Standards for Fetal Abdominal Circumference and Estimated Fetal Weight in Shanghai, China

Introduction: While current evidence suggests that fetal skeletal parameters do not vary considerably across populations, there is uncertainty about the standards of abdominal circumference (AC) and estimated fetal weight (EFW), which are the basis of clinical guidelines for the management of fetal growth restriction. The use of different algorithms for mapping ultrasound parameters to fetal size parameters also make it difficult to make local applications of growth standards developed elsewhere. Therefore, we aimed to construct standards of AC and EFW for use in China. Methods: This is an analysis of pregnant women who received antenatal care from 2009 to 2013 in Shanghai, China. A healthy cohort of 2529 singleton pregnancies defined by maternal, pregnancy and fetal characteristics and birth outcomes were selected for analysis. Ultrasound measurement was used to estimate gestational age and the Hadlock algorithm was used to calculate EFW. Results: A linear increase of AC was observed from 16 to 35 weeks, after which the increase slightly flattened out. An accelerating increase of EFW was observed from 16 to 27 weeks, after which the increase became linear and then flattened out around 36 weeks. Model diagnostics showed sufficient goodness-of-fit of the standards. Conclusion: Standards of AC and EFW are available to monitor fetal growth for pregnancy in China. Comparability with other cohorts cannot be conclusive as different analysis methods were used across studies.


Introduction
Ultrasound measurement of fetal biometry is an important and commonly used tool in antenatal care. It measures biometric parameters such as biparietal diameter (BPD), head circumference (HC), abdominal circumference (AC) and femur length (FL). Estimated fetal weight (EFW) can be calculated based on multiple biometric parameters using established formulae [1]. It is important not only to measure fetal biometric parameters accurately, but also to interpret the results appropriately. It helps to predict or diagnose fetal health problems.
Fetal growth restriction (FGR), conventionally defined as estimated fetal weight that is less than the 10 th percentile for gestational age and small-for-gestational-age (SGA), conventionally defined as birth weight less than the 10 th percentile for gestational age, are conditions that have been associated with increased risk of perinatal morbidity and mortality and increased metabolic and cardiovascular disease risk in later life [2][3][4][5]. The reduction of such risk depends on improved antenatal detection and careful clinical management [6]. Various studies have demonstrated that AC and EFW are preferred parameters to diagnose FGR or predict SGA [7,8]. Below 10 th percentiles of AC or EFW standards has also been recommended as diagnostic criteria by recent national clinical guidelines [2,6,9]. the past two decades in China. To assist clinical management of FGR and SGA in mainland China, growth standards of AC and EFW using most recent data are needed.
The aim of this study is to construct standards for AC and EFW for the Chinese population in mainland China, using ultrasound observations obtained in Shanghai, China, during 2009-2013.

Materials and Methods
Data for analysis were extracted from a routine clinical and ultrasound database for standard health care procedures of pregnant women who had their antenatal visits at the maternal unit of International Peace Maternity and Child Health Hospital, Shanghai, China from 2009 to 2013 [18]. Patient identifiers were removed from the dataset before analysis and ethical approval for the analysis was granted by the hospital's Institutional Review Board.
Women were advised to have ultrasound scans at gestational ages of 16 th week, 24 th week, 34 th week and 38 th week according to standard schedule of antenatal care at Shanghai, China. Crown-rump length (CRL), biparietal diameter (BPD), femur length (FL), anteriorposterior trunk diameter (APTD) and transverse trunk diameter (TTD) were measured in the antenatal ultrasound scans according to a protocol widely used in China [19]. Briefly, CRL was measured based on frozen image taken on stretched body at first antenatal visit for gestational age estimation. The maximum distance between crown and rump were recorded. AC was based on measurements of APTD and TTD and computed by the formula (APTD+TTD)*1.57 [20]. Head circumference was measured using the ellipse facility on the outer border of the skull. Femur length was measured using a longitudinal view of the fetal thigh closest to the probe and with the femur as close as possible to the horizontal plane. Fetal ultrasonography was performed using a GE Voluson E8 color Doppler ultrasonic diagnostic apparatus (GE Ultrasound, USA), equipped with a 3.5 MHz transabdominal transducer.
For the purpose of developing growth standards, a 'healthy' cohort of women with singleton pregnancy who conceived naturally was selected for analysis. Mothers or pregnancies with the following characteristics were excluded: Maternal age <18 or >35 years, selfreported pre-pregnancy BMI<17.0 kg/m 2 or >27.0 kg/m 2 at their first antenatal visits, cesarean section delivery due to clinical indication (not due to lifestyle choice), previous miscarriage and diagnosis of preeclampsia, pregnancy-induced hypertension, diabetes and gestational diabetes. In addition, fetuses with the following characteristics or birth outcomes were also excluded: Birth weight <2.3 kg or >4.5 kg, delivery before 37.0 weeks or after 42.0 weeks, still birth, fetal death and/ or congenital fetal anomaly indicated by ultrasound scans. Birth weight of 2.3 kg was chosen as the cut-off because 2.3 kg was the first percentile of birth weight in healthy newborns according to the WHO Multicentre Growth Reference Study Group [21].
Gestational age (GA, in exact weeks) at participants' first antenatal ultrasound visit was estimated by CRL using formula . . + = if CRL was available and not larger than 84 mm [22,23]. Otherwise, GA at first antenatal ultrasound visit was estimated by a combination of BPD, AC and FL using formula .
AC was estimated by APTD and TTD using formula .
Ultrasound biometric data from the first antenatal visits were used only for GA estimation, not for estimation of fetal growth trajectories.

