Were stillbirth to be listed among the leading causes of death in the United States, it would rank 11th, just ahead of chronic liver disease and cirrhosis. Were it a cancer, it would rank sixth because it annually causes about 20% more deaths than all leukemias combined. And these figures almost certainly underrecord this quasi-vital event.1
With such numbers, one might expect a robust public health interest in prevention, but such is not the case. Three million annual deaths are not sufficient to make stillbirth an international health priority.2 Still, the Centers for Disease Control held workshops in 2005 on stillbirth surveillance,3 and this year the Lancet has published a series of papers on stillbirth as a global public health problem.4,5 Interesting epidemiologic research on stillbirth using large US databases has been published in recent years,6–8 and NICHD and several universities are collaborating in a national stillbirth surveillance and case-control study, the latter including near-universal autopsy and placental assessment—essential prerequisites to any major advance in understanding.9
Although scientific interest seems to be on the upswing, we remain remarkably ignorant of the causes of stillbirth. And, as the paper by Auger et al10 in this issue indicates, we are uncertain even as to how best to describe its frequency.
It is now nearly a quarter century since Pat Yudkin and colleagues11 at Oxford pointed out that the appropriate denominator for stillbirths at gestations before term is the population of fetuses in utero at the time of the stillbirth. This approach takes, as the denominator for stillbirths at a given week of gestation prior to term, all live births in the population minus all births delivered prior to the gestational week under consideration, an approach referred to as “fetuses-at-risk.”
The conventional way of describing stillbirth rates at early gestation had been to use live births of the same gestational age as the denominator. But as a denominator, gestational age-specific live births fail on two counts. First, they are not the source population of the numerator. Second, they represent another adverse perinatal outcome, premature live birth. It should be noted that once all, or nearly all, fetuses have been delivered, the fetuses-at-risk approach converges with the conventional approach of using live births (or, more properly, live births and stillbirths), as the denominator. Thus, stillbirth rates at term, or total stillbirth rates at all gestations, are unaffected by use of the fetuses-at-risk approach.
A methodologic innovation is valuable by itself, but the Oxford team also drew an important clinical conclusion from their observation. The stillbirth risk, as they termed their new statistic, rises in frequency with increasing gestation, while the stillbirth rate (their term for the conventional frequency measure) declines. This served to confirm obstetrical practice, which increases clinical surveillance in pregnancy as term approaches.
The specific contribution of the paper by Auger et al10 is to show empirically how the conventional approach to preterm stillbirth expression is flawed. Maternal education is inversely associated with virtually every perinatal outcome; therefore, it seems counterintuitive to suppose that this relationship would be reversed for stillbirths born prior to the 28th week of gestation. Yet, this is the finding when the conventional denominator is used, not when the fetuses-at-risk approach is used. It is evident that low maternal education is associated with both stillbirth and preterm birth before 28 weeks, but somewhat more so for the latter manifestation of perinatal adversity. The apparent protective role of maternal education against early stillbirth is thus nothing more than an indication of its weaker association with one pathology than another.
Nonetheless, the fetuses-at-risk approach has been criticized by Feldman12 and by Cheung,13 both of whom point out that the gestation-week-specific risk of stillbirth focuses on the immediate risk in the gestational week of interest, ignoring all stillbirths yet to occur in later gestations to the at-risk population of fetuses. They advocate including these later stillbirths in the estimate. Cheung argues that the stillbirth rates calculated by the Yudkin et al method “… do not indicate that mortality among fetuses at risk is lower in the preterm period than in the term period. They indicate simply that in the preterm period, death is less imminent.”13 By analogy, mortality is not really lower in the young, merely less imminent. All epidemiologic expressions of mortality are age-delimited and do not consider all deaths that eventually will occur in the cohort under study.
Does the fetuses-at-risk concept have broader applicability? In some sense, all phenomena consequent upon delivery before term might legitimately use fetuses still in utero at the time of birth as the denominator, but this position can lead to its own counterintuitive results, such as the finding that the risk of cerebral palsy increases with gestational age. Joseph et al14 find that applying the fetuses-at-risk concept to cerebral palsy, the risk of this highly gestational age-related outcome among live births (births <28 weeks experience 50 times the risk of cerebral palsy as births at term) is reversed to a 10-fold higher risk at term than at 28 weeks when fetuses at risk are the denominator.
