Correspondence: Dr J. Gardosi, Perinatal Research, Audit and Monitoring (PRAM), Department of Obstetrics and Gynaecology, Queen's Medical Centre, Nottingham NG7 2UH, UK.
Objective To examine the length of gestation according to menstrual and ultrasound scan dates, and the rate of induction of labour in a unit with a routine induction policy for prolonged pregnancy.
Design Retrospective analysis of computer files of 24,675 pregnancies delivered in a teaching hospital between 1988 and 1995, which had a record of the last menstrual period and a dating ultrasound scan. Detailed survey of 168 casenotes of consecutive inductions of labour to establish the indications given.
Setting Teaching hospital with policies of routine mid-trimester ultrasound scan and routine induction for prolonged pregnancy at 290 to 294 days.
Main outcome measures Gestational age at delivery by menstrual history and ultrasound biometry in spontaneous and induced labours
Results The single largest category of reasons given for induction of labour was prolonged pregnancy.‘Post-term pregnancy’, from the date of expected delivery as recorded in the notes, together with ‘maternal request’ and ‘social factors’, were the reasons given for induction of labour in 71.3% of cases. Menstrual dates systematically overestimated gestational age at term when compared with scan dates. After 41 weeks, this difference exceeded the confidence limits for second trimester scan dating error, suggesting that most pregnancies which are considered ‘prolonged’ according to menstrual dates are in fact mis-dated. The median gestational age for induced labours was 286 days by last menstrual period but only 280 days by scan, and most (71.5%) inductions done post-term (> 294 days) according to menstrual dates were not post-term if scan dates alone are used to calculate the gestational age. The average induction rate over the seven year study period was 16.6%. It was higher when there was any gestational age error in either direction (16.8%) compared with when menstrual and scan dates were in complete agreement (13.7%, OR 1.27, CI 1.09–1.47, P < 0.001). The induction rate was highest (up to 21.8%) in the cases where menstrual dates overstated gestational age without exceeding the usual limits for adjusting dates according to scan. Such over-estimation within tolerance limits of 7, 10 or 14 days occurred in 37.1%, 45.8%, or 52.6% of all pregnancies, respectively.
Conclusions Most pregnancies undergoing post-term induction are not post-term when assessed by ultrasound dates. Regardless of whether prolonged pregnancy is considered to be a risk factor requiring intervention, the proportion of pregnancies considered ‘post-term’ can be reduced considerably by a dating policy which ignores menstrual dates and establishes the expected delivery date on the basis of ultrasound dates alone.
Induction of labour is an intervention to expedite delivery when there is concern about the pregnancy. Prolonged pregnancy is seen as a reason for routine induction and protocols in many units advocate delivery after 41 or 42 weeks. In part, such protocols are derived from considerations about fetal wellbeing and meta-analyses indicating that induction reduces perinatal mortality1. Furthermore it has been suggested that induction of labour for prolonged pregnancy results in a reduction in caesarean sections2,3. There is often also maternal concern about delay past the expected date of delivery4.
There is considerable variation in the way that the expected date of delivery is determined. It is known that the last menstrual period (LMP), even when recalled with confidence, can result in considerable dating error5. Using scan dates will result in fewer pregnancies being considered post-term6. Various methods are in use to adjust to ultrasound dates (e.g. when the discrepancy between scan dates and menstrual history is 7, 10 or 14 days). However, predicting the expected date of delivery by the dating scan alone has been found to be more accurate than any method which also takes LMP into account7. This is mainly because menstrual dates error is heavily skewed towards over-estimating the true gestational age8, whereas scan error is normally distributed9. The purpose of this study was to analyse differences between menstrual history- and scan-derived pregnancy dates, and the effects of such variation in the presence of a routine induction policy for prolonged pregnancy.
Data from the computerised obstetric database were analysed in anonymised form, starting from 1988 when routine mid-trimester dating of pregnancies was introduced. Multiple pregnancies and congenital abnormalities were excluded, which left a total of 34,052 computerised obstetric data files for further examination, constituting all normally formed singleton live deliveries over the period of April 1988 to May 1995 at the Queen's Medical Centre. Data files with incomplete menstrual and scan dates were excluded. Accordingly, 9377 (27.5%) of pregnancies had to be excluded because no LMP was given (n= 2244; 6.6%), no dating scan measurement of biparietal diameter between 13 and 22 weeks was recorded (n= 6155; 18.1%) or both these data were absent (n= 978; 2.9%). This left a total of 24,675 records for analysis. The average rate of operative delivery during this period was 28.7%. This included 14.9% forceps and ventouse deliveries and 13.8% caesarean sections, of which 5.4% were elective and 8.3% emergency operations.
The expected date of delivery is usually calculated in our unit by using the menstrual dates, unless they are outside ±10 days of the ultrasound scan dates, in which case the latter are used to calculate expected date of delivery. However in practice a tolerance of ±14 days is also often applied before the menstrual dates are disregarded in favour of scan dates.
