Participación

The model made a number of assumptions that need to be kept in mind when interpreting the results. 1. We were able to perform the analysis only for the interventions for which we had sufficient information

on all outcomes required in the model. This does not mean the excluded interventions are not cost-effective, just that we have no evidence. Therefore, our conclusions on the cost-effectiveness of the included interventions needs to be interpreted within the set of interventions that we were able to include. However there were no interventions that were identified by the NMA as being effective that were not included in the cost-effectiveness analysis. Furthermore, only a subset (86) of the studies provided information on both VD within 24 hours and CS for the joint modelling required in the economic model. Therefore, the economic evaluation is based on fewer studies than the NMA presented inChapter 3for VD within 24 hours.

2. It is assumed that the proportion of babies who are admitted to NICU depends on mode of delivery (CS or VD), but not on whether a VD was within 24 hours of induction or not. Of those admitted to NICU, we assumed that the proportion of babies cared for in intensive (19%), high dependency (7%) or transitional care (74%) would not vary depending on method (vaginal vs. CS) or timing of delivery (<24 hours;>24 hours), or intervention.

3. It was also assumed that the length of stay in intensive, high dependency and transitional care was fixed at 2, 1.5 and 2 days, respectively, based on the data from the Liverpool Women’s Hospital. 4. It was assumed that long-term costs and benefits would be equal across induction methods, and that

any variation would be captured in the time between induction and discharge.

5. The NMA gave estimates on the rate of instrumental delivery, Apgar score<7 at 5 minutes and uterine

hyperstimulation, but these were assumed to be unnecessary in the model, as the differences in costs and benefits would be captured in the other outcomes included.

6. Some important outcomes, such as post-partum haemorrhage, were not reported as an outcome in trials and therefore could not be included in the economic model.

7. Although we would have liked to, we did not have enough evidence on parity to explore cost-effectiveness in primiparous and multiparous women separately.

Conclusions

In summary, the base-case analysis found that all of the methods of induction were cost-saving compared with no treatment. It is noteworthy that there is considerable uncertainty in our cost-effectiveness

estimates, with the majority of the interventions having very similar utility values, and mainly differing in total costs.

With this caveat, buccal/sublingual misoprostol and titrated oral misoprostol were identified as being the interventions with the highest expected net benefit and the highest probability of being cost-effective. At any willingness-to-pay value of>£23,000 per unit increase in utility, titrated low-dose oral misoprostol

solution seems to be the intervention that is most likely to be the most cost-effective for use on the UK NHS. Given that we were able to analyse only two subgroups (intact membranes and unfavourable cervix), and the number of interventions compared–and studies included–were lower than in the base case, the results of subgroup analyses should be interpreted cautiously (i.e. as hypothesis generating).

In the subgroup of women with intact membranes, and limiting to interventions feasible on the NHS, i.v. oxytocin with amniotomy was identified as being the intervention with the highest expected net benefit and the optimal intervention at any willingness-to-pay value. However, there was again a lot of uncertainty in this estimate, with buccal/sublingual misoprostol and titrated (low-dose) oral misoprostol also with a moderate probability of being most cost-effective.

Buccal/sublingual misoprostol and titrated low-dose oral misoprostol solution were found to be the interventions that were most likely to be cost-effective in women with an unfavourable cervix.

The majority of the interventions, with a few notable exceptions, such as intracervical PGE2, result in similar expected utility and vary mainly in terms of cost. There is a considerable degree of uncertainty in these estimates, demonstrated by the wide confidence intervals around the values.

There is a need to study further utilities on both mother and baby outcomes from both mother and baby perspectives. This research should be conducted using preference-based measures on large samples and with uncertainties fully reported. We would urge future trials in this area to present results according to mutually exclusive clinically relevant subgroups (e.g. parity, membrane and cervical status, previous CS) to allow more evidence to inform subgroup analyses. We would also urge trialists to report results in a format that allows the construction of the number of vaginal deliveries within 24 hours, the number CSs and the number of vaginal deliveries after 24 hours. It would also be useful to report the NICU admissions according to mode of delivery. Haemorrhage and sepsis (antibiotic usage) are also important adverse outcomes that have consequences for the economic evaluation but which are inconsistently reported. The value-of-information analysis suggests that the decision is very sensitive to uncertainty in the model inputs, and there is potential value in reducing this uncertainty through future research studies. Further large well- conducted trials may be a worthwhile use of resources, but it is essential to collect information on costs and utilities, as well as transition probabilities for mode of delivery and NICU admission.

Chapter 5

Discussion