1. Top of page
  2. Abstract
  7. References

Objective To study early pregnancy characteristics as possible risk factors associated with preterm birth.

Design Retrospective analysis of prospectively collected maternity data.

Population 21,069 singleton deliveries with record of a specified last menstrual period and a midtrimester dating scan.

Setting Catchment area of tertiary centre serving a general maternity population.

Methods Univariate and multivariate analysis. Variables included: maternal age; height; weight at first visit; parity; ethnic group; cigarette smoking and alcohol consumption recorded in early pregnancy; history of abortion; history of preterm birth; and discrepancy between menstrual dates and ultrasound dates.

Main outcome measures Adjusted odds ratios for factors associated with preterm birth, stratified according to parity (nulliparae vs multiparae) and gestational age (early preterm, 24–33 weeks; late preterm, 34–36 weeks; all preterm, < 37 weeks). Population attributable risk (aetiologic fraction) of the significant variables for preterm birth.

Results The overall preterm (< 37 weeks) delivery rate according to scan dates was 7.0%. Preterm birth was associated with young (< 20 years), short (≤ 155 cm) and underweight (≤ 52 kg) mothers, non-Europeans, cigarette smokers, previous abortion or previous preterm delivery, and a prolonged menstruation-conception interval. Preterm births which followed the spontaneous onset of labour (72%) had results which were similar to the overall group, while there were too few iatrogenic preterm deliveries for separate analysis. Logistic regression showed that associations varied in different parity and gestational age groups. For nulliparae, smoking was not associated with preterm birth, but it was strongly associated with multiparous women (adjusted OR 1.8, 95% CI 1.6–2.1). A past history of premature delivery had the highest risk for birth before 34 weeks in the index pregnancy (adjusted OR 5.1, 95% CI 3.4–7.6). A discrepancy between menstrual and scan dates of greater than +7 days, suggestive of a prolonged interval between last menstruation and conception, was present in 23.3% of all pregnancies, and was associated with an increased risk of preterm delivery in all gestational age categories for nulliparae (adjusted OR 1.5, 95% CI 1.3–1.8) and multiparae (adjusted OR 1.9, 95% CI 1.6–2.2). Because of its high prevalence, this variable constituted a relatively high population-attributable risk for premature birth for both nulliparae (10.7%) and multiparae (16.6%).

Conclusions A discrepancy of more than +7 days between menstrual and scan dates, indicating a prolonged interval between last menstruation and conception, is a significant predictor of preterm birth. This effect is independent of other factors such as maternal age, height, weight and smoking which are also associated with prematurity. In a maternity population with ultrasound scan dates and recorded last menstrual period, this variable can be easily calculated and used as a marker for increased surveillance.


  1. Top of page
  2. Abstract
  7. References

Preterm birth is the major cause of neonatal mortality, but despite many efforts, little progress has been made in its prediction. Preterm labour is considered to have heterogeneous origins1 and has strong associations with pregnancy complications such as spontaneous rupture of membrances, pre-eclampsia, and antepartum haemorrhage. In low risk pregnancies the strongest indicators rhage. In low risk pregnancies the strongest indicators have been found to be nulliparity and male gender of the fetus, but known risk factors were found to explain only a small fraction of spontaneous preterm birth2. To our knowledge, no recent study has analysed the markers for preterm delivery which are known to the clinician at the beginning of pregnancy in order to allow assignment of the appropriate level of antenatal care towards early recognition of preterm labour.

We used a database of routinely scan dated pregnancies to undertake such an analysis. There is evidence that many pregnancies are misdated, especially at the extremes of gestation3,4. Where equipment and trained personnel are available, ultrasound dating in the second trimester is better than certain menstrual dates alone or in combination with scan dates5. Furthermore, analysis of assisted reproduction technique pregnancies with exact dates has demonstrated the reliability of commonly used ultrasound formulae6,8, which have a smaller and normally distributed error than menstrual history. It has therefore been recommended that ultrasound should be used as the gold standard for dating pregnancies9.

Uncertain gestation dates are more likely to be associated with adverse pregnancy outcome, and exclusion of such cases from analysis may give an incorrect representation of the true rate of preterm delivery10. Yet dates given by mothers often show considerable gestation error even when they are sure about their last menstrual period11. It is unknown to what extent specified,‘certain’ yet incorrect dates are associated with preterm delivery. It has been our clinical impression that women whose menstrual dates have been adjusted because they are more than a week different from scan dates were more likely to have problems later in pregnancy. We therefore used ultrasound biometry for dating but also included in the analysis a variable which represents the discrepancy between menstrual and scan dates.


