Among 3325 women enrolled in the study, 240 (7.22%) of them had MS-AFP ≥ 2.5 MoM, and 3085 (92.78%) had MS-AFP < 2.5 MoM. Participants’ basic characteristics such as maternal age, weight, body mass index, parity, smoking history, gestational diabetes mellitus, neonatal gender and delivery mode, and screening markers β-HCG and nuchal translucency (NT) were all not significantly different between women with MS-AFP ≥ 2.5 and < 2.5 MoM. However, women with elevated MS-AFP had lower PAPP-A levels (P = 0.001) and higher proportion of assisted reproductive technology (P = 0.003). (Table 1)
A total of 594 pregnancies resulted in APOs, among which 181 in preterm birth, 32 in stillbirth, 81 in preeclampsia, and 362 in SGA. The incidence rate of overall APOs among women with MS-AFP ≥ 2.5 and < 2.5 MoM were 29.58 vs. 16.95% (P = 0.000). The incidence rate of each APO in the study population and its corresponding rate in the elevated and normal MS-AFP groups were 5.44% (11.67 vs. 4.96%, P = 0.000) for preterm birth, 0.96% (1.25 vs. 0.94%, P = 0.500) for stillbirth, 2.44% (6.25 vs. 2.14%, P = 0.001) for preeclampsia, and 10.89% (17.92 vs. 10.34%, P = 0.001) for SGA. Women with MS-AFP ≥ 2.5 MoM had increased risks (OR, 95% CI) of preterm birth (2.53, 1.65~3.88), preeclampsia (3.05, 1.71~5.43) and SGA (1.90, 1.34~2.69), while the risk of stillbirth was not significantly increased (1.33, 0.40~4.41). (Table 2)
In order to figure out the association of individual MS-AFP level with the incidences of APOs, the ROC curve was used to evaluate the predictive efficiency. The areas under ROC curves (AUROC) were 0.569 for overall APOs (P = 0.000), 0.572 for preterm birth (P = 0.001), 0.597 for stillbirth (P = 0.060), 0.579 for preeclampsia (P = 0.015) and 0.565 for SGA (P = 0.000) (Figure 2). It could be seen that the AUROC was small for each APO, and the significance was not obtained for stillbirth probably due to its lower incidence than that of the other three APOs.
Considering that MS-AFP might perform better for more severe APOs such as SGA < 3rd percentile, or preterm birth at < 32 weeks, or severe preeclampsia, we further evaluated the performances of MS-AFP for these subdivisions of APOs. It could be seen that MS-AFP did not perform remarkably better as predictors of more severe APOs (Table 3). When preterm birth was divided into spontaneous (n=145) and iatrogenic (n=36), a mildly improved AUROC was obtained: 0.631 for iatrogenic (P = 0.007) vs. 0.556 for spontaneous (P = 0.023) (Table 3). Since SGA consists also of group of healthy small fetuses, we tried to use ultrasound definition that was put forward by Gordijn et al in 2016 to discriminate fetal growth restriction (FGR) cases [21]. 104 cases of FGR were defined in 362 SGA, and the performance of MS-AFP for FGR was mildly improved, with an AUROC of 0.602 (P = 0.000) compared to 0.565 for SGA (P = 0.000) (Table 3). To sum up, first-trimester MS-AFP was not a good predictor for these APOs, even if the subdivisions of APOs were applied for analysis.
To concretely show the discriminatory power of MS-AFP for preterm birth, a Kaplan-Meier curve was used to record the gestational weeks at delivery for all participants. Women with MS-AFP ≥ 2.5 MoM had an earlier overall distribution of gestational weeks at delivery (P = 0.004). However, the actual difference was minuscule; the most significant difference in gestational weeks at delivery were concentrated in 36~37 weeks in which the fetuses are close to maturity and are less prone to adverse outcomes (Figure 3a). A boxplot was used to show the birth weights of all neonates. Women with MS-AFP ≥ 2.5 MoM had a lower overall distribution of neonatal birth weights (P = 0.000). However, the actual difference was also minuscule; the medians of neonatal birth weight were 3200 and 3050g respectively in the two groups (Figure 3b).