The rationale for heightened fetal surveillance in monochorionic (MC) compared to DC twins is early detection of TTTS and IUGR, to allow timely treatment in an attempt to prevent adverse perinatal outcome, principally by amnioreduction, endoscopic laser ablation of anastomoses, bipolar cord occlusion, or early delivery. Although the frequency of monitoring has not been evaluated in randomised trials for any type of twin pregnancy, it is common practice to monitor MC twins more frequently than DC twins. Despite a vigilant policy in our centre of fortnightly ultrasound surveillance for MCDA pregnancies, we observed that unexpected late fetal deaths still occurred in seemingly uncomplicated MCDA twins.

The term “uncomplicated” was used to denote pregnancies without features of TTTS on ultrasound that also had appropriate and concordant fetal growth, as well as normal growth velocity in each of two structurally normal twins. Appropriate growth was defined as an estimated fetal weight (EFW) greater than the fifth centile for gestational age [11]. The percentage EFW difference (ΔEFW) was calculated as the difference in weight, divided by the larger twin's weight, multiplied by 100, with growth defined as concordant if the ΔEFW was less than 25% [12]. Growth velocities were plotted against the charts of Chitty et al. [13–15]. The absence of TTTS in these pregnancies was denoted by normal and evenly distributed amniotic fluid volume (deepest vertical pool 2–8 cm in each sac) with concordant bladder dynamics [16]. These pregnancies also had normal umbilical artery (end diastolic frequencies present), umbilical vein (no pulsations), and/or ductus venosus (positive a wave) Doppler waveforms in each twin.

Uncomplicated MCDA pregnancies were monitored according to a standard protocol, which comprised routine first trimester nuchal translucency assessment and chorionicity determination, a detailed anomaly scan and fetal echocardiography at 20 wk, and then fortnightly scans for growth, amniotic fluid, and Doppler (umbilical artery, umbilical vein, and/or ductus venosus). Ductus venosus Doppler was added as a routine component in 1999 [16], and chorionic plate Doppler for compensatory artery-to-artery anastomosis (AAA) detection was added in 1995 [17]. Elective delivery was scheduled in otherwise uncomplicated pregnancies at between 36 and 37 wk of gestation.

We excluded pregnancies complicated by TTTS, IUGR, structural abnormalities, and twin reversed arterial perfusion, as well as high-order multiple, monoamniotic, and conjoined pregnancies. As this was a study of fetal death after viability, we also excluded pregnancies that delivered before 24 wk of gestation and those for which the pregnant woman returned to her local hospital for continued monitoring for geographical reasons. Six pregnancies with unavailable outcome data were also excluded.

The rate of fetal death in continuing uncomplicated MCDA pregnancies was derived for each 2-wk gestational block, beginning at 24 wk. It was calculated as the number of IUDs that occurred within the 2 wk following the beginning of week n divided by the number of continuing uncomplicated pregnancies at the beginning of week n. As we have previously applied to singleton pregnancies, the prospective risk of IUD was calculated as the total number of IUDs at or beyond the beginning of week n divided by the number of continuing pregnancies at or beyond the beginning of week n [18]. Prospective risk was not determined after 36 wk, as elective delivery resulted in too few continuing pregnancies beyond this time point.

To ensure that no features of IUGR or TTTS had been missed in pregnancies with an IUD, we compared antenatal indicators of these conditions at the last scan before the IUD in affected pregnancies with those in uncomplicated MCDA pregnancies unaffected by an IUD. The variables examined for IUGR were the abdominal circumference, head:abdominal circumference ratio, EFW, and ΔEFW. For TTTS, the amniotic fluid index (AFI) was used as the antenatal indicator; the deepest vertical pool of liquor in each twin's sac could not be compared, as they were not consistently recorded when the AFI was reported as normal and evenly distributed. The comparison used a 1:3 gestation-matched case versus control design (cases are pregnancies with an IUD; controls are those without). Variables in cases were from the last scan before the IUD; control scans were from the next three uncomplicated MCDA pregnancies appearing in the database for which the scan matched within 2 d of the case scan.

As the presence of an AAA reduces the risk of TTTS by a factor of four [17], and reduces severity in affected cases [19], we also compared the frequency of AAA detection by Doppler between cases and controls to determine whether functional AAAs were less common in pregnancies with an IUD; if so they would be more prone to develop TTTS than those unaffected by an IUD.

