The trial recruited and randomised women with an uncomplicated twin pregnancy at 36+6 weeks of gestation, with all women enrolled having ultrasound confirmation of gestational age before 20 weeks of gestation. Although access to first-trimester ultrasound in the setting of a multiple pregnancy was commonplace among the centres participating in the trial, we did not specifically collect this information. Similarly, our trial protocol was pragmatic in that it did not mandate fetal monitoring regimens or their frequency, which were known to vary considerably across the participating centres. However, as indicated in our manuscript (p.965), eligibility was determined by individual clinicians, with participating women having no contraindication to continuation of the pregnancy.
At the time of the inception of our trial protocol, there was less clinical variation in the management of a woman with a twin pregnancy based on pregnancy chorionicity. As reported in our manuscript (p.968), we conducted a pre-specified analysis, which did not indicate statistically significant differences in the risk of infant health outcomes by treatment group between monochorionic and dichorionic twin pregnancies. In our discussion, we acknowledge the relative lack of statistical power of our trial to reliably detect differences in timing of birth between late mean gestational ages of 37 and 38 weeks.
Our pre-specified composite outcome included measures reflective of ‘adverse’ neonatal outcome, including being born small for gestational age, in recognition of the strong association between being small for gestational age (including both growth restriction and low birthweight) and the impact on future health, which as indicated in our manuscript are being increasingly reported in twin infants. As we highlight, to detect a 66% reduction in adverse outcome at term as suggested by our data, would require a sample size of approximately 1100 women with an uncomplicated twin pregnancy at term. We acknowledge the ongoing debate in relation to the use of composite outcomes in clinical trials; reflecting, as described by Ross,4 an inevitable compromise between what is ideal and what may be achievable. Obviously, while improving perinatal mortality and reducing rarer causes of term morbidity may be considered the ideal goal, the necessary sample size in excess of 20 000 women with an uncomplicated twin pregnancy at or beyond 37 weeks of gestation was not feasible given the limited resources that were available.
As our conclusions state, elective birth at 37 weeks of gestation was associated with a significant reduction in the risk of birthweight below the third centile, with no identified increase in the risks associated with early birth for either women or their infants. We have indicated the limitations to our trial, and leave it to our peers to judge whether in the first instance these findings are sufficiently robust to be used to recommend changes to their clinical practice. If not, is there sufficient clinical equipoise and commitment to conduct the large-scale, international randomised trial that would be required?