Ultrasound dating at 12–14 or 15–20 weeks of gestation? A prospective cross-validation of established dating formulae in a population of in-vitro fertilized pregnancies randomized to early or late dating scan

Authors


Abstract

Objectives

To determine the accuracy of established ultrasound dating formulae when used at 12–14 and 15–20 gestational weeks.

Methods

A total of 363 singleton pregnancies conceived after in-vitro fertilization (IVF) and randomized to a dating scan at 12–14 or 15–20 gestational weeks were studied. ‘True’ gestational age was calculated on the basis of the day of oocyte retrieval and was compared to gestational age calculated on the basis of seven dating formulae including the fetal biparietal diameter (BPD) and three dating formulae including BPD and fetal femur length (FL). The mean of the differences between estimated and true gestational age (systematic measurement error) and their SD (random measurement error) were calculated for each formula.

Results

Three formulae showed systematic errors of less than −0.7 days at both early and late scanning. Two formulae overestimated gestational age at both early and late scanning by 5.7 and 3.1 vs. 2.3 and 2.8 days, respectively, while five formulae manifested very different systematic errors at early and late scanning. The formulae used for clinical management underestimated gestational age by a mean of 3 days when dating was performed at 12–14 weeks, and by a mean of 0.8 days when dating was done at 15–20 weeks. The random error was on average 1 day less when the scan was carried out early (2 vs. 3 days; P < 0.0005). Mean true gestational age at delivery in IVF pregnancies with spontaneous start of labor was 279 days (SD 12.9); excluding preterm deliveries it was 281 days (SD 8.1).

Conclusions

Ultrasound dating formulae originally intended for use in the middle of the second trimester do not necessarily perform well when used for dating earlier in gestation. The systematic and random error of any dating formula must be assessed for the gestational age interval in which the formula is intended to be used. Copyright © 2004 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Determination of gestational age by ultrasound is a routine clinical procedure in many countries and is usually performed at 15–20 gestational weeks. However, during the past decade the use of serum screening and fetal nuchal translucency (NT) screening for Down syndrome1–3 has resulted in more pregnancies being examined and dated at 11–14 weeks.

Reliable information about gestational age is necessary to diagnose fetal growth restriction and is essential for optimal management of pre- and post-term pregnancies. Ultrasound dating in the second trimester is more accurate than the last menstrual period (LMP) for predicting delivery date4–7. Many commonly used mathematical models using measurements of the fetal biparietal diameter (BPD) and the fetal femur length (FL) were designed to calculate gestational age from 14 gestational weeks8, 9. Dating formulae and charts based on BPD measurements in earlier gestation have been designed, some of them being based on transvaginal measurements10, 11 and others on transabdominal measurements4, 12. Campbell and coworkers showed that prediction of delivery date based on BPD measurement performed at 12–15 weeks resulted in a higher proportion of deliveries within ± 1 week of the predicted date than predictions performed at later gestation4. To the best of our knowledge, no other research team has prospectively validated any first- or second-trimester formula with regard to their accuracy in estimating gestational age when dating is performed in the interval between 12 and 14 weeks.

In Sweden, a randomized controlled trial, the NUPP Study, was undertaken to compare the medical, psychological and economical effects of a routine fetal scan at 12–14 weeks with those of a scan at 15–20 weeks. The aim of the present study was to determine the accuracy of different dating formulae in estimating gestational age when dating was performed at 12–14 and 15–20 weeks. For this purpose, 10 published dating formulae based on BPD or BPD/FL were applied to a population of singleton pregnancies conceived after in-vitro fertilization (IVF). In addition, we analyzed the effects of using those formulae that had the best performance in our IVF pregnancies in a larger population comprising spontaneously conceived pregnancies.

Methods

Nupp Study

The NUPP Study is a multicenter, randomized controlled trial involving the ultrasound units at eight major Swedish hospitals (Huddinge University Hospital, UltraGyn/Danderyd Hospital, South Stockholm General Hospital, Karolinska Hospital, Södertälje Hospital, Malmö University Hospital, Helsingborg Hospital and Uppsala Academic Hospital). The trial was approved by the ethics committees at the Karolinska Institute in Stockholm and at the medical faculties of Lund University and Uppsala University. It enrolled 40 000 pregnant women who were randomly allocated to ultrasound examination at 12–14 or 15–20 gestational weeks. Randomization started in March 1999 and finished in October 2002. Randomization was performed using a computerized Internet-based program developed by Medscinet AB, Sweden in which also demographic data, data from ultrasound examinations and invasive tests and information on pregnancy outcome were continuously registered.

