The association between temporal changes in the use of obstetrical intervention and small- for-gestational age live births

Temporal changes in SGA rates among singletons in the United States followed a non-linear pattern, with substantial declines between 1990–92 and 2002–04 and a small increase between 2002–04 and 2008–10. These declines in SGA rates were not consistent with the observed pattern of changes in the frequency of risk factors for fetal growth restriction such as maternal smoking, which decreased monotonically from 1990–92 to 2008–10. The SGA pattern was also not consistent with changes in older maternal age and nulliparity, both of which increased in a linear fashion between 1990–92 and 2008–10. The complex pattern of changes in SGA rates was more consistent with changes in the gestational age distribution; rates of late preterm birth, and early term birth increased substantially between 1990–92 and 2002–04 before declining in the period between 2002–04 and 2008–10. In our study, the sensitivity analyses restricted to smokers showed similar overall patterns in obstetrical intervention and SGA indicating that changing maternal characteristics (such as declines in maternal smoking) cannot explain temporal trends in SGA rates. Kitagawa decomposition analyses using the fetuses-at-risk approach supported these findings by showing that the decline in the SGA rate between 1990–92 and 2002–04 was entirely due to large changes in the gestational age distribution, whereas the small increase in SGA rates between 2002–04 and 2008–10 was due to adverse changes in gestational age-specific rates (which overwhelmed smaller positive changes in the gestational age distribution). Increases in gestational age-specific rates of SGA between 2002–04 and 2008–10 adversely impacted overall SGA rates-presumably the adverse effects of increases in older maternal age, and nulliparity on fetal growth were stronger than those of declines in maternal smoking and increases in pre-pregnancy weight.

There is some uncertainty in the literature with regard to the utility of the fetuses-at-risk approach for modeling postnatal phenomena such as SGA live births [16‐19]. The traditional model, which shows stable SGA rates across pregnancy, receives some support from observed increases in birthweight with increasing gestational age that result from increases in uterine artery blood flow and other physiologic changes that occur through the course of pregnancy. Changes in fetal growth over the last two decades under this model are best viewed as the consequence of increases in some risk/protective factors (e.g., nulliparity, older maternal age, and pre-pregnancy obesity) and decreases in others (e.g., maternal smoking). However, the constancy of growth restriction across gestation is difficult to reconcile with changes in fetal and perinatal mortality with increasing gestation.

The fetuses-at-risk model, on the other hand, shows that rates of SGA increase with advancing gestation. The utero-placental system’s ability to support the fetus declines over the course of pregnancy and this viewpoint is supported by numerous animal and human studies which show that blood flow per unit fetal weight declines toward later gestation [20‐23]. Early delivery, whether at late preterm, early term or late term gestation, will result in lower rates of SGA. This does not necessarily imply improved neonatal health status; indeed the DIGITAT trial which showed a reduction in SGA (<3rd percentile) from 30.6 % in the expectant management group to 12.5 % in the labour induction group did not show a significant difference in neonatal mortality or morbidity rates [5].

The reasons for the observed increase in obstetrical intervention are poorly understood. A proportion of this increase can be explained by changing maternal characteristics such as increased rates of delayed childbearing and obesity which increase pregnancy complications such as hypertension and diabetes [24]. Additionally, with technological advances in fetal surveillance and monitoring, clinicians are now more aware of subtle changes in fetal wellbeing [2]. However, in general clinical practice it is estimated that less than one-third of fetal growth restriction is identified antenatally [25]. It is estimated that much of the increase in obstetrical intervention for term deliveries appears to have been for informal medical indications or non-medical reasons [8, 26, 27]. Also, a lower threshold for intervention in recent times may relate to the safety of labour induction and the relative rarity of severe neonatal morbidity and mortality after 34 weeks of gestation [2]. None of this contradicts the proposition that SGA rates, which increase with advancing gestation under the fetuses-at-risk model, will decline with a left-shift in the gestational age distribution. In recent years, clinical guidelines have recommended that elective deliveries be scheduled in the late-term period (i.e. 39 weeks) [28]. While this has likely contributed to changes in the gestational age distribution with reductions in late preterm/early term births, both randomized controlled trials and observational studies of routine clinical practice have not found a significant impact on short-term maternal or neonatal health outcomes [29, 30]. The impact of these policies on long-term neurological development is currently unknown.

Limitations of our study include use of data from US birth certificates which may contain some transcription and other errors. Several studies have confirmed that the accuracy of this data varies by specific data element [31‐33]. Additionally, US birth certificates do not collect the indication for labour induction or cesarean delivery [34] and validation studies show that birth certificates over-estimate the rate of obstetrical intervention [34]. While misclassification is a concern, no studies have indicated that this misclassification is differential over time [32]. The increased use of ultrasound for pregnancy dating may have impacted gestational age trends in our study [35‐38]. Ultrasonography for pregnancy dating (which would be reflected in the clinical estimate of gestational age) leads to an artefactual shift in the gestational age distribution with fewer post-term births compared to menstrual-based estimates of gestational age. Other limitations included our inability to distinguish elective cesarean delivery (which shortens gestational duration) from emergency cesarean delivery (which does not materially alter gestational duration). Also, our data source did not include information on pre-pregnancy weight; increases in pre-pregnancy obesity are another potential explanation for the temporal changes in SGA rates. Information on other interventions such as the use of assisted reproductive technologies which have been shown to result in shorter lengths of gestation and lower birth weights than naturally conceived pregnancies [39‐41], was not consistently available throughout the study period and hence not examined as a potential explanatory factor for the temporal changes in SGA rates. Our study assumed that increases in early delivery rates were due to increases in obstetric intervention. However, some increase in spontaneous preterm birth and spontaneous early term birth may have occurred because of increases in other factors such as multiple births, older maternal age and parity. Studies which have documented a left shift in the gestational age distribution for spontaneous deliveries show a shift from delivery at ≥40 weeks to delivery at 39 weeks of gestation [42, 43]. However, these studies also show that the rate of scheduled births has increased, and the gestational age at which obstetric interventions are used has decreased [43]. Finally, we used neonatal growth curves instead of fetal growth curves, although the former are known to underestimate SGA at preterm gestation. We chose to use neonatal curves because our study was based on SGA as assessed among live births; fetal growth curves provide ultrasound-based fetal weight and this corresponds poorly with birth weight, especially at the extremes.