Statistical methods for development of growth standards
Since there are multiple measurements per pregnancy, mixed effect linear regression models for repeated measurements were used to develop the standards [25,26]. In our study, women had different number of antenatal ultrasound visits. The approach of mixed effect models weights data from women with more ultrasound measurements less heavily. This prevents statistical parameters from gravitating towards women with more ultrasound measurements. A Box-Cox transformation was applied to account for non-normal distribution in biometric measurements (AC and EFW), with adjustment for a cubic function of GA and indicator variables for women [24]. Fractional polynomials was applied to find a power function of GA such that a linear relationship between the function of GA and the transformed biometric measurements could be formed. A linear mixed effect model of the transformed biometric measurements was then generated to estimate parameters that were needed for z-scores and percentiles (5th, 10th, 50th, 90th and 95th percentiles), as described by Royston [25].
Detrended Q-Q plot, also known as worm plot, was created for checking the normality of z-scores [27]. Proportions of observed AC and EFW below the 10 th or above the 90 th percentiles across all gestational ages were calculated to assess if they did cover 10% of the observations as intended, with 95% confidence interval (CI) obtained by generalized linear models with an identity link function and binary distribution. Logistic regression was used to check whether the proportions below (or above) the specified percentiles were independent of gestational ages as intended. Huber-white robust standard error estimator was used in the analyses to adjust for correlated data (multiple ultrasound scans per woman). Statistical analyses were performed using the R 3.1.2.

Results
Out of a total of 11463 women with singleton pregnancies in the database, 2529 women qualified for the healthy cohort for the purpose of the analysis. Excluding the ultrasound data from the GA dating visit, there were 6837 and 6832 ultrasound measurements of AC and EFW for the development of the standards, respectively. Five criteria that accounted for 70% of the exclusion were maternal age <18 or >35 years old, self-reported pre-pregnancy BMI<17.0 or >27.0, delivery before 37.0 weeks or after 42.0 weeks, cesarean section delivery due to clinical indications and miscarriage in the past.

Statistical modelling of repeated measurements
The Box-Cox transformation gave estimates of power transformation parameter  (Table 1).

Fetal growth charts
The standards for AC and EFW are shown in Figures 1 and 2 in terms of percentiles. For AC, an approximately linear increase was observed Step Parameter Value  Differences between percentiles of EFW increased rapidly over time. At 20 th , 24 th , 36 th and 40 th weeks, median EFW was around 308 g, 619 g, 2511 g and 3167 g, respectively (Figures 1 and 2).