Study populations

For the purpose of comparing dating at 12–14 and 15–20 weeks all singleton IVF pregnancies included in the NUPP Study from April 1999 to May 2002 were considered for inclusion. Inclusion criteria were BPD 21–31 mm at the 12–14-week scan, BPD 32–60 mm at the 15–20-week scan and a liveborn baby with no fetal malformations or chromosomal anomalies. After exclusions our IVF population comprised 363 women, 169 randomized to a dating scan at 12–14 weeks and 194 to a dating scan at 15–20 weeks (Figure 1). The early and late scan groups were similar with regard to age (mean age 34.1, range 20–45, years in both groups) and parity (73% vs. 72% nullipara).

Figure 1.

Characteristics of pregnancies conceived after in-vitro fertilization (including details of exclusions). BPD, biparietal diameter; IVF, in-vitro fertilization.

In a second stage we analyzed the effects of using the formulae with the best performance in our IVF population for dating in a larger population of pregnancies conceived spontaneously. This population comprised 4197 women included in the NUPP Study between September 1999 and May 2000 fulfilling the following inclusion criteria: singleton pregnancy, BPD at the dating scan in the same intervals as described above for the IVF pregnancies, spontaneous onset of labor at ≥37 weeks and a liveborn baby with no fetal malformations or chromosomal anomalies. Of these 4197 women, 2070 were examined at 12–14 weeks and 2127 at 15–20 weeks. Mean age was 29.5 and 29.3 years (range, 16–43 years in both groups), and the proportion of nullipara was 48.5% and 48.2%, respectively.

Ultrasound examinations

The dating scans were scheduled to take place at 12–14 weeks (early scan group) or at 15–20 weeks (late scan group) according to the day of oocyte retrieval in IVF pregnancies and according to the LMP in spontaneously conceived pregnancies. The examinations were performed by specially trained midwives with 1–25 (mean, 11) years' experience of routine ultrasound examinations at 15–22 weeks. Before the study started all the midwives had received a certificate of competence in the theory and practice of the 11–14-week scan from The Fetal Medicine Foundation, London, UK. All examinations were performed transabdominally using any of the following ultrasound systems: Aloka 1400, Aloka SSD 2000, Aloka 4000, Aloka 5000 (Aloka Co. Ltd, Tokyo, Japan), Acuson XP 10 or Aspen (Siemens Acuson Inc., Mountain View, CA, USA) with 3.5–5-MHz curvilinear transducers. At the 12–14-week scan the examination included measurements of fetal crown–rump length (CRL), BPD, FL and NT. At the 15–20-week scan only BPD and FL were measured. Fetal anatomy was examined in both groups. The BPD was measured from the outer to the inner edge of the parietal bone in a transverse plane which aligns the cavum septi pellucidi and third ventricle with the thalami13, 14. The near field femur was measured in a straight line from one end of the diaphysis to the other15, 16. At the early scan FL was not always measured. For all parameters the mean of three replicate measurements was used.

Calculation of gestational age

In this paper, gestational age was used synonymously with menstrual age. In the IVF pregnancies gestational age at the dating scan/delivery was calculated by adding 14 days to the conceptual age, i.e. the number of days between the date of oocyte retrieval and the date of the ultrasound scan/delivery. Gestational age according to oocyte retrieval was regarded as the true gestational age. Gestational age in the IVF population was also estimated using seven ultrasound dating formulae based on BPD measurements and three formulae based on BPD and FL measurements (Table 1). For clinical management gestational age was estimated using the BPD formula of Persson and Weldner at the 12–14-week scan and the BPD/FL formula of Persson and Weldner at the 15–20-week scan8.

Table 1. Formulae used to calculate gestational age in days
ReferenceFormula
  1. The hash (#) symbol denotes the formula number where a reference contains multiple formulae. BPD, biparietal diameter; FL, femur length.