  1. Top of page
  2. Abstract
  7. References


The maternity population at our institution is unselected, in that virtually all pregnancies in this urban and suburban catchment area are booked for hospital delivery within the National Health Service. Antenatal care is usually in a shared care arrangement between hospital specialists and community clinics with general practitioners and midwives.

The data were derived from anonymised computer files of 32,154 consecutive, singleton live births between 1988 and 1995. During this time the policy of a routine second trimester ultrasound scan became increasingly established in our unit, including about 75% in 1988 to over 90% of all pregnancies during the 1990s. We excluded all cases (n= 3685) where neither a last menstrual period nor an ultrasound dating scan was recorded, which included pregnancies with late presentation for antenatal care and referrals from other units. This left 28,469 computer files for which we had a record from early pregnancy and from delivery. This consisted of 5859 cases where a last menstrual period was specified (i.e. calendar day, month, year) but no early ultrasound scan was recorded; 1541 cases with no menstrual dates but an ultrasound scan with a biparietal diameter measurement between 28 and 55 mm, equivalent to 14 to 21 completed weeks of gestation12; and 21,069 cases where there was a record of both the last menstrual period and a mid-trimester ultrasound scan. This latter category was the focus of the current analysis.

The data included maternal height and weight as measured at the first hospital visit. These data were grouped into categories nearest to the 90th and 10th centiles for height and weight of mothers of term pregnancies. Body mass index was calculated from weight/height2 (kg/m2). Ethnic group comparisons were made between (Anglo-) Europeans (89.2%) and non-Europeans (10.8%), which included mothers from the Indian subcontinent (Indian and Pakistani, 6.3%) and from the West Indies (Afro-Caribbean; 2.1%) and others (2.4%). Data on smoking and alcohol consumption were derived from the obstetric history taken by a midwife at the first hospital visit. Smoking was categorised as 1–9 and 10+ but grouped together in the multivariate analysis to give sufficient numbers. Alcohol consumption was reported as moderate (occasional or social, up to 10 units per week) by 31.6% of mothers. Only 14 admitted to regular drinking and one to heavy drinking, and these 15 were excluded as too few for further analysis.

Pregnancy dating

Gestational age at birth was calculated by computer and expressed in days, using two methods:

  • 1
    Menstrual dates were derived from the first day of the last menstrual period, as specified by the mother and recorded by a midwife in the antenatal booking clinic. If the mother is not sure about her dates, the usual practice is not to record an approximate date, as dates based on the routine ultrasound can subsequently be used.
  • 2
    A second trimester ultrasound scan is policy for all pregnancies. In some pregnancies a scan is performed in the first trimester for various indications and a crown-rump length is usually measured, but a repeat visit for a detailed structural scan is then still arranged for 18–20 weeks. For purposes of this analysis, the second trimester biparietal diameter measurement was used to date all pregnancies. If more than one biparietal diameter measurement was recorded, the one taken at the time of the detailed anomaly scan was selected. The standard biparietal diameter ultrasound formula by Campbell and Newman12 is used. In an analysis of assisted conception pregnancies, we have found this formula to perform well in routine, multi-operator use, with a systematic error of less than a day and a normally distributed random error of ± 4.26 days7.

The ultrasound dates were used to determine whether a delivery was preterm. This was based on previous findings that dating by ultrasound scan is more accurate than a specified last menstrual period alone or in combination with the scan5. However, the difference between menstrual dates and scan dates was also calculated. By convention, ultrasound dating formulae add a standard 14 days to the fetal biometry-based conception age to express gestational age. A positive difference between last menstrual period and scan-derived gestational ages can therefore indicate a menstruation-conception interval which is more than 14 days, and a negative difference can indicate an interval which is less than 14 days. We chose a cutoff of greater than one week as a menstrual—scan dates discrepancy of potential clinical relevance. For a random error of 4.26 days for scan dating7, a range of ± 7 days represents z= 7/4.26 or ± 1.643, which is equivalent to a 90% confidence interval. A discrepancy ≤ 7 days was termed a ‘shortened menstruation-conception interval’, and a discrepancy > +7 days was a ‘prolonged menstruation-conception interval’. These two derived variables were included in the analysis.

