16th World Congress on Ultrasound in Obstetrics and Gynecology Poster abstracts 24–30 day cycle. The influence of fetal gender, maternal smoking and maternal height on gestational age (GA) at the time of CRL mea- surement was assessed using chi-square test. 4,595 deliveries were spontaneous and 1,299 induced (missing information in 31 cases). The relative risk (RR) of preterm delivery (≤ 258 days) using LMP and CRL was estimated for girls compared with boys in women with a spontaneous delivery. Results: The table shows the distribution of the difference between GA by LMP and GA by CRL on fetal gender, maternal smoking and maternal height. In non-smoking women the RR of preterm delivery was unchanged in girls compared with boys when comparing CRL (RR 0.81) to LMP (RR 0.79). In smoking women, however, the RR of preterm delivery increased from 0.50 by LMP to 0.80 by CRL. In tall women the RR increased from 0.38 using LMP to 0.54 using CRL, while the RR was unchanged in women < 173 cm. Conclusions: Fetal gender, maternal smoking and maternal height significantly influence gestational dating by CRL. Using CRL will overestimate GA in tall non-smoking women with a male fetus and underestimate GA in short, smoking women with a female fetus. The effect of the three parameters is additive. P13.08 Comparison of reference charts for estimated fetal weight and for actual birth weight L. J. Salomon , J. P. Bernard, Y. Ville Universit´ e Versailles Saint-Quentin, Service de Gyn´ ecologie-Obst´ etrique. Centre Hospitalier Intercommunal de Poissy St Germain en Laye, France Objectives: To build and compare reference charts and equations for estimated fetal weight (EFW) and for birthweight (BW) based on a large sample of fetuses and neonates of a single health authority. Materials and Methods: Biometric data were obtained at 15 to 36 weeks from routine screening examination over 3 years. Exclusion criteria were a known abnormal karyotype or congenital malformation and multiple pregnancies. No data were excluded on the basis of abnormal biometry. EFW was calculated based on Hadlock’s formula. We used a non parametric approach (LMS method) to compute new reference charts for EFW. This chart was compared to that of BW at between 24 and 36 weeks of gestation during the same period. Results: 19,015 fetuses were included. New charts and equations for Z-scores calculations are reported based on LMS methods. Comparison with the chart built on 58,942 singletons born during the same period in the same health authority showed that EFW chart was above that of BW at 24 to 36 weeks’. Differences were up to 350 grams or 20% of EFW. At between 27 and 33 weeks, the 50th percentile for BW grossly compared to the 10th centile for EFW. Conclusion: We present new reference charts and equations for EFW. EFW is computed throughout gestation based on measurements in healthy fetuses. However, before full term, BW charts reflect a significant proportion of growth retarded fetuses that were born prematurely. Reporting EFW on BW charts is therefore misleading. P13.09 Customized fetal weight estimation at term S. Fiore 1 , M. Cortina-Borja 2 , F. Severi 3 , E. Ferrazzi 1 1 DSC Sacco University of Milan, Italy, 2 Institute of Child Health, University College London, United Kingdom, 3 COG University, Siena, Italy Objective: To improve ultrasonography- base prediction of fetal weight at term, accounting for routinely measurable maternal and pregnancy-specific variables. Methods: Data on 1208 single pregnancies were prospectively col- lected; to estimate term birth weight (BW). 594 uncomplicated pregnancies were analysed. (gestational age (GA) at delivery between 37 and 42 weeks) with complete information on: Fetal Abdominal Circumference, Head Circumference, Femur length, neonatal Gender and BW and GA at delivery; Maternal Body mass Index (BMI), GA at ultrasound exam, Maternal Weigh increase during pregnancy, Maternal Age Statistical methods We fitted linear regression models using the square root to stabilise the variances and achieve Normality for BW. An interaction term was included in order to reflect possible synergy between fat/bone growth and quadratic terms for fetal vari- ables to allow for growth faster than linear in these measurements An optimal model was found using a predictive information criteria. (S-Plus 2000). Results: Baseline data included 300 female and 294 male born to Caucasian women with median BW gr 3365 (range 1930–5000); median maternal age 31 (range 16–44); maternal BMI is 26 (range 19–37); median maternal weight increase12 kg (range 4–27). We found a significant association with maternal BMI (p < 0.001) and gender (p < 0.0001). The final model is (SexNeonate is 0 = girl or 1 = boy). Conclusions: Matching the observed BW vs our prediction, and calculating the difference as a %of the observed BW, we estimated that 86.7% of the predicted BW fell between, b 10% of the observed BW. The mean absolute percent (MA %)error in BW prediction was 5.37%, remaining constant also for BW > 3500 gr. (n157) Our results compare favorably with previous published results (Table) MA % Error BW ± 10% Watson (1988) 8.2% 66% Chauhan (1992) 15.6 42% Sherman (1998) 8.1% 69% Hendrix (2000) 16.5% 32% Range (previous papers) 8.1–16.5% 32–69% Our results 5.2% 87.2% Including BW > 3500 gr 5.1% 87.6% P13.10 Impact of the modelling method on reference charts and equations for biometry L. J. Salomon 1 , J. P. Bernard 1 , B. De Stavola 2 , M. Kenward 2 , Y. Ville 1 1 Department of Obstetrics and Gynecology, Universit´ e Versailles Saint-Quentin, France, 2 Medical Statistics Unit, London School of Hygiene and Tropical Medicine, United Kingdom Objective: Biometry assessment, during gestation or at birth, is usu- ally based upon a comparison of measured values with predicted values derived form reference charts or equations. However, the impact of statistical methods used to compute such charts have not been extensively evaluated. Materials and Methods: Data were obtained from the birth registry of the Yvelines (a French Territorial division of 1.4 millions peo- ple) over 3 years. Two modelling methods were compared. The most widely used consists of polynomial regression and fit of variability as described by Chitty and Altman. Z-scores can be easily derived from the predicted means and standard deviations. The second one (LMS) was described by Cole and consists of summarising the changing distribution by three curves representing the median (M), coefficient of variation (S) and skewness (L). Using penalised likelihood the three curves can be fitted as cubic splines by non-linear regression and Z-scores for any measurement can be calculated from the L, M and S values. Results: 58 942, 56 956 and 56 909 measurements for weight, size and head circumference from 25 to 42 weeks’ respectively were used. Z-scores obtained from both methods were significantly differ- ent and this would result in different numbers of measurements to be classified as abnormal. The differences were however not clinically relevant for birth weight and birth size, whereas LMS provided better results on the skewed head circumference measurements. Therefore 590 Ultrasound in Obstetrics & Gynecology 2006; 28: 512–614