BPD formulae
 Hadlock et al. (1982)147*(6.8954 + 2.6345*BPD*0.1 + 0.008771*(BPD*0.1)3)
 Hadlock et al. (#1) (1984)97*(9.54 + 1.48*BPD*0.1 + 0.1676*(BPD*0.1)2)
 Selbing and Kjessler (1985)1714 + 44.65 + 1.07*BPD + 0.0138*BPD2
 Persson and Weldner (#1) (1986)839.1 + 2.10*BPD
 Mul et al. (#1) (1996)1844.17 + 1.99*BPD
 Mul et al. (#2) (1996)1846.56 + 1.87*BPD + 0.0013*BPD2
 Altman and Chitty (1997)127* exp (0.04557* BPD − 0.0061838* BPD* ln (BPD) + 1.985)
BPD/FL formulae
 Hadlock et al. (#2) (1984)97*(10.50 + 0.197*BPD*FL*0.01 + 0.9500*FL*0.1 + 0.7300*BPD*0.1)
 Persson and Weldner (#2) (1986)849.0 + 1.21*BPD + 1.02*FL
 Mul et al. (#3) (1996)1850.16 + 1.36*BPD + 0.78*FL

When testing the formulae in the larger pregnant population, the estimated day of delivery was at 280 completed gestational days. A delivery at ≤258 completed days was considered to be preterm and a delivery at ≥294 completed days to be post-term. Preterm delivery rate was defined as the number of pregnancies with spontaneous start of labor at ≤258 days divided by the total number of pregnancies with spontaneous start of labor. Post-term delivery rate was defined as the number of pregnancies with spontaneous start of labor at ≥294 days divided by the total number of pregnancies with spontaneous start of labor.

Statistical analysis

Measurement error was defined as the difference in days between the gestational age at the dating scan calculated on the basis of ultrasound fetometry and the true gestational age according to oocyte retrieval. The mean measurement error was compared between the early and late scan group using Student's two-sample t-test and an F-test. A t-test against zero was used to analyze if the mean measurement error was statistically significant. The mean of the measurement errors reflects the systematic measurement error (i.e. systematic under- or overestimation of gestational age), whereas the SD of the measurement errors reflects the random measurement error. The statistical significance of differences in proportions was determined using the Chi-square test or McNemar's test, where appropriate. The statistical significance of differences in gestational age at delivery was calculated using Student's two-sample test. Values of P < 0.05 were considered statistically significant. Statistical analysis was carried out using the Statistical Package for the Social Sciences (SPSS Inc, Chicago, IL, USA, 2001).

Results

Dating in the IVF population

According to oocyte retrieval, mean gestational age at the scan was 13 + 2 weeks at the early scan and 17 + 6 weeks at the late scan. The distribution of BPD at examination is shown in Figure 2. FL was measured in only 62 women in the early scan group and mean gestational age at the dating scan in this subgroup was 13 + 3 weeks.

Figure 2.

Biparietal diameter at the dating scan.

Measurement errors ranged from −9 to +12 days. The measurement errors for the formulae tested and the statistical significance of differences in measurement error between the early and the late dating group are shown in Table 2. Three formulae—all using BPD measurements only—had a small negative systematic measurement error of <0.7 days both at early and late dating (Selbing and Kjessler (1985)17, Mul et al. (#1) (1996)18, Mul et al. (#2) 1996)18), and two of these had systematic errors that did not differ significantly from zero (Selbing and Kjessler (1985)17, Mul et al. (#2) (1996)18). Two formulae resulted in substantial overestimation of gestational age in both groups (Hadlock et al. (#1) (1984)9, Altman and Chitty (1997)12). For six formulae, including one of the formulae that overestimated gestational age in both groups, the measurement error differed significantly between the 12–14 weeks and the 15–20 weeks groups, the measurement error being significantly larger in the early scan group for five of the six formulae. The formulae used for clinical management underestimated gestational age by a mean of 3 days when dating was carried out at 12–14 weeks and by a mean slightly less than 1 day when dating was performed at 15–20 weeks. The SD of the measurement error was approximately 2 days in the early scan group and 3 days in the late scan group. This difference in SD was statistically significant for all formulae (P < 0.0005).

Table 2. Measurement errors of dating formulae when applied at a dating scan at 12–14 and 15–20 gestational weeks in pregnancies conceived after in-vitro fertilization
ReferenceMeasurement errorP*
12–14-week scan15–20-week scan
MeanSDRangeMeanSDRange
  • *

    P values refer to comparison of means.

