Average birth weights are known to be lower than intrauterine (fetal) weights early in pregnancy [21, 22], strongly suggesting that preterm births are undergrown relative to fetuses who remain in utero at the same GA. Birth weights observed at preterm gestational ages are thus “abnormal” overall. The differences observed in Swedish births (Table 1) between preterm live birth weights and estimated intrauterine fetal weights were similar to those reported by Secher et al.  and Hutcheon et al. . At 32 weeks, Hutcheon et al. estimated that the median live birth weight is 120 g (i.e., about 0.5 SD) lower than the median fetal weight.
As shown in our study, stillbirth weights were at least one SD below live birth weights at preterm GAs. The fact that stillbirths weigh less at delivery than live births at the same GA has been previously documented [34, 35]. Previous studies have also reported that among stillbirths whose time of death is known, most are delivered within 24 hours of fetal death and that the median interval from confirmation of cardiac activity until documentation of death is 7 hours [12–14]. The fetus may lose weight in utero if the death occurs several days or weeks before delivery [3, 35]. This is not likely for most stillbirths, however, since the delay between death and delivery is often substantially lower [12–14]. The difference we observed between live births and stillbirths at preterm gestations might be due to the larger proportion of congenital anomalies in preterm stillbirths than among preterm live births . After excluding all births with reported congenital anomalies, however, preterm stillbirths continued to weigh substantially less than preterm live births.
In our analysis of the Swedish data, we were surprised not to observe significant birth weight differences between antepartum and intrapartum stillbirths at preterm GAs. This result does not agree with previous reports by Chard  and Alberman et al.  Alberman et al., however, did not adjust birth weights for gestational age, and antepartum stillbirth gestational ages averaged 2 weeks lower than those of intrapartum stillbirths , while Chard included only stillbirths with GAs 24–32 weeks . The absence of difference in weight between antepartum and intrapartum stillbirths should perhaps not be so surprising at preterm gestations. Why would an intrapartum stillbirth occur at early gestation unless it was seriously compromised before going into labour? In contrast, at GAs of 37–42 weeks, we observed a large difference in weight between antepartum and intrapartum stillbirths but no difference in weight between intrapartum stillbirths and live births.
Assuming little or no weight loss among most antepartum stillbirths from death to delivery, preterm stillbirths appear to be more growth-restricted than live births at the same GA. It has been shown that size at birth is affected by growth velocity early in gestation and that growth restriction diagnosed at delivery is preceded by slower first- and early second-trimester fetal growth [37, 38]. As shown in Figure 1, the difference in birth weight between live births and stillbirth increased with advancing GA, suggesting that fetuses who subsequently died in utero grew more slowly before their death, i.e., that fetal growth restriction in stillbirths is a cumulative process.
An important limitation of our study is the lack of a good ultrasound-based fetal weight reference at preterm gestations, which would improve estimation of the difference between the birth weights of live births and the fetal weights of their counterparts who remain in utero. Hadlock’s fetal growth curve , based on a modest number of White fetuses from the 1980s, does not appear to fit well with our study sample (U.S. cohort), because birth weight-for-gestational-age z-scores among preterm live births were not substantially below 0 at most preterm GAs, as has been observed in previous studies [21, 22]. Another limitation is the lack of information on the precise timing of fetal death; we are thus unable to identify preterm fetuses who died long before the time of diagnosis. This is in particularly true for very preterm stillbirths, owing to less frequent prenatal care visits at early gestations. Furthermore, lack of information on the causes of stillbirths prevents us from exploring the relation between stillbirth weights and causes, and thus the relative weight deficits of stillbirths that are “unexplained.” Finally, the data used in this study are fairly old; the latest U.S. data available for fetal death are from the 2005. There is no reason to suspect, however, that the difference in birth weight between preterm live births and stillbirths would have changed in more recent years.
Our study is further limited by the use of nation-wide vital statistics databases, in which coding errors are known to occur [39, 40]. In particular, inaccurate estimation of gestational age in the U.S. has been widely reported and discussed [24, 25]. However, our sensitivity analysis based on the Swedish data, in which gestational age is primarily based on second-trimester ultrasound estimates and has been well-validated , solidifies our findings. Reporting of fetal deaths is not uniform across the states; a few states report fetal deaths for all periods of gestation and for birth weight <350 g . The use of Table 1 in Alexander et al's excluded mostly extremely low birth weight (65% of the excluded stillbirths were <350 g and nearly 3/4 were < 500 g) due to such non-uniform reporting and thus caused unbalanced exclusion between live births and stillbirths. Without the exclusion, the difference in birth weight between live births and stillbirths at preterm would have been larger. The exclusion of California due to lack of data on clinical estimate of gestation age also limits the generalizability of our study. The proportion of births in California is about 13% of the total U.S. births. Based on our previous study of birth weight trends at term , in which menstrual estimates of gestation were used in a sensitivity analysis, such an exclusion is likely to have little impact on our results.
Many preterm births are growth-restricted. Our study suggests that growth restriction is a cumulative process and that those fetuses who subsequently die preterm in utero grow even more slowly then those born alive at the same GA. Reduced placental blood supply and the consequent reduction in delivery of oxygen and nutrients to developing fetuses may both restrict fetal growth early in gestation and increase the risk of subsequent stillbirth. Suboptimal placental function and poor growth early in pregnancy limit fetal growth for the remainder of pregnancy. Early and routine ultrasound scans to monitor fetal size and growth during pregnancy might theoretically reduce the risk of stillbirth. This is perhaps the most important clinical implication of our findings. Before implementing routine clinical monitoring, however, an improved intrauterine fetal weight reference is required. Large-scale NIH- (U.S. National Institutes of Health) and Gates Foundation-funded studies of ultrasound-based fetal size assessment currently under way should provide much-needed data on normal fetal growth patterns. Future research should attempt to understand the pathophysiological mechanisms underlying growth restriction, as a basis for developing and testing preventive and therapeutic interventions to reduce stillbirth risk.