The dates of the last menstrual period, if recalled by the mother with reasonable certainty, are recorded at pregnancy booking. Over 90% of women in our population have at least one ultrasound scan between 14 and 22 weeks, at which time gestational dates are calculated from biparietal diameter measurement according to conventional dating tables10.
In cases where the pregnancy goes post-dates, there is a policy to induce labour with vaginal prostaglandin and/or oxytocin infusion. This is usually done at 41+ weeks (290 days) or 42 weeks (294 days). The actual reasons for induction were not recorded in the computer files. To ascertain the indications given for induction of labour in our population, a detailed casenote audit of all deliveries over a three-month period (October to December 1994) was carried out. There were 1071 deliveries during this time, including 168 (15.7%) where labour was induced.
Data analysis was performed with software including Microsoft Excel 5.0 and SPSS for Windows 6.1.1.
Table 1 shows a breakdown of the reasons for induction of labour as recorded in the notes of the 168 consecutive inductions of labour. The data are grouped according to whether delivery was before or after the expected delivery date as determined by the clinicians involved in the case. Prolonged pregnancy was the single most common cause for inducing labour. Together with the pregnancies induced for maternal requests and/or social reasons, 71% of inductions past the expected date of delivery were carried out for no other medical or obstetric indication. From 290 days, the corresponding rate for this category was 89%.
Table 1. Reasons stated for induction of labour in 168 of 1071 (15.7%) unselected singleton consecutive deliveries. 524% of the induced deliveries were before and 47.6% on or after the expected date of delivery = 280 days as recorded in the notes. Values are given as n (%). 1UGR = intrauterine growth restriction
Reason for induction
< 280 days n= 88
> 280 days n= 80
Spontaneous rupture of membranes
IUGR or reduced fetal movements
Other medical / obstetric causes
The overall elective delivery rate over the seven-year study period was 22.0%, which included 16.6% inductions and 5.4% elective caesarean sections. The gestational age distribution was skewed towards early deliveries. The modal length of pregnancy was 283 days by LMP and 281 days by scan (Table 2); the median was 281 (LMP) and 278 (scan), which increased slightly to 282 and 279 days, respectively, if preterm deliveries (< 37 weeks) were excluded. The median for induced labours was 286 by LMP but only 280 days by scan, and if preterm inductions were excluded, the figures were 288 and 282 days, respectively. For elective caesarean sections, which are usually aimed to be done at 38 weeks of gestation, the mode was 267 days (38 weeks and 1 day) by menstrual as well as scan dates.
Table 2. Length of gestation according to last menstrual period (LMP) and mid-trimester ultrasound scan. HDD = expected date of delivery
% > EDD
Yo > EDD+ 10
% > EDD+ I4
Using scan dates to calculate gestational age reduces the proportion of pregnancies classified as post-term (Table 2). Among spontaneous labours, the post-term (> 294 days) delivery rate was 9.5% according to LMP but only 1.5% by scan. Among induced labours, there was a reduction from 19.3% to 5.5% when scan instead of menstrual dates were applied (i.e. 71.5% of inductions post-term according to LMP were not post-term according to scan). However, most of the differences between menstrual and scan dates were within commonly accepted tolerance margins of 7, 10 or 14 days before dates are reset according to the scan: only 33.0% of cases were outside ±7 days agreement, 22.4% outside ±10 days, and 14.6% outside ±14 days. Most variation was due to over-estimates by menstrual dates, which up to the 7, 10 and 14 day limits, respectively, amounted to 37.1%, 45.8% and 52.6% of all pregnancies.
Table 3 shows the proportion of labour inductions and elective caesarean sections according to the magnitude of the discrepancy between LMP and scan dates. Caesarean section rates showed no significant differences in the various categories. However the induction rates were significantly higher when menstrual dates over-estimated gestation length than in those cases (13.7%) when the two dates were in complete agreement (Table 3). Analysis of this group in smaller categories showed induction rates of 17.2% when menstrual dates were greater than scan dates by 1–3 days; 19.6% (4.7 days), 21.8% (8–10 days), 18.6% (11–14 days), and 16.3% (> 14 days). Induction rates thus appeared to rise alongside a positive LMP - scan difference up to 10 days, from when clinicians tend to increasingly disregard menstrual dates in favour of the scan results. There was a slight, nonsignificant trend towards lower induction rates when menstrual dates under-estimated scan dates. Overall, 16.8% of labours were induced when there was any discrepancy in either direction, compared with 13.7% when menstrual and scan dates were equal (OR 1.27, 95% CI 1.09–1.47, P < 0.001).