  • This formula was used clinically at dating scans at 12–14 weeks.

  • This formula was used clinically at dating scans at 15–20 weeks. The hash (#) symbol denotes the formula number where a reference contains multiple formulae. BPD, biparietal diameter; FL, femur length.

BPD formulae(n = 169)  (n = 194)   
 Hadlock et al. (1982)140.82.06−5.3 to 5.32.13.18−6.4 to 11.80.0005
 Hadlock et al. (#1) (1984)95.71.78 0.2 to 10.03.13.01−4.8 to 12.10.0005
 Selbing and Kjessler (1985)17−0.41.87−6.1 to 4.0−0.23.25−8.9 to 10.20.51
 Persson and Weldner (#1) (1986)8−3.12.22−9.2 to 8.6−0.83.12−9.2 to 8.60.0005
 Mul et al. (#1) (1996)18−0.72.10−7.0 to 3.9−0.23.02−8.2 to 8.70.10
 Mul et al. (#2) (1996)18−0.42.04−6.7 to 4.1−0.63.00−8.5 to 8.30.62
 Altman and Chitty (1997)122.31.98−3.8 to 6.82.83.13−5.6 to 12.30.07
BPD/FL formulae(n = 62)  (n = 194)   
 Hadlock et al. (#2) (1984)93.81.69−0.3 to 8.40.62.71−6.9 to 8.80.0005
 Persson and Weldner (#2) (1986)8−3.32.10−7.8 to 1.5−0.92.83−8.5 to 8.00.0005
 Mul et al. (#3) (1996)18−1.21.99−5.2 to 3.10.22.80−7.0 to 9.00.0005

Pregnancy duration according to oocyte retrieval and ultrasound estimation in the IVF population

Pregnancy duration in the IVF pregnancies is presented in Table 3 and pre- and post-term rates in Table 4. There was no difference in mean pregnancy duration at delivery between the early and the late scan group irrespective of whether gestational age was calculated on the basis of oocyte retrieval or on the basis of the dating formulae with the best performance (Mul et al. (#2) (1996)18, Selbing and Kjessler (1985)17) at the dating scan. Among the 245 women with IVF pregnancy and spontaneous start of labor, 14 (5.7%) women gave birth preterm (≤258 days) and 14 (5.7%) post-term (≥294 days) according to oocyte retrieval, with no statistically significant difference in pre- and post-term rates between the early and the late scan group (Table 4). According to ultrasound dating using the two dating formulae with the best performance (Selbing and Kjessler (1985)18, Mul et al. (#2) (1996)18), pre- and post-term rates were very similar to those found when gestational age was calculated from oocyte retrieval. When we applied the BPD formula that we used clinically (Persson and Weldner (#1) (1986)8) to the 12–14 weeks scan group the post-term rate was much lower (0.9% vs. 4.5%), but the difference was not statistically significant (P = 0.12).

Table 3. Duration of pregnancy according to oocyte retrieval and to ultrasound estimation of gestational age in 363 pregnancies conceived after in-vitro fertilization
Method of calculation of gestational ageDuration of pregnancy (days)
12–14-week scan15–20-week scanAll
MeanSDMedianRangeMeanSDMedianRangePMeanSDMedian
  1. The hash (#) symbol denotes the formula number where a reference contains multiple formulae.

Oocyte retrieval
 All276.915.99280207–302276.713.12280224–3020.9276.814.50280
 Spontaneous onset of labor278.314.63282207–302279.611.35282231–2990.5279.012.95282
 Spontaneous onset of labor ≥37 weeks281.08.27283259–302281.67.90282259–2990.6281.48.06282
Ultrasound dating
 Spontaneous onset of labor ≥37 weeks
  Mul et al. (#2) (1996)18280.68.44281257–302281.28.36281260–3010.5280.98.39281
  Selbing and Kjessler (1985)17280.68.43281257–302281.68.46282261–3030.4281.28.44282
  Persson and Weldner (#1) (1986)8277.98.46278255–3002818.40281260–3010.006279.68.55280
Table 4. Numbers and rates of pre- and post-term deliveries according to oocyte retrieval and to ultrasound fetometry in pregnancies conceived after in-vitro fertilization with spontaneous onset of labor
Method of estimation of gestational ageDeliveries (n (%))
Preterm (≤258 days)Post-term (≥294 days)
12–14-week scan (n = 113)15–20-week scan (n = 132)12–14-week scan (n = 113)15–20-week scan (n = 132)
  1. The hash (#) symbol denotes the formula number where a reference contains multiple formulae.

