Main findings
There has been great heterogeneity in the study design, statistical analysis and presentation of results in reported formulae for calculating GA [14]. Some of the studies included nonselected or low-risk pregnant women, but most of them did not use adequate quality control standards. In other studies, retrospective analyses were used to obtain data from clinical workstations, with a high risk of bias. The purpose of our study was to establish a formula for calculating GA for naturally conceived fetuses at 11–14 weeks of gestation in the mainland of China. Unified enrollment standards, clinical procedures, data collection procedures and strict quality control procedures were used, so that the examination results could be promoted widely. All data in this study were measured specifically for the purpose of this study, and were obtained from a prospective study rather than a retrospective clinical database. According to the study design recommended by Altman et al. [20], low-risk pregnant women with a naturally conceived singleton pregnancy, and a clear LMP date were selected, and each fetus was measured only once during pregnancy. All pregnant women were followed up until the birth of the fetus, and data were excluded only when the fetus presented with congenital malformations or intrauterine death, to avoid generating an abnormal database.
In this study, GA was calculated according to the first day of the LMP. We selected only pregnant women who had records of regular menstrual cycles of 28–30 days for at least 1 year before pregnancy. They had not taken ovulation-inducing or contraceptives drugs or other estrogenic hormones for at least 6 months before pregnancy. Other methods used previously to calculate the GA have included the date of oocyte collection in ART cycles, serum human chorionic gonadotropin levels, elevated luteinizing hormone levels, the timing of embryo transfer in ART cycles, ultrasound detection of follicular rupture, cervical mucus morphology, basal body temperature increases, and the date of sexual intercourse. Some studies have used the GA of fetuses produced by in vitro fertilization (IVF) as a gold standard. However, fetuses produced through IVF cannot be identical biologically to naturally conceived ones because there might be differences between the dates of ovulation and conception. IVF-derived fetuses might also have growth difference during the first trimester. The biological characteristics of pregnant women following ART might be different from low-risk women with natural conceptions. Therefore, we believe that the application of formulae for evaluating GA derived from ART-conceived fetuses to naturally conceived fetuses is probably invalid.
The formula for calculating GA obtained in this study was based on univariate linear regression analysis, similar to that of McKennan et al. [9]. Systematic prediction errors and random prediction errors were used to evaluate the differences in formulae for estimating GA between this study and others. Sladkevicius et al. [21] summarized 21 CRL-based dating formulae, among which three were selected from naturally conceived fetuses with relatively large sample sizes, including the studies by Robinson et al. [7] in 1975, Hadlock et al. [6] in 1991 and von Kaisenberg et al. [11] in 2002. However, Sahota et al. [10] pointed out that there was actually no formula for calculating GA in the study by von Kaisenberg et al. [11], which was incorrectly derived by Sladkevicius et al. [21] based on a size estimation formula. Napolitano et al. [14] carried out a systematic analysis on formulae for calculating GA, and four studies with scores higher than 18 points (29-point maximum) were selected: Sahota et al. [10], Verburg et al. [12], Robinson et al. [7], and McKennan et al [9]. In 2014, Papageorghiou et al. [13] proposed a “worldwide” CRL-based dating formula from data of 4321 fetuses. The results of our study was compared with the above six studies [6, 7, 9, 10, 12, 13] and our formula lay in the middle of them. Three studies overestimated the GA, and three underestimated it, compared with our study. Our results were very consistent with the recent high-quality studies, such as those by Papageorghiou et al. [13], Sahota et al. [10] and Verburg et al. [12]. The prediction differences were + 0.14, + 0.28 and − 0.26 days, respectively. However, the results of this study had relatively large differences from some relatively old papers, such as those of Robinson et al. [7] and Hadlock et al. [6], which might be related to the poor resolution of the instruments used in their studies. We believe that there is no clinically significant difference in CRL-based GA formulae derived from Chinese and non-Chinese fetuses [10, 14]. At the same time, this suggests that fetal growth and development are similar between different populations when methodological standards are high and appropriate selection is made.
Strengths and limitations
It should be noted that one disadvantage of estimating GA by measuring fetal CRL alone using ultrasound is the unknown biological variation of this measure during the first trimester of pregnancy. Therefore, we recommend collecting all information (including LMP and assessing its reliability) from pregnant women at their first visit in the first trimester [22]. When the GA calculated by measuring CRL by ultrasound and that from LMP are basically consistent, the GA can be calculated according to the date of the LMP. However, when the timing of the LMP is very accurate and reliable, and there is a big difference from the GA calculated by CRL, clinics should suspect that the fetus might have pathological abnormalities causing disorders of growth and development, which need to be further monitored and diagnosed [23, 24]. At the same time, calculated variabilities in the GA, such as SD values or percentiles, should be explained to pregnant women as estimated prediction error.