Discussion
In this large cross-sectional study based on US births, we evaluated the impact of gestational age on delivery mode for pregnancies complicated by SGA. Because the gestational age when delivery becomes indicated may influence labour outcomes, we performed a causal mediation analysis to quantify the mediating and interaction effects of gestational age on exposure to SGA and delivery mode. The principal finding of this study is that the risk for caesarean delivery with TOL at term gestations and caesarean delivery without TOL across all gestations for SGA births is primarily attributable to exposure to SGA rather than mediation or interaction effects of gestational age.
The caesarean delivery without TOL outcome includes women who had prelabour caesareans. The intent of performing a prelabour caesarean for some obstetricians is to avoid impending complications as a consequence of labour, including category 2 fetal heart rate tracings that may precipitate an emergent caesarean delivery. In this analysis, caesarean rates without TOL were higher for SGA births at each gestational age strata except for 39-40 weeks’ gestation. The decomposition analysis revealed that exposure to SGA by itself (and independent of gestational age) was the driver of this risk.
The caesarean delivery after TOL outcome includes those who required caesarean deliveries for abnormal fetal heart rate patterns and labour complications. The rates of caesarean delivery after TOL were higher for SGA births across gestational age strata, but the highest rates were observed ≥41 weeks’ gestation. When these rates were highest at term gestations, exposure to SGA was the dominant driver of caesarean delivery risk.
Prior studies have shown high rates of caesarean delivery for SGA pregnancies. For example, in a study of 2885 non-anomalous, singleton, cephalic presenting SGA fetuses at 25-34 weeks in New York (1995-2003), the overall caesarean delivery rate was 57.9% (26). That study was limited by small size, but also did not stratify the analyses by gestational age ranges, caesarean history, or caesarean indications. As a result, a picture emerges of caesarean risk for SGA pregnancies that may fuel concerns about high rates of emergent deliveries. While the results of this study also found higher overall caesarean delivery rates from 22-44 weeks’ gestation for SGA versus AGA pregnancies (34.3% for SGA <5th percentile and 36.4% for SGA <3rd percentile versus 29.6% for AGA), a more nuanced picture emerges when caesarean rates are broken into gestational age ranges. For example, there were extremely high rates <32 weeks’ gestation (over 80%), but the rates improved as gestational age advanced. This breakdown by gestational age provides improved insight into the caesarean risk for SGA pregnancies.
The results of this analysis also demonstrated that labour outcomes for SGA pregnancies are impacted by indication for delivery. The overall high caesarean rates for SGA fetuses in the US were largely driven by prelabour procedures. Approximately two-thirds of all Caesareans were classified as prelabour procedures, but the risk was highest <32 weeks’ gestation. Further, Caesarean delivery rates without TOL were considerably higher at each gestational age strata compared to rates with TOL, except at ≥41 weeks’ gestation. These findings give credence to apprehensions raised by the NIH Consensus statement “Preventing the First Caesarean Delivery,” that “concern[s] about vaginal delivery coupled with relative indifference regarding the risks of Caesarean may lead to a decision that is not based on clinical evidence” (1). The results of this analysis suggest that too many patients in the US are not given the opportunity to have TOL and spontaneous vaginal delivery.
In fact, while the rates of Caesarean delivery after TOL are higher for SGA versus AGA pregnancies, the magnitude of risk is less than some providers may recognize. Overall rates of caesarean with TOL were 9.1% for SGA <5th percentile and 10.2% for SGA <3rd percentile versus 5.9% for AGA births. The caesarean with TOL risks were notably highest for SGA and AGA births after 41 weeks gestation. Also, the risks of caesarean delivery with TOL was lowest for SGA versus AGA births at 39-40 weeks’ gestation and with the smallest risk difference at this gestation age range. These observations corroborate the main findings of the ARRIVE trial (27) that induction of labour in the 39th week reduces Caesarean delivery rates.
The results of this mediation analysis suggest that different mechanisms drive the risk of caesarean delivery after TOL at preterm versus term gestations. At preterm gestations, interaction of SGA with gestational age has a larger effect on the risk, whereas exposure to SGA itself has a large effect at term gestations. Unfortunately, this dataset cannot explain how the interaction affect contributes to caesarean delivery after TOL rates in preterm gestations. It is possible that the fetal heart rate monitoring profile of preterm SGA fetuses could provide insight, but this has not been elucidated. A 2010 Cochrane review that sought to assess the effects of a policy of elective caesarean delivery versus expectant management for growth restricted fetuses could only identify 6 trials with 122 patients (28). The included studies were marred by recruitment problems, which limited the conclusions of the systematic review. The manifestation of gestational age-SGA interaction as it relates to fetal tolerance of labour at preterm gestations requires further evaluation.
Exposure to SGA alone (i.e., the CDE) explains increased risks of caesarean delivery after TOL for term and late-term SGA births. In this case, it is likely that the impact of pathologic growth restriction drives the risk and likely reflects underlying ischemic placental disease (IPD) (6, 29). IPD, which also includes preeclampsia and placental abruption, has been implicated in over 50% of all clinician-initiated preterm deliveries and may play a more substantial role in term gestations (6, 29).
The main strength of the study is the application of a causal mediation approach, a cutting edge statistical methodology (30), to evaluate the individual contribution of SGA and the impact of gestational age on caesarean risk. Many studies in the obstetrical literature inappropriately attempt to adjust for gestational age as a confounder, leading to biased associations since gestational age may be on the causal pathway for many adverse perinatal outcomes (31, 32). Using causal mediation analysis, the study results clarify the impact of gestational age as an intermediate on the SGA-labour outcome paradigm to elucidate the causal relationship between exposure and outcome.
We explicitly outlined the assumptions that we used to make causal claims in this mediation analysis (12). Further, these assumptions were assessed by a sensitivity analysis that evaluated the extent unmeasured confounding could bias our results. The E-values were 2-3 fold larger than the observed effect estimates, suggesting that it would take very large effect sizes of unmeasured confounders to nullify the associations, a situation that is unlikely. This methodological approach strengthens the study, and suggests robustness of the findings.
The study has some limitations. The data could not provide insight into spontaneous versus clinician-initiated preterm deliveries. We wanted to assess labour outcomes, including abnormal fetal heart rate patterns, but this information is not readily available from the data source. In order to capture women who had caesarean deliveries for fetal indications after unsuccessful TOL, we had to include all women who were coded as having caesarean delivery after TOL. Some of these patients required caesarean delivery for labour dystocia as well as maternal indications. Although we were able to avoid confounding due to prior caesarean delivery by limiting the analytic cohort to women of live birth parity 2, we recognize that our primary outcome has some limitations.