Oocyte retrieval6 (5.3)8 (6.1)5 (4.5)9 (6.8)
Ultrasound fetometry
Mul et al. (#2) (1996)187 (6.2)8 (6.1)5 (4.5)8 (6.1)
Selbing and Kjessler (1985)177 (6.2)8 (6.1)5 (4.5)8 (6.1)
Persson and Weldner (#1) (1986)88 (7.1)8 (6.1)1 (0.9)7 (5.3)

Pregnancy duration in spontaneously conceived pregnancies

Table 5 shows pregnancy duration according to ultrasound fetometry in spontaneously conceived pregnancies. Using two dating formulae that showed very small systematic and random errors (Selbing and Kjessler (1985)17, Mul et al. (#2) (1996)18) both at early and late dating in our IVF population, pregnancy mean duration was significantly longer in the early scan group than in the late scan group (P = 0.02 and P = 0.01, respectively). When the BPD formula of Persson and Weldner was used (Persson and Weldner (#1) (1986)8) pregnancy duration was significantly longer in the late scan group than in the early scan group (P = 0.0001).

Table 5. Duration of pregnancy according to ultrasound measurement of the biparietal diameter in 4197 spontaneously conceived singleton pregnancies with spontaneous start of labor at ≥37 gestational weeks
Dating formula usedDuration of pregnancy (days)
12–14-week scan15–20-week scanAll
MeanSDMedianRangeMeanSDMedianRangeMeanSDMedian
  1. The hash (#) symbol denotes the formula number where a reference contains multiple formulae.

Selbing and Kjessler (1985)17282.18.01283261–305281.68.04282259–305281.88.03282
Mul et al. (#2) (1996)18282.18.01282262–305281.48.02281259–304281.78.02282
Persson and Weldner (#1) (1986)8279.48.03280259–302281.08.02281259–302280.28.06280

Discussion

Our results show that dating formulae originally designed to estimate gestational age in the middle of the second trimester are not necessarily applicable earlier in gestation. When applied in an IVF population with presumably known gestational age, three of the BPD formulae that we tested (Selbing and Kjessler (1985)17, Mul et al. (#1) (1996)18, Mul et al. (#2) (1996)18) performed well with small systematic and random errors at both early and late dating. Other BPD formulae were good either at early or at late dating but not at both (Persson and Weldner (#1) (1986)8, Hadlock et al. (1982)14), or they considerably overestimated gestational age at both early and late dating (Hadlock et al. (#1) (1984)9, Altman and Chitty (1997)12). Both formulae of Persson and Weldner, which are widely used for second-trimester dating in Scandinavia, performed well when dating was carried out at 15–20 weeks, but substantially underestimated gestational age when dating was performed at 12–14 weeks.

An ideal dating procedure is associated with no systematic error and a small random error. To achieve this we need a formula that performs well throughout our dating interval. Moreover, the ultrasound examiners must take precise measurements, and the technique must be the same as that used when the formula was designed. In the present study the systematic error of some formulae was very similar in the early and late scan groups, whereas that of other formulae differed substantially between the two groups. This suggests that the differences in systematic error between the early and late scan groups are explained by differences in performance between formulae and not by differences in measurement technique. Random error can be explained by biological variations in fetal size and by the precision of the measurements. In the present study the random error was very similar for all formulae tested. It was on average 1 day smaller at early dating than at late dating. Smaller random errors at early dating scans than at late dating scans were also found by others12, 17. It probably reflects smaller biological variation in fetal size at earlier gestation, even though a difference in measurement technique between early and late scans cannot be excluded.

One of the best formulae in our study was that of Selbing and Kjessler. Their formula was designed on the basis of 970 measurements of BPD taken at 9–22 weeks in women in whom gestational age according to the LMP had been confirmed by a CRL measurement17. Mul and colleagues designed dating formulae that were based on measurements taken at 14–24 weeks in 64 IVF pregnancies18. Despite the formulae of Mul and coworkers being designed on the basis of measurements taken later in gestation and in only 64 pregnancies, we confirmed the accuracy of their formulae both in our early and late dating group. The formulae by Persson and Weldner were designed on the basis of 67 measurements taken longitudinally between 11 and 24 weeks in 14 women with reliable ovulation date according to basal body temperature charts. According to our results, both their BPD formula and their BPD/FL formula underestimated gestational age by on average 3 days when dating was done at 12–14 weeks.