Finally, we reviewed the autopsy reports for any undetected pathology. In particular, we sought phenotypic features of TTTS such as organ hypertrophy, particularly cardiac hypertrophy in the larger twin.

From this cohort of 151, there were ten unexpected fetal deaths in seven uncomplicated MCDA pregnancies (three double deaths and four single deaths) after 24 wk, giving an overall incidence of 4.6% per pregnancy (CI 1.9%–9.9%). Deaths occurred at a median gestation of 34⁺¹ wk (range 28⁺⁰ to 36⁺³).

Table 1 shows the rate by 2-wk blocks of fetal death per pregnancy, which, with the limits of small numbers, appeared to increase after 32 wk. The prospective risk of unexpected fetal death, also detailed in Table 1, remained relatively stable between 24 and 34 wk of gestation, 1/22 (4.5%) and 1/30 (3.3%) pregnancies, respectively.

Ultrasound findings on the last scan before IUD diagnosis and the clinical presentation at IUD are detailed in Table 2. Apart from one case that presented with reduced fetal movements, the finding of an IUD was incidental in six of the seven pregnancies at a routine scan. Growth was appropriate for gestational age and concordant, and the AFI was normal and evenly distributed in all seven affected pregnancies at the last scan prior to the diagnosis of IUD. All fetal deaths were diagnosed within 2 wk of a normal scan (median 1⁺⁴ wk, range 1⁺⁰ to 2⁺⁰).

In the 1:3 gestation-matched case–control comparison, there was no significant difference between IUD-affected pregnancies and controls in any of the antenatal indicators of IUGR or TTTS. The mean Δ value for the abdominal circumference was 20 mm (CI −58 to +98), for the head:abdominal circumference ratio was −0.02 (CI −0.04 to 0.00), for EFW was −10 g (CI −142 to +122), for ΔEFW was −0.5% (CI −4.8 to +3.8), and for AFI was 1.2 cm (CI −1.4 to +3.8). AAA detection rates were similar in the two groups (54% in pregnancies with IUD and 60% in matched controls without an IUD). All the IUDs occurred in pregnancies monitored after the introduction of routine chorionic plate Doppler for AAA detection, as similarly did the selected 3:1 control pregnancies.

Sebire et al. suggested that the main risk of fetal death in MC pregnancies was before 24 wk, as after this, the rate of perinatal loss was only slightly higher in MC than in DC pregnancies (4.9% versus 2.8% of pregnancies) [3]. This, however, was a referral population with a high incidence of TTTS and other pathologies, and no recent ultrasound in many cases, and so is not comparable with our study of otherwise uncomplicated MCDA pregnancies which suffered an unexpected death within 2 wk of a normal scan.

Our data suggest instead that even intensively monitored, apparently healthy MCDA pregnancies remain at substantial risk of IUD after 24 wk (4.6% of pregnancies and 3.3% of fetuses). IUDs after 24 wk occurred in the third trimester, and predominantly after 32 wk of gestation, at which time the prospective risk of subsequent IUD was 1/23 pregnancies. Our findings provide useful information for counselling parents with MCDA pregnancies, and notwithstanding the limitation of small numbers, may be used to inform decisions regarding the optimal timing of delivery.

Analysis of the antenatal findings on fortnightly scans shows that deaths after 24 wk occurred unassociated with any antenatal evidence of either TTTS or IUGR, and thus were unexpected. Only two of the seven pregnancies complicated by an IUD had a definite cause at autopsy. Late onset TTTS is rare but not unknown, and here seems to have been unpredictable. Recent case reports [25,26] suggest that thrombosis of a previously patent compensatory anastomosis, either AAA or reverse arteriovenous, can result in haemodynamic imbalance and thus acute onset TTTS, as suggested in one of the two cases in this study. The fetal deaths in the present study occurred despite strategies aimed at preventing them, specifically, fortnightly ultrasound and Doppler surveillance in a tertiary fetal medicine unit, and elective delivery at 36–37 wk).