As a rounding down of the first day of the last menstrual period (e.g. to record it as the first of the month if the exact day of the month was not known or not certain) would exaggerate a positive menstrual-scan dates discrepancy, we studied the frequency distribution of the given first day of the last menstrual period in the database of 21,069 births. The results showed that there was some evidence of rounding, but this was not confined to the first day of the month. Compared with the average percentage of last menstrual period dates which fell on each day from 1st to 30th of the month (3.3%), there were modes in the frequency distribution on day 1 (4.9%), day 10 (3.8%), day 15 (4.4%), day 20 (4.3%) and day 25 (3.8%). When grouped by weeks, the distribution of reported last menstrual periods was even throughout the month, with 23.9% falling on days 1–7, 23.4% on days 8–14, 24.3% on days 15–21, 23.1% on days 22–28, and 6.9% on days 29–31.

Prematurity categories

Preterm was defined as < 37–0 weeks (< 259 days). The analysis included two preterm gestational age categories: early preterm (< 34.0 weeks) and late preterm (34.0–36.6 weeks). The early preterm cutoff at 34 weeks was used because:

  • 1
    It represents a clinically significant milestone in our institution, before which tocolysis was usually attempted for preterm labour unless contra-indicated.
  • 2
    Steroids were routinely administered in cases of threatened preterm delivery up to this gestation and not beyond.
  • 3
    It is the gestational age before which a baby is routinely admitted to the special care unit.

This cutoff point left sufficient numbers in the early preterm subgroup for statistical comparisons. We conducted analyses for all preterm deliveries and separately also for those which were preceded by spontaneous onset of labour (78.3%) but not for nonspontaneous i.e. iatrogenic preterm deliveries (21.7%) as the numbers were too low. We did not attempt to differentiate between onset of labour with or without spontaneous rupture of membranes, as often the time sequence is not accurately known and/or not well recorded.

Birthweight categories

Birthweight centiles were calculated using coefficients to adjust for nonpathological pregnancy characteristics known to affect birthweight13 maternal height, weight, ethnic group, parity, baby's gender and gestational age (based on ultrasound dates).

Statistical tests included univariate and multivariate analyses with stepwise logistic regression, using the Statistical Package for the Social Sciences (SPSS) and Microsoft Excel with added functions programmed in Visual Basic. Odds ratios and confidence intervals were derived using standard formulae14 and P= 0.05 was used as cutoff for statistical significance. Logistic regression with preterm delivery as dependent variable was stratified for each preterm gestation category and run separately for nulliparae and multiparae. Variables known to be associated with preterm delivery were entered in all analyses: maternal age, height, weight, ethnic group and smoking, and for multiparae, history of preterm delivery. The following additional variables were entered and excluded stepwise if nonsignificant: alcohol consumption, history of abortion, and shortened or prolonged menstruation-conception interval.


  1. Top of page
  2. Abstract
  7. References

The preterm delivery rate for the study group was 6.3% according to last menstrual period and 7.0% according to scan dates (Table 1). It was higher in the groups which were excluded from analysis: pregnancies with no record of a last menstrual period had a prematurity rate of 9.4% by scan. For pregnancies with no dating scan, it was higher still, 10.2%, although this figure is based on menstrual dates which are likely to be unreliable in late-presenting pregnancies.

Table 1.  Preterm delivery rates of 28,469 pregnancies with a record of menstrual or ultrasound datess. LMP = last menstrual period as recorded at the first hospital visit; SCAN = dating scan based on measurement of the biparietal diameter between 28 and 55 mm (equivalent to 14 to 21 weeks of gestation). Values are given as n or n (%).
  Early preterm (≤ 34 weeks)All preterm (≤ 37 weeks)
LMP only5859228 (3–9)596 (10–2)
SCAN only154143 (2.8)145 (9.4)
LMP and SCAN21,069368 (1.7)399 (1.9)1326 (6.3)1485 (7.0)

The analysis was restricted to pregnancies which had specified last menstrual period dates as well as ultrasound scan dates. In Table 2 the relative proportions of characteristics known at the booking visit are listed, together with the preterm delivery rate (calculated according to scan dates). Premature birth was significantly associated with young, short, underweight mothers, cigarette smokers, and those with a past history of abortion or preterm delivery. Moderate alcohol drinkers had a reduced level of preterm births which reached borderline significance (Table 2).

Table 2.  Univariate analysis of factors associated with preterm birth.
 All birthsPreterm (≤ 37 weeks)
 nColumn %n(%)Odds ratio (95% CI)
  1. *Baseline category for odds ratios.