Extending the fetuses-at-risk concept to nonfetal outcomes requires ignoring the required transitions that permit live birth (delivery) or the acquisition of a disability (neonatal survival). These transitions create risks so distinct from remaining in utero, particularly for the preterm, that they cannot be ignored. To select the appropriate controls in case-control studies, we often try to imagine the cohorts from which cases and controls arise. Reversing this thought experiment, we can ask whether a fetus in utero at 32 weeks (who is most likely to be born at term) would make an appropriate control in an etiologic study of cerebral palsy for a case born at 32 weeks. Because more than 95% of births at 32 weeks do not develop cerebral palsy, such a case-control study would in the main identify risk factors for preterm birth, not for cerebral palsy. It seems essential that controls be selected from the universe of fetuses who have gone through the experience required for caseness, that is, birth at 32 weeks, but who nonetheless did not acquire cerebral palsy. This makes live births a much more reasonable denominator for cerebral palsy rates than fetuses at risk.
Stillbirth remains an enigma—difficult to study; difficult even to describe. As long as stillbirths are as frequent as infant deaths (about 25,000 each in the United States annually), stillbirth merits the serious attention of epidemiologists.
1. Greb AE, Pauli RM, Kirby RS. Accuracy of fetal death reports: comparison with data from an independent stillbirth assessment program. Am J Public Health. 1987;77:1202–1206.
2. Frøen JF, Cacciatore J, McClure EM. Stillbirths: why they matter. Lancet. 2011;377:1353–1366.
3. Duke CW, Correa A, Romitti PA. Challenges and priorities for surveillance of stillbirths: a report on two workshops. Public Health Rep. 2009;124:652–659.
4. Cousens S, Blencowe H, Stanton C, et al.. National, regional, and worldwide estimates of stillbirth rates in 2009 with trends since 1995: a systematic analysis. Lancet. 2011;377:1319–1330.
5. Flenady V, Koopmans L, Middleton P, et al.. Major risk factors for stillbirth in high-income countries: a systematic review and meta-analysis. Lancet. 201116;377:1331–1340.
6. Salihu HM, Sharma PP, Ekundayo OJ, et al.. Childhood pregnancy (10–14 years old) and risk of stillbirth in singletons and twins. J Pediatr. 2006;148:522–526.
7. Luo ZC, Liu S, Wilkins R, et al.. Risks of stillbirth and early neonatal death by day of week. CMAJ. 2004;170:337–341.
8. Ananth CV, Basso O. Impact of pregnancy-induced hypertension on stillbirth and neonatal mortality. Epidemiology. 2010;21:118–123.
9. Parker CB, Hogue CJ, Kocha MA, et al.. Stillbirth Collaborative Research Network: design, methods and recruitment experience. Paediatr Perinat Epidemiol. 2011;25:425–435.
10. Auger N, Delézire P, Harper S, et al.. Maternal education and stillbirth mortality: estimating gestational age-specific and cause-specific associations. Epidemiology. 2012;23:247–254.
11. Yudkin PL, Wood L, Redman CW. Risk of unexplained stillbirth at different gestational ages. Lancet. 1987;1:1192–1194.
12. Feldman GB. Prospective risk of stillbirth. Obstet Gynecol. 1992;79:547–553.
13. Cheung Y-B. On the definition of gestational-age-specific mortality. Am J Epidemiol. 2004;160:207–210.
14. Joseph KS, Allen AC, Lutfi S, et al.. Does the risk of cerebral palsy increase or decrease with increasing gestational age? BMC Pregnancy Childbirth. 2003;3:3–8.
ABOUT THE AUTHOR
NIGEL PANETH is a pediatrician and perinatal epidemiologist in the Department of Epidemiology at Michigan State University. His principal research interest is in the prevention of perinatal mortality and the major disabling neurodevelopmental disabilities, especially cerebral palsy, with a particular focus on links between these outcomes and pregnancy and perinatal exposures.