Given the increasing prevalence of twins and the higher rates of late-preterm and early-term fetal death, compared with singletons, the pragmatic question is: when should twins be delivered in the absence of standard maternal/fetal criteria?
This is a challenging research question. Because rates of fetal and neonatal death near full term are low, investigators must randomise enormous numbers of women to iatrogenic birth at, say, 36–37 weeks of gestation, versus expectant management. Were such a trial performed, it might only estimate the risks/benefits of planned birth for that week. Even if resources were available, covariates such as chorionicity and maternal comorbidities might be troublesome, as subgroup analyses would have diminished power. Significant limitations in the existing epidemiological data include imprecise timing of fetal death, defaulting to birth. This overestimates the rates of later stillbirth, making planned birth appear more favourable. Another problem is missing chorionicity: risks are higher in monochorionic compared with dichorionic placentation.
To overcome these deficiencies, Dr Wood and colleagues performed an analysis of a 15–year period from the Alberta Perinatal Health Project database to find 17 875 twins alive at 23 weeks of gestation, a cohort that included 202 fetal and 244 neonatal deaths (Tables S1, S2). With this sample they were able to construct a graph of the weekly risk of fetal death in undelivered gestations, observing a spike in the fetal death rate at 38 weeks of gestation (Figure S1). They also computed a weekly risk ratio of neonatal versus fetal death: a ratio of <1 favoured delivery. Although this point estimate fell below unity at 36 weeks of gestation, by 38 weeks of gestation the ratio became statistically significant (Figure S3). This effect was more pronounced earlier, in the monochorionic subgroup (Figure S2).
They further observed that small for gestational age (birthweight <5th centile) and an inter-twin discordance of >20% were significant predictors of fetal death. Note that birthweights and not sonographic estimates, which were available to clinicians, were used. Other important limitations, as noted by the authors, include missing information on pregnancy (?mis)management: e.g. was ultrasound used to detect growth restriction and discordance so that fetal surveillance could avert fetal death and prolong gestation? Perhaps the biggest limitation is the missing neonatal morbidities that would accrue from late-preterm and near-term births. This imposes ‘trade-off’ decisions: delivering late preterm averts fetal death but incurs neonatal morbidity. Thus, although the investigators performed a thorough death-metric comparison, the imponderable question is: how many cases of neonatal morbidity balance one fetal death?
These limitations notwithstanding, this is a timely analysis and makes a significant contribution to understanding the natural history and gestational age at birth of twins: they confirm that the rate of fetal death in twins is very low. The authors also correctly assert that with such low rates of fetal death, especially in dichorionic twins, clinicians and patients might not be persuaded that early delivery is necessary, in light of the known higher morbidities. Perhaps the best way to use these findings clinically is to recognise and not exceed their lower bounds (i.e. 36 weeks of gestation for monochorionic twins; 38 weeks of gestation for dichorionic twins) when considering planned twin birth without justifiable maternal/fetal indications.
Disclosure of interests
Dr Owen has no conflicts of interest to report.