  2. Excluded because of missing data or likelihood of erroneous entry. Accepted ranges: maternal age 10–60; height 140–200 cm; weight 40–140 kg; menstrual-scan dates discrepancy −100 to +200 days.

Total21,069 1485 (7.0) 
  Nulliparae908643.1685 (7.5)1.14 (1.03–1.27)
  Multiparae*11,98356.9800 (6.7) 
  Iatrogenic458021.7421 (9.2)1.47 (1.30–1.65)
  Spontaneous*16,48978.31064 (6.5) 
Maternal age    
  ≤2016457.8171 (10.4)1.60 (1.35–1.90)
  20–34*17,84684.81204 (6.7) 
  > 3515677.4109 (7.0)1.03 (0.84–1.27)
Maternal height (cm)    
  < 155242411.8225 (9.3)1.41 (1.21–164)
  156–171*16,12278.21090 (6.8) 
  > 172208210.1135 (6.5)0.96 (0.79–1.15)
Maternal weight (kg)    
  <5220499.9231 (11.3)1.80 (1.55–2.10)
  52–80*16,35079.01076 (6.6) 
  >80229711.1154 (6.7)1.02 (0.86–1.21)
Body mass index (kg/m2)    
  <20.0203810.0203 (10.0)1.56 (1.33–1.83)
  20.1–29.4*15,94678.31056 (6.6) 
  ≥29.4237211.7173 (7.3)1.11 (0.94–1.31)
Ethnic group    
  European*18,80089.21285 (6.8) 
  Non-European226910.8200 (8.8)1.32 (1.13–1.54)
  Indian/Pakistani13166.3118 (9.0)1.34 (1.10–1.64)
  Afro-Caribbean4492.144 (9.8)1.48 (1.08–2.03)
Other5042.438 (7.5)1.11 (0.80–1.55)
Cigarettes per day    
  1–919099.1150 (7.9)1.26 (1.05–1.50)
  10+314714.9317 (10.1)1.65 (1.45–1.88)
  Any505624.0467 (9.2)1.50 (1.34–1.68)
  None*16,01376.01018 (6.4) 
Alcohol consumption    
  Moderate664031.6436 (6.6)0.89 (0.80–1.00)
  None*14,41468.41049 (7.3) 
History of abortion    
  Yes462922.0372 (8.0)1.20 (19.06–1.36)
  No*16,44078.01113 (6.8) 
History of preterm delivery (11,983 multiparae)    
  Yes4463.784 (18.8)3.51 (2.74–4.50)
  No*11,53796.3716 (6.2) 
Menstrual-scan dates discrepancy    
  ≤−7 days10014.854 (5.4)0.86 (0.65–1.14)
  Within ± 7 days*14,93371.9927 (6.2) 
  ≥+7 days484523.3496 (10.2)1.72 (1.54–1.93)

A discrepancy between menstrual dates and scan dates of more than a week was the case in 29.1% of pregnancies. Most of these (23.3%) were towards over-estimation (i.e. a prolonged interval between reported last menses and conception). This group had a significantly higher rate of preterm delivery (OR 1.7, 95% CI 1.5–1.9).

Table 3 shows a comparison of menstrual-scan dates discrepancy in various categories of pregnancy characteristics, gestation lengths and birthweights. The skewness towards over-estimation of dates by last menstrual period can be seen in all categories, and is pronounced in young and short mothers, smokers and mothers with a prior history of preterm delivery. In the gestational age at birth comparisons, the agreement of the two methods within ± 7 days is higher for term than preterm deliveries (72.6%–vs 62.8%). This is mainly due to a high degree of over-estimates of menstrual dates in each of the preterm gestation categories, with overall 33.6% of preterm births having had a prolonged last menstruation to conception interval. Table 3 also shows that a prolonged menstruation to conception interval was associated with a lower mean birthweight (3274 g), compared with when the discrepancy was within ± 7 days (3381 g; mean difference −107 g, 95% CI −126 to +88; P < 0.001). In contrast, the mean birthweight was not different when menstrual dates under-estimated the scan based gestational age by > 7 days (3401 g; mean difference +20 g, 95% CI −18 to +57; P= 0.33). But pregnancies with a prolonged menstruation to conception interval were not more likely to result in small-for-ges-tational-age babies (501/4846, 10.3%), compared with all other pregnancies (1556/15,933, 9.8%) (OR 1.07, 95% CI 0.96 to 1.18; P= 0.13). This indicates that the lower birthweight in pregnancies with a prolonged last menstruation to conception interval was due to the higher rate of prematurity.

