Knowledge of gestational age is needed for optimal obstetric care, including evaluation of fetal growth and management of preterm and postterm pregnancies.1,2 The first day of the last menstrual period (LMP) has been the basis for calculating the day of delivery, which is supposed to occur 280 days later. However, an epidemiologic study of more than 400,000 singleton births found that the mode and median of the time between the first day of a certain LMP and the day of spontaneous delivery might be as long as 282 days.3 Another study confirmed that.4
Measurement of the biparietal diameter (BPD) during routine second-trimester scanning was better than LMP for predicting the day of delivery.4–6 However, more pregnancies are examined during the first trimester, especially to screen for nuchal translucency7,8 and to search fetal structural defects.9 Although the measurements of crown-rump length, biparietal diameter (BPD), and femur length (FL) are feasible at 11–16 weeks, their accuracy in predicting the day of delivery has been investigated infrequently.10
The aim of our study was to estimate which ultrasound measurement, crown-rump length, BPD, or FL, is best for predicting the day of delivery at 8–16 weeks' gestation. We also compared them to prediction by certain and uncertain LMP.
Materials and Methods
During 1993–1998, all pregnant women residing in the catchment area of Jorvi hospital (western part of greater Helsinki) were invited to have transvaginal sonographic screening at 13–14 completed weeks of gestation.7 The invitation was presented during the woman's first antenatal visit. A total of 22,077 pregnant women were examined.
Exclusion criteria were crown-rump length less than 15 mm (less than 8 weeks) (n = 24), BPD over 36 mm (more than 16 weeks 6 days) (n = 276), multiple gestation (n = 376), nonviable pregnancy (n = 91), and fetal malformation (n = 240). Women who had elective cesarean (n = 1313) or induction of labor (n = 2536) also were excluded, leaving 17,221 singleton pregnancies with spontaneous onset of labor as the study population.
Last menstrual period was reported as “certain” by 13,541 (78.6%) women and “uncertain” by 3680 (21.4%); LMP was considered uncertain if the woman could not confirm it with certainty, or her menstrual period was on average beyond 24–32 days, or when her pregnancy ensued less than 4 months after stopping hormonal contraception. Women with no menstrual history were included in the uncertain group; the approximate duration of pregnancy for them was estimated by uterine palpation or results of pregnancy tests. The day of delivery was estimated according to Naegele's rule (first day of LMP + 280 days) and modified Naegele's rule (first day of LMP + 282 days).
Sonograms were done using Hitachi equipment with 6.5-MHz transvaginal and 5.0-MHz transabdominal transducers (model EUB-415; Hitachi Medical Corporation, Tokyo, Japan). Transvaginal sonography was done in 99% of women. In 11%, transabdominal sonography also was done to the improve visualization of the upper fetal structures. In clinical practice, the assessment of gestational age by ultrasound was based on crown-rump length up to 11 weeks 6 days (BPD under 20 mm) and on BPD or FL for older fetuses.11–13 Among the 13,541 pregnancies with certain LMP, all three measurements (crown-rump length, BPD, and FL) were available in 11,678. Both crown-rump length and BPD were available in 12,042; in the remaining 1499, either crown-rump length or BPD was available together with FL.
Before 11 weeks' gestation the number of measurements of BPD or FL was too small (under 100) for calculations. The data for most regression analyses were limited to at least 37 weeks to correct for the otherwise irreparable skewness caused by premature deliveries. Another reason to exclude premature deliveries from the regression analyses was our general intent to predict births with spontaneous onset; premature deliveries can occur because of a triggering agent, such as infection. The deliveries were classified according to World Health Organization (WHO) standards as term, preterm (before 259 days), or postterm (at or later than 294 days).
The analyses were done using SAS (SAS Institute Inc., Cary, NC) and NCSS 2000 (NCSS Inc., Kaysville, UT). Duration of pregnancy from screening to spontaneous delivery was predicted by univariate linear regression models for crown-rump length, BPD, and FL. Nonlinear models and multiple regression using two or all three of the available ultrasonic measurements also were tested. Residuals (differences between predicted and observed values) were considered as indicators of prediction accuracy. The accuracy of the models was compared with that based on LMP to assess the best predictor for day of delivery.
The differences in proportions were tested with exact probability tests or with the χ2 test. Normally distributed continuous variables were analyzed by analysis of variance. The chosen level of statistical significance was .05. All tests were two-sided.