Mul and coworkers tested several established dating formulae in their 64 IVF pregnancies18. Performing the scans at 14–24 weeks, they found an overestimation of gestational age by 1 day for the BPD formula of Selbing and Kjessler, an underestimation by 0.3 days for the BPD formula of Persson and Weldner, and an overestimation by 3–4 days for the formulae of Hadlock and coworkers. These figures are in accordance with the results in our late dating group, with the exception that the formula of Selbing and Kjessler performed better in our hands. Possibly, the discrepancy is to be explained by a difference in gestational age at dating. Mul and coworkers performed their dating at 14–24 weeks, whereas almost all women in our late group were scanned before 20 weeks. Our results in the late dating group are also in accordance with those of Chervenak and coworkers, who tested 38 previously published dating formulae, including seven of the formulae tested in our study, in 156 singleton IVF pregnancies19.

The BPD and BPD/FL formulae designed by Persson and Weldner have been validated in two studies. Geirsson and Have20, analyzing 36 IVF pregnancies, found a mean difference between estimated and true gestational age of −0.9 days using the BPD/FL formula of Persson and Weldner. They calculated true gestational age from the day after oocyte retrieval. Had they calculated true gestational age from the day of oocyte retrieval, their results would have been similar to those of another study21 comprising 253 IVF pregnancies where the difference between estimated and true gestational age was −1.9 days using the BPD formula and −2.1 days using the BPD/FL formula of Persson and Weldner. In both studies cited20, 21, the dating scans were carried out in the second trimester. Thus, both in our study and in those of Geirsson and Have and Wennerholm and coworkers, the formulae of Persson and Weldner underestimated gestational age when dating was performed in the second trimester.

Two formulae were associated with a substantial overestimation of gestational age (Hadlock et al. (#1) (1984)9, Altman and Chitty (1997)12). These formulae were both designed on the basis of a large number of measurements taken in women with reliable menstrual dates, but the measurements were taken from 14 to 42 weeks (Hadlock et al. (#1) (1984)9) or from 12 to 42 weeks (Altman and Chitty (1997)12). Individual variations in fetal size increase with the progress of pregnancy9, 14, 17. This probably explains why a formula designed to fit the duration of a whole pregnancy performs poorly when the period of dating is limited to 12–20 weeks.

Normal pregnancy duration in humans is not known. It is most often stated to be 280 days, which corresponds to a slight modification of Naegele's rule22. In our study, mean pregnancy duration including all singleton IVF pregnancies was 277 ± 14.5 days. If only pregnancies with spontaneous onset of labor were included, mean pregnancy duration was 279 ± 12.9, with a preterm rate of 5.7%. Many preterm deliveries represent pathological pregnancies. Therefore, it seems appropriate to estimate ‘normal’ pregnancy duration both in IVF pregnancies and in pregnancies conceived spontaneously only on the basis of pregnancies with spontaneous onset of labor ≥259 days. In doing so, we found pregnancy duration in IVF pregnancies to be 281 ± 8.1 days. IVF pregnancies may not be representative of normal pregnancy, even after exclusion of preterm deliveries, inductions of labor and fetal malformations. Mean age and the rate of nullipara are usually higher in women who have conceived after IVF than in those who have conceived spontaneously23, 24. However, case-control studies comparing IVF pregnancies and naturally conceived pregnancies matched for age, parity, time and location of birth, and number of fetuses have shown very similar obstetric outcome concerning pregnancy duration and maternal and fetal complications25, 26. Therefore, calculating gestational age from oocyte retrieval in IVF pregnancies is probably the closest we can come to true gestational age. We also applied two dating formulae that had shown small systematic and random errors when estimating gestational age in our IVF pregnancies (Selbing and Kjessler (1985)17, Mul et al. (#2) (1996)18) to a larger population of spontaneously conceived pregnancies. We found mean and median pregnancy duration to be slightly less than 282 days. We believe that our results provide scientific evidence that 281 or 282 gestational days—not 280 days—should be used as the estimated day of delivery. This is supported by a study of Tunón and coworkers, where median pregnancy duration was 281 days according to ultrasound dating in 9400 women with regular menstrual cycles, reliable LMP and spontaneous onset of labor6, and by Bergsjö and Denman, who found median pregnancy duration to be 282 days according to LMP in a study of 383 000 women with known LMP27. In both studies cited preterm deliveries were included. Mean pregnancy duration is affected by inclusion of preterm deliveries. Therefore, median pregnancy duration may be a better estimate in populations including preterm deliveries.