We acknowledge a number of limitations of our study. First, there is the retrospective nature of the analysis. Notwithstanding this, clinical and ultrasound findings were entered prospectively into our database, and we reviewed all cases entered over the 12-y period since its inception. Second, caution is advised in view of small numbers. However, this is the first series to our knowledge to report this outcome in otherwise normal MCDA twins, and the deaths were spread relatively evenly over the study period.Third, there is a lack of comparative data in other twin pregnancies. Use of a DC control group was precluded by they fact that our protocol for DC pregnancies differs from that for MCDA pregnancies, as in most centres, with DC monitoring performed outside the fetal medicine unit and at less frequent intervals (every 4 wk rather than every 2 wk), and with elective delivery 1 to 2 wk later than for MCDA pregnancies. In terms of complicated MC twin pregnancies, few with TTTS or IUGR and two live fetuses remain undelivered after 32 wk. We chose not to use the brain:liver ratio as an indicator of IUGR at post-mortem because of its insensitivity [27]. Birth weights were also not used as an indicator of discordant IUGR in view of the variable interval from death to diagnosis, and the even greater interval between death and delivery (up to several weeks) following single IUD, in which varying degrees of post-mortem weight loss and continued growth of the surviving twin might confound measured birth weights.

Our findings have a number of clinical implications. The high rate of unexpected third trimester fetal death might be obviated by a range of preventative strategies. One would be to increase the frequency of monitoring. Although growth is only usefully measured every 2 wk, more frequent surveillance could include amniotic fluid volume and distribution, and fetal Doppler waveforms. This might allow earlier identification of the occasional case of late onset TTTS, and thus prevention of IUD through timely delivery. However, it is not clear how more frequent monitoring would prevent unexplained IUDs, just as this strategy does not appear to prevent them in singleton pregnancies [28,29].

Another preventative approach might be earlier delivery. Complicated MCDA twin pregnancies, such as those with TTTS or abnormal umbilical artery Doppler, are already delivered in many centres at 32 wk. At this stage neonatal survival is now comparable to that at term [30,31]. Minor neonatal morbidity is still likely at 32 wk, but the chance of major respiratory and neurological problems is reduced substantially by administration of antenatal steroids [32]. Elective vaginal delivery would increase the chance of failed induction, but there is an increasing view that elective caesarean section is preferable to preterm induction of labour in MCDA twin pregnancies, both to avoid a caesarean section being performed as an emergency and to obviate the risks of acute intertwin transfusion during labour [33]. Accepting that elective delivery of all MC twins at 32 wk would carry an attendant neonatal morbidity, the complications of iatrogenic prematurity could be lessened if an intermediate gestation were chosen, e.g., 34 wk. Although neonatal morbidity would be reduced, so would the number of fetal deaths prevented. Acknowledging the limitation of small numbers in this study, 80% of fetal deaths at 32 wk or greater and 60% at 34 wk or greater, respectively, would have been prevented if fetuses were delivered before these gestation time points. When expressed per pregnancy, one case of IUD would be prevented for every 23 MC pregnancies delivered at 32 wk and one for every 30 pregnancies at 34 wk. The above figures underestimate the potential gain of such strategy, because single IUD in MCDA pregnancy also exposes the surviving co-twin to a substantial risk of brain injury and thus long-term handicap from acute intertwin transfusion [6–8].

Major neurodevelopmental sequelae are now rare in otherwise well babies delivered after 32 wk. MC twins are at increased risk of cerebral palsy compared to DC twins, but this is mainly prior to 32 wk and in pregnancies complicated by TTTS and IUGR [34,35]. Prematurity per se is not considered responsible for the high risk of neurological morbidity in MC twins, but rather the haemodynamic imbalance unique to MC placentation [36,37]. Thus, uncomplicated MC twins should not be at substantially increased risk of neurological sequelae from elective premature delivery at or after 32 wk. Indeed, elective delivery may reduce their risk of neurodevelopmental injury, as single IUD in MC twins is a well-established risk factor for cerebral palsy [6,38].

In conclusion, apparently uncomplicated MCDA pregnancies remain at substantial risk of fetal loss as they approach 32 wk of gestation, whereafter one in 23 (4.3%) will suffer an unexpected IUD. This risk is exacerbated by the 40%–50% risk of co-twin death or neurological injury following a single IUD [6]. If our findings are confirmed in other observational series, our suggestion that earlier delivery might prevent this adverse outcome could be tested by randomised trial.