The mean gestational age at screening was 13.5 weeks (range 8–16 weeks 6 days; Table 1). Of the 17,221 women, 9566 (55.5%) gave birth within 7 days of the day estimated by the LMP, 11,809 (68.6%) within 10 days, and 13,507 (78.4%) within 14 days. Of the deliveries, 760 (4.4%) were preterm and 470 (2.7%) postterm, as estimated by ultrasonography. The mean duration of pregnancy was 278.6 days with standard deviation (SD) of 12.5. The median was 281 days, and the mode 283 days, as estimated by ultrasonography. In pregnancies with certain LMP (n = 13,541), the mean duration of pregnancy estimated from ultrasonography was 279.7 days, the median 281 days, and the mode 282 days, with a mean prediction error of 10.2 days.
Table 2 shows the prediction errors for ultrasonography and certain LMP data throughout 8–16 weeks in pregnancies of at least 37 weeks' duration. Using univariate linear regression, crown-rump length, BPD, and FL showed a similar average prediction error of 7.4–7.7 days, whereas the error by LMP data reached 10 days. The proportion of variation explained by the univariate regression models (R2) amounted to 34–39%. When premature deliveries were included, prediction errors by ultrasonography deteriorated to 9.6–9.9 days, and the R2 values to 26–29%, whereas the error by LMP increased to 11.7 days.
The data were tentatively split in half by random numbers (0 or 1). Group 0 was first used to create the regression models, which were then used for actual predictions for group 1. The test was repeated with the groups reversed. The resulting intercepts and regression coefficients as well as the average prediction errors and the R2 values did not differ significantly from those shown in Table 2 for the whole data set.
A linear model appeared sufficient for our data, because there was no significant increment in R2 when a second-degree polynomial model was applied. The prediction accuracy did not improve when any two or all three of the available ultrasonographic variables were included as predictors in multivariate regression analysis.
Figure 1 depicts the average errors of predicting the day of delivery as a function of gestational age at screening. Crown-rump length was the best predictor at 11–12.5 weeks (corresponding to at most 60 mm), whereas BPD was superior at 13 weeks and thereafter (at least 21 mm). Femur length was almost as good as crown-rump length at 11–12 weeks and almost as good as BPD at 13–16 weeks. The estimate of the day of delivery by certain LMP was always less accurate than any of the three sonographic variables, with a prediction error of 9–11 days. Compared with the best ultrasound predictor, LMP was about 1.7 days less precise at 12–14 weeks and 3.5 days less precise at 11–16 weeks (Figure 1). An obvious explanation for the upward swing of the LMP curve in both directions is that the women were invited to screening at 13–14 weeks as estimated by menstrual history only. The more that gestational age by ultrasonography differed from that by LMP, the greater was the prediction error by LMP.
For women with certain LMP, the ultrasound estimate of the day of delivery (by crown-rump length when 15–60 mm or by BPD when 21 mm or greater) was later than LMP + 280 in 6988 women (58.0%), earlier in 3898 (32.4%), and equal (the same day) in 1156 (9.6%). Compared with LMP + 282, the estimate by ultrasonography was later in 4785 women (39.7%), earlier in 6207 (51.5%), and equal in 1050 (8.7%).
Using LMP + 280, ultrasound was closer to the actual day of birth in 6016 cases (55.3%), whereas LMP was closer in 4029 (37.0%) (P < .001), and they were equally close in 841 (7.7%). If LMP + 282 was used, ultrasound was closer in 6004 cases (54.6%), LMP was closer in 4138 cases (37.6%) (P < .001), and they were equal in 850 (7.7%). As the difference in estimated gestational age between the two methods increased in either direction, ultrasound gave a progressively better estimate, as shown in Figure 2.
The numbers of preterm and postterm pregnancies in different groups are given in Table 3. When LMP was used to estimate the day of delivery, the proportion of postterm pregnancies was 9.1–10.3%, regardless of whether the LMP had been reported as certain or uncertain. The number of postterm pregnancies was slightly less (6.8%) when LMP + 282 was used. In contrast, using ultrasound (crown-rump length or BPD measurements) the proportion of postterm pregnancies was only 2.7% (P < .001). The percentages of preterm deliveries were similar when predicted by ultrasonography or by certain LMP, but there were more preterm births in women with uncertain LMP (Table 3).