It is important for every ultrasound unit to regularly check its quality of dating scans. Ideally this should be done by comparing gestational age as estimated by fetometry with ‘true’ gestational age in pregnancies with known date of conception, e.g. in singleton IVF pregnancies. However, for most ultrasound units this is not a realistic option, because there are too few singleton IVF pregnancies. An alternative is to calculate the difference between the actual date of delivery and the estimated date of delivery, and to use the mean of these differences to reflect ‘systematic’ measurement error, and the SD of the differences to reflect the random measurement error. A positive mean would reflect systematic overestimation of gestational age at the dating scan, and a negative mean would reflect systematic underestimation. Our results suggest that the SD of ‘normal’ pregnancy duration in singletons is about 8 days. Thus, in an ultrasound unit with perfect dating scans the SD of pregnancy duration as calculated from ultrasound fetometry should be approximately 8 days (including only singletons with spontaneous start of labor ≥37 weeks). However, the mean of the differences between actual delivery date and estimated delivery date is an unreliable measure of systematic measurement error because these differences are highly dependent on how estimated day of delivery is defined. Both our results and those of others suggest that the estimated day of delivery should be at 281 or 282 completed days of gestation—not 280 days. Should we have used 280 days as the estimated day of delivery in our population of spontaneously conceived pregnancies we would have had a positive mean difference between actual delivery date and estimated delivery date even when using the best dating formulae. This would have suggested that we overestimated gestational age at the dating scan, while in fact we almost certainly underestimated it (as shown by the results in our IVF population). The post-term rate is another possible measure of the quality of dating scans. Our results indicate that the post-term rate among ‘normal’ singleton pregnancies is approximately 6%. A post-term rate substantially higher than this may reflect systematic overestimation of gestational age at the dating scan, whereas a substantially lower rate might reflect systematic underestimation of gestational age—provided that post-term rate is calculated in the same way as in the present study. The BPD formula of Persson and Weldner may be used as an example. This formula underestimated gestational age by on average 3 days in the early IVF scan group and was associated with a post-term rate of 0.9%, while in fact the ‘true’ post-term rate according to oocyte retrieval was 4.5%. However, it is important to keep in mind that also the post-term rate is highly dependent on how it is defined and calculated. To the best of our knowledge there is no universally agreed definition of post-term rate. If the post-term period is defined to start later than at 294 gestational days the post-term rate will be lower than if it is defined to start at 294 days. Tunón and colleagues reported a 2.2% post-term delivery rate in a population of 10 000 women with spontaneous onset of labor, but they defined post-term as ≥296 days6. Inclusion/exclusion of labor inductions and Cesarean sections will also affect the post-term rate.

In conclusion, ultrasound dating formulae originally intended for use in the middle of the second trimester do not necessarily perform well when used for dating earlier in gestation. We have found three ultrasound dating formulae that seem to be associated with small systematic and random errors when used for dating both at 12–14 and 15–20 weeks (Mul et al. (#1) (1996)18, Mul et al. (#2) (1996)18, Selbing and Kjessler (1985)17). Our finding of smaller random error at early dating supports the hypothesis that early dating yields more precise estimates of gestational age than late dating. This is in agreement with smaller biological variation of fetal size in early pregnancy.

Acknowledgements

This work was supported by grants from the Stockholm County Council Public Health and Medical Services Committee R&D departments and the Karolinska Institute South Hospital. Colleagues and midwives at the participating ultrasound units at the Departments of Obstetrics and Gynecology at Huddinge University Hospital, UltraGyn/Danderyd Hospital, South Stockholm General Hospital, Karolinska Hospital, Södertälje Hospital, Malmö University Hospital, Helsingborg Hospital and Uppsala Academic Hospital are gratefully acknowledged for their valuable contribution to this study.

Ancillary