Interpretation
In our study, the SMFM definition identified 115 cases (4.91%) of FGR, the ACOG definition identified 63 cases (2.7%) of FGR, while the ISUOG definition identified 48 cases (2.05%) of FGR. However, when Molina et al.(12) and Roeckner et al.(13)analyzed the data of the Redefining Fetal Growth Restriction (RFGR) project, the ACOG definition identified 8.63% (91/1055) of FGR cases, the SMFM definition identified 13.0% (137/1054) of FGR cases, and the ISUOG definition identified 5.2.% (55/1055, 55/1054) of FGR cases, with percentages that were higher than the incidence observed in this study. The reason of the higher incidence was that the population enrolled in the RFGR program was at high risk of developing FGR. High risk factors included uterine height inconsistent with gestational age, hypertension, diabetes, and other chronic kidneys, blood vessel, and hemoglobin diseases. In their study, pregnant women with chronic hypertension and pre-pregnancy diabetes accounted for 14% and 8% of the sample, respectively. Nevertheless, the samples in our study were from the general population. The percentages of chronic hypertension and pregestational diabetes mellitus in this study were 1.1% and 1.3%, respectively Furthermore, our institution offered third trimester ultrasound as a routine protocol, which provided sufficient ultrasound biometric information for the calculation of EFW in the late pregnancy.
The reason why SGA was selected as a primary outcome is that SGA is associated with ANO and adult cardiovascular diseases and metabolic diseases (16; 17). In addition, it has a clear definition (birth weight less than the 10th percentile). Therefore, many studies have used SGA as one of the indicators of the accuracy of FGR predictions.(11; 12; 13). The results from this study indicate that the performance of the SMFM-FGR definition to predict SGA was higher than those of the ACOG-FGR and ISUOG-FGR criteria (AUC: 0.69 vs. 0.62 and 0.60), mainly because the sensitivity of the SMFM-FGR criteria to detect SGA was significantly higher than those of the ACOG-FGR and ISUOG-FGR criteria (40.8% vs. 24.5%, 20%). In 2018, Blue et al. compared the ability of RCOG (same as SMFM-FGR) and ACOG criteria for FGR to predict SGA and found that the SMFM-FGR criteria were slightly better than those of the ACOG-FGR criteria (AUC: 0.78 vs. 0.76)(11). When comparing the SMFM-FGR and ISUOG-FGR criteria, Roeckner et al. also found that the sensitivity of the SMFM-FGR criteria to predict SGA was significantly higher than that of the ISUOG-FGR criteria (54.7% vs. 28.8%), while the specificity was slightly lower than that of the ISUOG-FGR criteria (93.3% vs. 98.4%)(13).
As is known that not all SGA cases are pathological. Most of SGAs are constitutional. Using the SGA alone to assess compare the predictive ability of different FGR definition was comprehensive. Considering that FGR fetuses were at an increased risk of fetal and neonatal mortality and complications due to their reduced growth potential, there were also a few studies which used neonatal complications and mortality as primary outcome, thereby the concept of composite ANO being proposed(12; 13). When Molina et al. compared the ability of the ACOG-FGR and ISUOG-FGR criteria to predict ANO, the performance of the ISUOG-FGR criteria was superior to that of the ACOG-FGR criteria (sensitivity: 10.1% vs. 9.3%; specificity: 95.5% > 91.5%) (12). In our study, the sensitivity of the ACOG-FGR criteria for predicting ANO was slightly higher than that of the ISUOG-FGR criteria, but the specificity was slightly lower, and the AUC of the two groups was similar (0.552). When Roeckner et al. compared the SMFM-FGR and ISUOG-FGR criteria, they found that, although the sensitivity of the SMSM-FGR criteria was higher than that of the ISUOG criteria, the specificity was lower than that of the ISUOG criteria, and the AUC of the two groups was 0.51 vs. 0.53. In this study, the sensitivity of the SMFM-FGR criteria for predicting ANO was higher than that of the ISUOG criteria, but the specificity was slightly lower, and the final AUC for the SMFM-FGR criteria was slightly higher. However, the differences in the medical resources and capabilities of different institutions to deal with neonatal complications also led to differences in ANO between studies.
Many studies have shown that different growth curves produce different percentiles, thereby affecting the rates of detection of SGA and LGA.(18; 19). In addition, different ethnic groups have different growth potentials. For example, the average birth weight of newborns in India is 2.9 kg, and the average birth weight of newborns in the UK is 3.5 kg, with a difference of 600 g.(20). Therefore, the National Institute of Child Health and Human Development (NICHD) developed different growth curves for Caucasian, Black, Hispanic, and Asian fetuses.(21) Because there is no authoritative local growth curve in China, the EFW percentile in this study was determined based on the growth curve of Asian women developed by the NIICHD(21; 22). The percentiles for newborns are based on newly released data of newborns in nine cities in southern and northern China in 2020.(23).