In this study based on a large nonselected population, we found that ultrasonic prediction of the day of delivery between 11 and 16 weeks' gestation was 1.7–3.5 days more accurate than the use of LMP, depending on gestational age at screening. It has been documented that crown-rump length measurement has smaller error than BPD before 12 weeks in predicting the day of delivery,14 but for how long crown-rump length stays better has not been studied. We found that crown-rump length measurement of 15–60 mm (8 weeks 0 days to 12 weeks 4 days) can be considered the best determinant, whereas BPD (at least 21 mm) is more precise thereafter. We found, like other investigators,15,16 that FL also is a good gestational age determinant between 11 and 16 weeks' gestation.
In predicting the day of spontaneous delivery, simple linear regression models for crown-rump length, BPD, and FL were sufficient for our data on a relatively narrow band of gestational ages. Our equations cannot, however, be extrapolated beyond 8–16 weeks, when nonlinear prediction models might be needed. We did not find improved accuracy when any two or all three ultrasonic variables (crown-rump length, BPD, and FL) were included in the prediction models. The clinical implication is that it is not efficient to combine information from more than one adequately obtained ultrasonic measurement in estimating the day of the delivery.
It has been suggested that the prediction error between the estimated day of delivery and the actual day of delivery should be 7 to 8 days.14 In our study, the prediction error at 11–16 weeks' gestation was 7.3 days for the best ultrasound determinant, when premature deliveries were excluded. The corresponding error from LMP at 12–15 weeks was 9–9.6 days. When premature deliveries were included in the calculations, the prediction errors for both ultrasonography and LMP increased by about 2 days.
The physiologic variation in the natural duration of pregnancy makes it impossible to predict the day of delivery beyond a certain limit, no matter how accurately the gestational age is known. The prediction error therefore always includes variation in the time of onset of labor and methodologic error. Differentiation between these components is difficult, if not impossible, unless the day of conception is known with certainty.
We found that ultrasound was superior to LMP by at least 1.7 days in predicting the day of delivery, even when the difference between the methods was small (7 days or less), and that the greater the discrepancy, the less accurate was LMP, which is in accordance with other studies based on BPD in the second trimester.3,6,17 We examined the pregnancies earlier, which might further improve the accuracy of ultrasonographic dating compared with LMP.
It has been claimed that a modification (LMP + 282) of Naegele's rule (LMP + 280) might give a prediction of term, which is as accurate as that using BPD.18 Neither our results nor others support this.3 Using 282 days from LMP instead of 280, we did not note improvement in the accuracy of the LMP estimation compared with ultrasound.
The mean and median duration of pregnancies calculated by certain LMP were in accordance with those of other studies.3,6,19,20 In a Danish study with similar number of women with certain LMP, the mean and median duration calculated by BPD at 12–22 weeks' gestation were 278 and 280 days, respectively, with a prediction error of 11.8 days.3 In this study, we found a slightly longer mean duration (279.7 and 281 days, respectively) and a smaller prediction error of 9.6–9.9 days. The smaller prediction error in our study suggests that dating pregnancies by ultrasound is more accurate at 8–16 weeks than at 12–22 weeks. The prediction error in our series was lowest at 12–14 weeks' gestation.
Although our main indication for screening at 13–14 weeks was to identify fetuses with increased nuchal translucency,7 the opportunity to obtain fetal ultrasonic dimensions emerged as a useful adjunct. However, it might not be worth switching from midtrimester screening to 13-week screening for slightly better precision in dating.
Another improvement obtained by early ultrasound was that the number of postterm pregnancies was smaller, 2.7%, compared with 3.7% observed at 12–22 weeks.3 Accurate knowledge of gestational age is particularly important for proper identification of postterm deliveries, which often are monitored intensively for the risk of placental insufficiency, growth restriction, fetal distress, and intrauterine death. Monitoring an unnecessarily large and imprecisely defined risk group is costly and stressful for the staff and the women themselves. In this study, the proportion of postterm deliveries decreased from 10.3% estimated by LMP to 2.7% estimated by ultrasonography, even in women with certain LMP. Other studies have shown that ultrasound dating reduced the number of postterm pregnancies.3,6
As in other studies,3,6 we found no increase in the number of preterm deliveries when ultrasound was used instead of LMP to determine gestational age. There have been concerns that dating strictly by ultrasound only could lead to failure to diagnose early growth restriction. This did not seem to hold true.21 It has also been demonstrated that the perinatal mortality rate did not change when pregnancies were dated by ultrasound only.22 We do not see any reason to use LMP for dating when adequate ultrasonic data are available by midpregnancy.