Model diagnostics
Detrended Q-Q plots for standard of AC showed that most of the data were around the horizontal line and within the 95% CI, indicating approximate normal distribution of the z-scores. The proportions of observed AC classified below the 10 th percentile or above the 90 th percentiles of the standard were 10.7% (95% CI: 9.8 to 11.6%) and 10.0% (95% CI: 9.1 to 10.8%) as intended, respectively. Logistic regression assessing the association between GA and the classification of AC below the 10 th percentile gave p-value of 0.658. That between GA and AC above the 90 th percentile was 0.041. Despite the statistically significant association between GA and AC above the 90 th percentile, the odds ratio was only 0.987 per week increase in GA and it induced practically little deviation from the targeted 10% coverage (Figure 3).
Detrended Q-Q plots for standard of EFW showed that most of the data points were around the horizontal line and within the 95% CI, except in one of the 16 GA intervals. The z-scores mostly were normally distributed. The proportions of observed EFW classified below the 10 th percentile or above the 90 th percentiles of the standard were 10.2% (95% CI: 9.3 to 11.2%) and 9.2% (95% CI: 8.4 to 10.0%), respectively. Logistic regression assessing the association between GA and the classifications of EFW below the 10 th percentile and above the 90 th percentile gave p-values of 0.409 and 0.723, respectively (Figure 4). Tables 2 and 3 compare the median AC and EFW at different gestational weeks between our study and others. In terms of AC, the previous study conducted in central-south China and the Singaporean Chinese study published in 1995 used the same algorithm as our study to derive AC from APTD and TTD [11,17]. Throughout 20 th to 40 th weeks, both studies had medians similar to ours. The recently published Singapore Chinese study and a Caucasian study in London, UK used    direct AC measurements and they shared similar medians [12,28]. The medians in our study tended to be lower than theirs (Table 2).

Comparison with other growth standards
In terms of EFW (Table 3), our study used the Hadlock (I) formula that only involves AC, FL and BPD to derive EFW, as HC was not measured in our study [1]. A Congolian study, the Singapore Chinese study and a UK study used the Hadlock formula involving AC, FL, BPD and HC (Hadlock (II)) instead [12,28,29]. Esinler et al. compared different EFW algorithms and suggested that Hadlock (I) and Hadlock (II) had good and similar degree of accuracy [30]. Our study, the *The plot consists of detrended Q-Q plots for 16 subgroups of Z scores for AC with equal size ranked by earliest to latest GA from upper-left panel to lower-right panel.  Singapore Chinese study and the Congolian study showed similar EFW at 20 and 24 weeks, but at 36 weeks our study showed lower EFW. The UK study showed higher median of EFW throughout 20 th to 40 th weeks. The previously mentioned central-south Chinese study and Singapore Chinese study who shared similar medians of AC with our study did not provide standards for EFW, so we were unable to compare EFW with these two Chinese studies.

Discussion
We have developed standards for both AC and EFW for singleton pregnancies in Shanghai, China. By using only data from a 'healthy cohort' , the standards describe the growth trajectories for healthy pregnancy and birth outcomes. Model diagnostics show sufficient goodness-of-fit. Considering the agreement between the median AC curves from our study and the previous study conducted in centralsouth China, we believe that the standards from our study is widely usable in China.
In our study, AC was not directly measured, it was based on measurements of APTD and TTD. Median AC in this study tended to be lower than the Caucasian study in London, UK, where AC was directly measured [28]. This difference may be partly attributed to the difference between AC algorithms as direct measurement method of AC tends to result in values that are greater than the indirectly estimated AC [31][32][33]. AC derived from APTD and TTD is the current common practice in China and growth standards using this algorithm will provide clinical guidance to health care professionals in China. However, in places where AC is directly measured, caution must be taken when using the growth standards developed in this analysis.
Reasons of Chinese fetus having smaller EFW in the third trimester than other cohorts need to be further studied. Given that AC was similar between this study and the earlier Singapore Chinese study, the present population may not be truly much smaller in EFW than other cohorts. An alternative explanation could be the impact of different EFW algorithms. HC was not measured in this study. Therefore, we did not use the Hadlock formula involving HC which was applied in the studies in Congo, Singapore and UK. It may be that the algorithm used in our study underestimated EFW or the algorithm used in other study overestimated EFW in the third trimester. This is a speculation and further studies need to be done before any firm conclusion is made.
A strength of our study is that, even after stringent selection to define a healthy cohort for the purpose of developing fetal growth standards, there was still a sufficient sample size of 2529 pregnancies/fetuses included in our statistical analysis. The pregnancies/fetuses included were not supposed to represent the excluded pregnancies/fetuses. Instead, they were chosen to represent a healthy pregnancy cohort, which is commonly used in the development of growth standards. The large sample size offers sufficient precision for the development of the standards.

Conclusion
In summary, we have developed growth standards for clinical use to monitor fetal AC and EFW for Chinese population in China. Compared to standards developed elsewhere, our standards indicate that Chinese fetus in Shanghai, China, had similar AC compared to other Singapore and Chinese cohorts. Variation in EFW across populations cannot be conclusively determined due to differences in EFW estimation algorithms.