Main Text:
Sir,
We respond to comments made by Zhang et al about our recent simulation study published in BJOG.1 We thank the authors for providing us with their original SAS code, and we are grateful for their response.
The central finding of our paper was that if there is an acceleration phase in the first stage of labour then it would not be detected by the statistical methods that had been used. For the polynomial regression models, missing data at earlier cervical dilatations seen in shorter labours was a significant source of bias. This is an example of missing data not at random , a crucial violation of the methodological assumptions.2
We have provided the requested graph (Figure 1), noting that as for the analyses by Zhang et al , random intercepts were not assumed. We disagree with the contention that in Figures 3B, S4, and S5 some of the underlying curves are close to the underlying labour pattern, because the acceleration phase in the underlying data was lost. In these figures, a resemblance can only be seen from 5cm cervical dilatation onwards, and then only when examinations are performed more frequently than occurs in clinical practice. We suspect that the papers published by Zhang et al , based on retrospective data, relied on cervical examinations more than one to two hours apart.3
In our opinion an accurate understanding of the range of normal patterns of human labour is essential to inform future research and, ultimately, clinical practice. We agree with Zhang et al that variation in rates of cervical dilatation in the active phase of the first stage of labour has not been clearly defined, and that multiple patterns may co-exist. We can confirm that, by design, our simulated data cm-by-cm transit times followed a log-normal distribution. We agree that it is likely that not all labours enter the active phase at 4cm but note that assuming a range of points of onsets would not improve the ability of Zhang et al ’s model to detect a true acceleration phase.
We agree that a rigid one-curve-for-all may be too simplistic and that can have important clinical consequences. For example, the ACOG / SMFM Consensus states that active phase arrest should not be diagnosed before 6cm cervical dilatation, and that a caesarean section should not be performed for a prolonged passive phase even of more than 20 hours in nulliparous women.4 As pointed out by Cohen and Friedman and others,5 this recommendation potentially leads to excessive delay in performing a caesarean section for women with arrest disorders at less than 6cm cervical dilatation which may lead to increased complications such as post-partum haemorrhage, endometritis, uterine rupture and adverse perinatal outcomes.
Despite the findings of our simulation studies, we believe that regression methods have the potential to help us understand patterns of human labour and that further research is warranted. We also note that we did not perform simulations for some analyses such as the proposed partograms by Zhang et al and cannot comment on the accuracy of these analyses.3