1. Campbell S, Wilkin D. Ultrasonic measurement of fetal abdomen circumference in the estimation of fetal weight. Br J Obstet Gynaecol 1975;82:689–97.
2. Saari-Kemppainen A, Karjalainen O, Ylostalo P, Heinonen OP. Ultrasound screening and perinatal mortality: Controlled trial of systematic one-stage screening in pregnancy. The Helsinki Ultrasound Trial. Lancet 1990;336:387–91.
3. Bergsø P, Denman DW, Hoffman HJ, Meirik O. Duration of human singleton pregnancy. A population-based study. Acta Obstet Gynaecol Scand 1990;69:197–207.
4. Nguyen TH, Larsen T, Engholm G, Møller H. Evaluation of ultrasound-estimated data of delivery in 17 450 spontaneous singleton births: Do we need to modify Naegele's rule? Ultrasound Obstet Gynecol 1999;14:23–8.
5. Persson PH, Kullander S. Long-term experience of general ultrasound screening in pregnancy. Am J Obstet Gynecol 1983;146:942–7.
6. Tunon K, Eik-Nes SH, Grøttum P. A comparison between ultrasound and a reliable last menstrual period as predictors of the day of delivery in 15,000 examinations. Ultrasound Obstet Gynecol 1996;8:178–85.
7. Taipale P, Hiilesmaa V, Salonen R, Ylostalo P. Increased nuchal translucency as a marker for fetal chromosomal defects. N Engl J Med 1997;337:1654–8.
8. Snijders RJM, Noble P, Sebire N, Souka A, Nicolaides KH. UK multicentre project on assessment of risk of trisomy 21 by maternal age and fetal nuchal-translucency thickness at 10–14 weeks of gestation. Lancet 1998;351:343–6.
9. Whitlow BJ, Chatzipapas IK, Lazanakis ML, Kadir RA, Economides DL. The value of sonography in early pregnancy for the detection of fetal abnormalities in an unselected population. Br J Obstet Gynaecol 1999;106:929–36.
10. Campbell S, Warsof SL, Little D, Cooper DJ. Routine ultrasound screening for the prediction of gestational age. Obstet Gynecol 1985;65:613–20.
11. MacGregor SN, Tamura RK, Sabbagha RE, Minogue JP, Gibson ME, Hoffman DI. Underestimation of gestational age by conventional crown-rump length dating curves. Obstet Gynecol 1987;70:344–8.
12. Daya S. Accuracy of gestational age estimation by means of fetal crown-rump length measurement. Am J Obstet Gynecol 1993;168:903–8.
13. Sabbagha RE. Gestational age. In: Sabbagha RE, ed. Diagnostic ultrasound applied to obstetrics and gynecology. 3rd ed. Philadelphia: J.B. Lippincott, 1994:155–78.
14. Persson PH. Ultrasound dating of pregnancy—Still controversial? Ultrasound Obstet Gynecol 1999;14:9–11.
15. Benson CB, Doubilet PM. Sonographic prediction of gestational age: Accuracy of second- and third-trimester fetal measurements. AJR Am J Roentgenol 1991;157:1275–7.
16. Hadlock FP, Harrist RB, Deter RL, Park SK. A prospective evaluation of fetal femur length as a predictor of gestational age. J Ultrasound Med 1983;2:111–2.
17. Gardosi J. Dating of pregnancy: Time to forget the last menstrual period. Ultrasound Obstet Gynecol 1997;9:367–8.
18. Olsen O, Aaroe CJ. Routine ultrasound dating has not been shown to be more accurate than the calendar method. Br J Obstet Gynaecol 1997;104:1221–2.
19. Kieler H, Axelsson O, Nilsson S, Waldenstrom U. The length of human pregnancy as calculated by ultrasonographic measurement of the fetal biparietal diameter. Ultrasound Obstet Gynecol 1995; 6:353–77.
20. Mongelli M, Wilcox M, Gardosi J. Estimating the date of confinement: Ultrasonographic biometry versus certain menstrual dates. Am J Obstet Gynecol 1996;174:278–81.
21. Tunon K, Eik-Nes SH, Grøttum P. Fetal outcome when the ultrasound estimate of the day of delivery is more than 14 days later than the last menstrual period estimate. Ultrasound Obstet Gynecol 1999;14:17–22.
22. Tunon K, Eik-Nes SH, Grøttum P. Fetal outcome in pregnancies defined as post-term according to the last menstrual period estimate, but not according to the ultrasound estimate. Ultrasound Obstet Gynecol 1999;14:12–6.