Intrapartum oxytocin: time to focus on longer term consequences?

Despite the ubiquitous use of uterotonic agents during caesarean delivery, evidence-based recommendations are largely lacking 1. In addition, dosing of uterotonic agents, especially oxytocin, is highly variable and often subject to idiosyncratic requests from obstetric colleagues. Little guidance exists to help clinicians balance the risk of failing to prevent postpartum haemorrhage with possible adverse effects from unnecessary overdose. In this edition of Anaesthesia, Heesen et al. attempt to address this shortcoming by reviewing a broad and heterogeneous body of literature to provide sensible recommendations for safe and effective administration of the more commonly used uterotonics. We appreciate that these recommendations are an invaluable starting point, especially in institutions where uterotonic administration is unregulated. A few issues, which the authors of the consensus statement allude to, prevent us from a wholehearted endorsement of these recommendations. Our main concern is the quality of the literature upon which these recommendations are based. There is a notable degree of inconsistency between these studies in both the assessed interventions and the measured outcomes. This is exemplified by the different doses, routes and rates of administration of uterotonic agents in the seven studies included in a recent systematic review 2 of randomised trials that compared misoprostol with oxytocin during caesarean delivery (Table 1). Such heterogeneity will undoubtedly be reflected in the estimates of effect size and presents a significant threat to the validity of the synthesis of any such data. Similarly, one would expect bleeding-related outcomes to be of interest in studies investigating uterotonic agents. However, this is not always the case. For example, Su et al. 3 included 11 studies in their Cochrane review of the efficacy of oxytocin and carbetocin. Perhaps worryingly, only two of these studies had a primary outcome related to bleeding. Acharya et al. 23 UK Lokugamage et al. 24 UK Vimala et al. 25 India Lapaire et al. 26 Switzerland 5 IU bolus + 20 IU infusion at 2.5 IU.h−1 Oxytocin 5 IU bolus + 800 μg (oral) Eftekhari et al. 27 Iran 20 IU infusion at 12 IU.h−1 until uterus contracted Chaudhuri et al. 28 India Owonikoko et al. 29 Nigeria 20 IU infusion Rate not specified The second concern is the interpretation of data produced by dose-finding studies, especially those employing the up-and-down sequential allocation methodology. A sophisticated understanding of this methodology is warranted if its limitations are to be appreciated and appropriate inferences drawn. Up-and-down sequential allocation methodology was originally applied to phase-1 toxicity trials where, for ethical reasons, it was imperative to provide an estimate of toxic dose (LD50) while minimising the number of subjects exposed to potentially toxic doses. Their design requires two to three times fewer participants than traditional, non-sequential designs in order to provide estimates but this efficiency comes at the expense of increased risk of bias and reduced precision 4, 5. The biased coin design is an extension of the up-and-down sequential allocation design that provides estimates with established accuracy at different percentiles along the dose–response curve. It achieves this by biasing the response to treatment ‘success’ in favour of selecting the same, rather than a decreased, dose in the next participant. On the other hand, in the context of the binary outcome (adequate vs. inadequate uterine tone) consistently employed in uterotonic literature, the response to inadequate uterine tone is to increase the dose in the subsequent participant. By exerting a 9:1 bias for example, investigators can hone in on the ED90. However, biostatisticians urge some caution 6 in the inferences made on the basis of such studies, particularly those fitting the data to logistic regression models considering the non-independence of the sequentially allocated doses. They advise that such estimators are prone to bias and may produce more precise estimates (narrower confidence intervals) than justified by the data. There are also important issues regarding sample size, starting dose and incremental differences in dose allocation that have subtle influences on the information gathered, especially with respect to inter-individual variability. With consideration of these caveats, studies using this design provide valuable information and are an improvement on the ED50 studies of the past. Balki et al. and Carvalho et al. used a 9:1 biased coin design to estimate the ED90 (95%CI) bolus dose of oxytocin needed for adequate uterine contraction at caesarean delivery for labouring and non-labouring women: 2.99 (2.32–3.67) IU and 0.35 (0.18–0.52) IU, respectively 7, 8. To illustrate how best to interpret these data, let us focus on the latter of these two estimates. It might not be immediately obvious, but it is not the case that 90% of those who received a dose of 0.35 IU in this study had an adequate uterine response. In fact only, three different doses were studied and none of the participants received 0.35 IU. Of their 40 patients, 3, 31 and 6 received 0 IU, 0.5 IU and 1.0 IU, respectively. A more accurate description, therefore, would be that the authors fitted a logistic regression model to the data in order to estimate (with 95% confidence) that 90% of their study population would have had an adequate uterine response to a dose of 0.35 (± 0.17) IU. Whether this would be the same for the women that are scheduled for a caesarean delivery in your hospital depends to a large extent on how well they match the small, and strictly defined, population in Carvalho et al.'s study. Due to study restrictions on the weight, age, ethnicity etc., it may be reasonable to presume that they underestimate the ED90 relevant to your patients. Such a conservative approach is further supported by the fact that the data may justify wider confidence intervals than provided by the logistic regression. So, if you aspire for a high confidence (95%) that the dose you give your patient population exceeds their ED90, you might be advised, at the very least, to give a dose at the upper limit of their confidence intervals (0.52 IU). Given the established safety of this, and even twice this dose, a recommendation of 1 IU of oxytocin as an initial bolus in elective caesarean delivery seems justified. The same logic can be applied to both the choice of bolus in intrapartum caesarean deliveries and the accompanying recommendations for maintenance infusion. We could argue that their choice of a 3 IU bolus is low and that maybe a bolus of 4 IU is a more reasonable choice (95%CI: 2.32–3.67 IU). However, their advice to administer a repeat dose (of 3 IU) as early as two minutes after delivery if uterine tone is inadequate, would probably result in many patients receiving 6 IU (often unnecessarily) in the first three minutes, rendering this a relatively moot point. They do not provide evidence to support their choice of a 2-min assessment and we could only find three studies from the dose-finding literature 9-11 that provide data on assessments this early after oxytocin administration and none strongly support it. The expert consensus statement advises lower ‘maintenance’ infusion rates after an initial bolus: 2.5–7.5 IU.h−1 and 7.5–15 IU.h−1, in elective and intrapartum caesarean delivery, respectively. These may seem much lower than the dose ranges supported by studies by George et al. and Lavoie et al., who found the ED90 (95%CI) in non-labouring parturients to be 17.4 (9–25.8) IU.h−1 and 16.2 (13.1–19.3) IU.h−1, respectively 12, 13, whereas the latter study also estimated this parameter as 44.2 (33.8–55.6) IU.h−1 in their labouring cohort. However, these studies did not use a bolus dose. Therefore, although the specific ranges chosen by the expert consensus are not guided by strong evidence, their recommendation of a bolus followed by a lower rate of infusion seems a sensible approach. A third consideration is the wide variability in the use of uterine massage, time to closure of hysterotomy, uterine exteriorisation and the subjective nature of uterine tone assessment, and the lack of data from high body mass index patients. These important confounders are highly likely to influence the success of the chosen oxytocin regimen. Therefore, we recommend flexibility when implementing these recommendations to account for these variables. The elephant in the (delivery) room, however, is the lack of true population pharmacokinetic data to inform oxytocin dosing. Although plasma level of oxytocin is poorly correlated with uterine contractility during augmentation of labour 14, we posit that this may not be applicable in the setting of haemorrhage prophylaxis. The doses required to promote tetanic contraction after caesarean delivery are four- to five-fold higher than those used to facilitate the cyclical uterine contraction required to augment labour. Where uterine tetany is desired, we agree with the consensus recommendation that an oxytocin bolus followed by infusion is the best approach because it makes pharmacokinetic sense to ‘induce’ uterine contractility with a bolus and subsequently maintain it with an infusion. For the bolus dose, it would be, at the very least, useful to know which scalar is appropriate to modify doses in patients at the extremes of height and weight, for instance ideal/adjusted body weight. For these reasons, a sample-intensive pharmacokinetic study is desperately needed. However, the practical obstacles to obtaining frequent plasma levels around the time of oxytocin administration at caesarean delivery, and some technical aspects of the relevant assays 15, mean these data will be challenging to obtain. If there are significant differences in pharmacokinetics between normal weight and obese subjects, it would have major implications for the selection of oxytocin dose for the prophylaxis and treatment of postpartum haemorrhage. This could then be further refined using pharmacodynamic data obtained with a standardised and objective measure of uterine tone. Finally, we advise a holistic and philosophical approach regarding oxytocin use. Although we are more familiar with its role in labour and delivery, we want the readers to be cognisant of the fact that oxytocin, a potent neuropeptide, is responsible for shaping fundamental human traits such as sociality, compassion, empathy, love and trust 16-19. The sheer scale and number of these recent discoveries in neuroscience clearly suggest that a radical shift is required on how oxytocin is viewed – as a naturally occurring hormone of profound importance to human behaviour, and not solely as a pharmaceutical agent. Administering supra-physiological doses of oxytocin to prevent postpartum haemorrhage may have unintended consequences for the mother–infant dyad. For example, there is a modest inverse association between oxytocin exposure and breastfeeding success 20, 21, possibly due to downregulation of oxytocinergic signalling mechanisms. Another area where oxytocin might be increasingly relevant is postpartum depression; intrapartum exposure to oxytocin increased the risk of depressive or anxiety disorder by over 30% at 1 year postpartum, regardless of pre-pregnancy psychiatric diagnoses 22. Clearly, there is a pressing need for clinical studies in these domains which may ultimately have repercussions for how much oxytocin we administer. Therefore, in this relentless pursuit of uterotonic efficacy with oxytocin, it is quite possible that we might be missing the bigger picture. In summary, we believe that the consensus statement is an excellent trigger to help recalibrate clinical practice and generate discussion with obstetric colleagues. The least possible oxytocin dose, guided by good quality pharmacokinetic/dynamic data, should be the ultimate goal. Proper stewardship of this potent neuropeptide will be paramount in the future, as we uncover more information on its effects on maternal-neonatal well-being. AP is an International Advisory Panel member of Anaesthesia. No external funding or other competing interests declared.