour and descent and station of the foetal head (which may influence the course of labour).

Given the complex nature of labour data, measurement error could occur since length of duration is estimated based on the data available and labour is a continuous process where transition between phases are complex. Further, a sizeable proportion of the original sample was excluded due to missing key exposure; that is, active phase duration could not be measured. By design, this study did not measure the latent phase or analyse any relationships to it. Observing the natural course of labour in all women is not possible because some women enter in very advanced labour and other are removed by caesarean delivery. Avoiding selection bias is a challenge when studying labour duration, because for women with caesarean delivery the end is determined artificially and any estimation of first-stage duration or second-stage duration would be highly speculative. Consequently, women with missing timepoint for fully dilated cervix (end of exposure) were excluded from the analysis, a previously adopted method in seminal findings.1, 3 To what extend those with a first-stage caesarean delivery (n = 551, eFig2) may influence selection bias has previously been investigated elsewhere.3, 20

Women admitted to the hospital with cervical dilation of 6 cm or more were not included because the onset of active first stage occurred prior to admission. Including women with a cervical dilation of 3–5 may have biased results in favour of longer labours, assuming that those admitted with dilation of 6 or more differ in terms of labour progression/duration.3, 21-23 Thus, any estimation of their active first-stage duration would be speculation. Generalisability of these results may be limited in some populations depending on baseline demographic characteristics and obstetric settings. Finally, women allowed to have excessively long first, or second stages may be meticulously selected by experienced clinicians. Prolonged durations may appear safe in very selected women and not carry the same outcome when applied without careful selection.

4.4 Interpretation

Traditionally labour duration is divided into first and second stage. Several efforts have been made to understand how these two stages of labour influence birth outcomes.1, 4, 6, 24, 25 Labour is a continuous process, which needs be taken into consideration since duration spent at each stage jointly shape labour management decisions and subsequent maternal and neonatal birth outcomes. As we applied a non-parametrical modelling approach on the continuous duration of active first stage, we found a linear increase in second-stage duration up until active first-stage duration of 12 h. This was further shown in the detailed curves for the 80th and 90th quantiles (Figure 2). We applied multiple different approaches and they all demonstrated similar findings, suggesting that women with active first stage of labour duration exceeding approximately 12 h (90th percentile) no longer had a linear increase in second stage of labour duration. Notably, lack of association beyond 12 h applied a small group of women.

While it is logical that longer active first stage of labour duration would be associated with longer second stage of labour duration, this relationship has rarely been statistically tested and reported. Both Hellman and Prystowsky14, Nelson, McIntire and Leveno15 reported similar patterns, while their findings suggested a linear relationship without changeable trend or plateau. A few points may explain this disparity. The Hellman and Prystowsky14 study is 60 years old, with limited statistical inference. Further, obstetrical management is different today and when to intervene during second stage most likely has changed during the last 60 years. Nelson, McIntire and Leveno15 used the 95th percentile (15.6 h) as a cut-off for the exposure (prolonged duration), in comparison with our study where both exposure and outcome were analysed as continuous variables. The Nelson, McIntire and Leveno15 approach could not capture subtle changes since the model was restricted to a preconceived cut-off. Our statistical approach allows the model to fit the observed data rather than selecting data to fit a preconceived model with a cut-off. Also, cubic splines modelling helps to depict these subtle changes in the shape of the second-stage duration curve, which made the plateau visible (Figure 3).

One hypothesis for the non-linear increase found after 12 h is that women with longer active first-stage durations have higher likelihood for interventions in second stage (i.e. active management, OVD, CD). This phenomenon may reflect compliance to clinical guidelines for labour management during active labour and definitions of prolonged labour. This plateau is most likely the results of labour management norms and not a biological phenomenon. It is well established that both adherence to guidelines, providers attitudes and focus on labour progression impact use of interventions such as oxytocin augmentation and mode of delivery.7, 9, 26-30

The increased odds ratios for operative vaginal delivery or caesarean delivery in labours with longer active first stage durations may also be due to foetal and/or maternal distress. Some studies have shown an association between prolonged second-stage duration and adverse neonatal and maternal outcomes, however often without taking active first-stage duration into account.30-35 We found, consistent with others, that operative deliveries during the second stage of labour were indicated by slow labour in a majority of the cases, and the indication for foetal distress was found in approximately 41% of the operative deliveries performed during second stage (eTable 6).36, 37 Evaluating how increasing duration during first and second stage may alter the indications for operative delivery in second stage was beyond the scope of this study and needs to be further studied. Thus, previously reported data from the Stockholm Gotland Obstetric cohort show increasing risk of birth asphyxia-related complications and admission to NICU with discrete additional increments in length of second stage, not taking first-stage durations into account.32

In terms of generalisation, labour durations in this study are in accordance with seminal findings from contemporary cohorts of women.1, 3, 6, 38-40 Most recently, Blankenship, Raghuraman, Delhi, Woolfolk, Wang, Macones and Cahill27 reported 11.3 h for the 90th percentile during active first stage, including only women who reached complete cervical dilatation (n = 6823). The LaPs pragmatic trial7 reported the 95th percentile at 12.5 h during active first stage. The LaPs trial took place in Norway, with both populations and settings similar to Sweden.2 Hence, our study findings contribute to knowledge about how labour is managed in relation to both stages and duration in a contemporary cohort of primiparous women.

4.5 Future research implications

We have learned from our study that application of DAGs to broaden the understanding of the continuous labour process (duration) may be helpful since randomised clinical trials are difficult to perform within childbirth. More research is needed, and future studies need to carefully evaluate how second-stage duration and mode of delivery should be properly accounted for in studies investigating labour duration and associated birth outcomes.10, 11, 13, 18, 41