Table 3. Induction of labour (IOL) and elective caesarean section (ECS) rates in various error categories between dates from last menstrual period (LMP) and ultrasound scan (scan). Odds ratios and significance tests compare rates against the baseline rates when dating difference = 0
Difference in days LMP - scan
−14 to −8
1 to 7
8 to 14
Figure 1 shows the mean and range (95% confidence limits) of deviations of gestational age at birth according to LMP, compared with scan dates as reference. The 95% confidence interval of second trimester dating with a standard ultrasound dating formula10 is also plotted. This error margin is −9-3 to 7-4 days and is based on the accuracy of biparietal diameter measurement in 25 obstetric units, from a study of pregnancies achieved with assisted reproduction techniques9. The graph shows an increasing discrepancy between menstrual and scan dates with advancing gestational age. The line crosses the confidence limit for scan error at 41+ weeks and its slope approximates unity. This suggests that any LMP-based delivery date past this gestation is more likely to represent dating error than a true post-term pregnancy.
Gestational age at delivery is plotted in Fig. 2a and 2b for spontaneous and induced labours as well as for elective caesarean sections. In Fig. 2a, the dates are calculated according to menstrual history. The inductions rise to a peak at 294 days corresponding to the modal value shown in Table 2. This is followed by a rapid drop, reflecting the induction protocol for prolonged pregnancy. However, when calculating gestational age at delivery according to scan dates, it becomes apparent that most inductions were well before the post-term period (Fig. 2b; Table 2).
The data show that there are large differences between gestation length determined by menstrual history and ultrasound scan, which is consistent with previous find-ings6,8,11,12. Ultrasound dates are more accurate than menstrual dates alone or in combination with ultrasound in predicting the date of delivery7. The discrepancy between menstrual and scan dates increases with the apparent gestational age at birth according to LMP (Fig. 1): at 41+ weeks, the line crosses the confidence limit for scan error which indicates that from this gestational age it is more likely that menstrual dates are wrong than that the pregnancy is actually this far advanced. Even if the date of the last menstrual period is recalled with accuracy, delay in ovulation can result in over-estimation of the true gestational age, which results in an apparent prolongation of pregnancy13.
By dates according to the ultrasound scan, the overall rate of post-term pregnancy in our population was rela- tively rare (2.2%). Most labours were in fact induced before being actually post-term by scan (Fig. 2; Table 2). Of the pregnancies which were induced after expected date of delivery + 10 by menstrual dates, only 16.2 of the 36.4%, (i.e. 44.5%) were actually past this limit according to the scan-based EDD; for 42 weeks (294 days) as cutoff, this proportion was even smaller: 5.5 of 19.3% (i.e. 28.5%). Although various protocols exist to reset menstrual dates according to the scan if the discrepancy is greater than 7, 10 or 14 days, most dating error occurs within these margins and is associated with significantly increased induction rates (Table 3). The data suggest that even if the discrepancy between menstrual and scan dates is less than a week, ignoring menstrual dates altogether and using scan dates alone to date pregnancy would reduce the induction rate substantially. It would be reasonable to speculate that a successful application of a general policy to use scan dates alone would result in an overall induction rate similar to that when the menstrual - scan date difference is 0 (i.e. 13.7%) (Table 3). This would be a 2.9% drop from our induction rate of 16.6%, or about 120 fewer inductions per year, simply by changing the unit policy for calculating expected date of delivery.
Although it is impossible to know when induced pregnancies would have gone into labour spontaneously, the steep drop of both spontaneous and induced labour curves (Fig. 2b) suggest that most pregnancies go into spontaneous labour soon after the true expected date of delivery is passed. This can be illustrated by calculating the proportion of pregnancies still remaining at various points after the expected date of delivery and plotting curves for the likelihood of spontaneous onset of labour (Fig. 3). For example at 41+ weeks (290 days) by ultrasound, and assuming that there is no other indication for elective induction, the likelihood of delivery after spontaneous onset of labour is 60% within the next three days, and 90% within seven days. Although these figures are derived from births excluding elective deliveries, there is no reason to assume that the routine induction policy selected those pregnancies which would have continued for longer. This is supported by similar findings in a population with no routine induction policy, where the overall incidence of scan-based post-term pregnancy was also rare (2.9%) and where 90% of women who were still pregnant at 42 weeks had gone into spontaneous labour within the following week14.
It is often argued that women themselves insist on labour being induced once the due date has passed and the pregnancy seems to drag on. A study of mothers’ attitudes suggested that most are unwilling to accept conservative management of prolonged pregnancy past 41 weeks4. However, mothers need to know all the facts to be able to make an informed choice, and it remains to be seen how many would insist on induction if they were aware of the high likelihood of spontaneous onset of labour past a correctly calculated expected date of delivery, as illustrated in Fig. 3. Furthermore, it is most important that dates are calculated accurately at the beginning of pregnancy, after the dating scan, and not attempted to be ‘reset’ towards the end, when the pregnancy appears to be prolonged.
Regardless of obstetric and maternal views of the advantages and disadvantages of routine induction policies, our results suggest that most post-date inductions are unwarranted on the basis of gestational age. The incidence of prolonged pregnancies can be considerably reduced by establishing dates by ultrasound alone.