Table 3.  Discrepancy between menstrual dates and ultrasound scan dates in various categories. Values are given as %, unless otherwisse indicated.
 Menstrual-scan dates discrepancy
 ≤−7 daysWithin ±7 days≥+7 daysTOTAL
Total (n (%))1001 (4.8)14,933 (71.9)4845 (23–3)20,779 (100.0)
Maternal age    
Maternal height (cm)    
  ≤ 1554.969.525.6100.0
Maternal weight (kg)    
Body mass index (kg/m2)    
  < 20.04.369.226.5100.0
Ethnic group    
Cigarette smoking    
History of abortion    
History of preterm birth    
Gestation at delivery    
  Mean [SD] days279 [14]277 [13]272 [14]276 [14]
  ≤ 34 weeks3.862.433.8100.0
  34–36 weeks3.662.933.5100.0
  ≤ 37 weeks3.762.833.6100.0
  >37 weeks4.972.622.5100.0
Birthweight (g)    
  Mean [SD]3401 [555]3381 [557]3274 [588] 
  ≤ 10th centile5.370.324.4100.0
  > 10th centile4.872.023.2100.0

Tables 4 and 5 show adjusted odds ratios resulting from logistic regression for nulliparae and multiparae, respectively, with preterm delivery as the dependent variable, stratified in three categories (24–33 weeks, 34–36 weeks, < 37 weeks). A separate analysis was also run for deliveries following spontaneous onset of labour, which constituted the majority of all births (78.3%, Table 2) and of preterm births (1064/1485 = 71.6%). Analysis of spontaneous labours revealed the same variables with similar odds ratios for preterm birth, but wider confidence intervals for some of the borderline variables which was likely due to lower numbers. The results are not detailed here but are available on request. There were insufficient numbers of nonspontaneous onset deliveries (i.e. iatrogenic: induction of labour or elective caesarean section) for multivariate analysis.

Table 4.  Logistic regression analysis of factors influencing preterm delivery in nulliparae. Adjusted odds ratios of significant factors are listed. Categorical variables: maternal age (≤20, >35); height (<155, >172); weight at first visit (<52, >80); ethnic group (non-European); cigarette smoker (any); alcohol drinker (mild or moderate); history of abortion; menstrual-scan dates discrepancy (≤−7 days =‘shortened MCI’; ≥+7 days: ‘prolonged MCI’). MCI = menstruation-conception interval, derived from the difference in gestational age between menstrual dates and ultrasound dates.
Nulliparae (n= 9086)PAdjusted odds ratio95% CI
Early pretenn (24–33 weeks)   
  Maternal weight ≥ 80 kg0.0131.781.13–2.80
  Prolonged MCI0.0221.451.06–1.99
Late preterm (34–36 weeks)   
  Maternal age ≤ 20 years0.0051.431.15–1.82
  Maternal weight ≤ 52 kg0.0031.521.15–2.01
  History of abortion0.0131.351.07–1.72
  Prolonged MCI≤ 0.0011.551.26–1.90
All preterm (≤ 37 weeks)   
  Maternal age ≤ 20 years0.0071.341.08–1.65
  Maternal weight < 52 kg0.0061.401.10–1.79
  Maternal weight > 80 kg0.0451.311.01–1.72
  Prolonged MCI≤ 0.0011.521.28–1.82
Table 5.  Logistic regression analysis of the factors influencing preterm delivery in multiparae. Variables entered were the same as in Table 4, plus ‘history of preterm delivery’. MCI = menstruation-conception interval, derived from the difference in gestational age between menstrual dates and ultrasound dates.
Multiparae (n= 11,983)PAdjusted odds ratio95% CI
Early pretenn (24–33 weeks)   
  Cigarette smoker0.0131.471.09–1.99
  History of abortion0.0041.551.15–2.09
  History of preterm delivery≤ 0.0015.073.40–7.57
  Prolonged MCI≤ 0.0011.911.42–2.57
Late pretem (34–36 weeks)   
  Maternal age ≤ 20 years0.0321.581.04–2.40
  Maternal weight ≤ 52 kg≤ 0.0011.801.41–2.30
  Cigarette smoker≤ 0.0011.921.61–2.30
  History of preterm delivery≤ 0.0012.551.85–3.53
  Prolonged MCI≤ 0.0011.881.57–2.25
All preterm (≤ 37 weeks)   
  Maternal age≤ 20 years0.0261.541.05–2.26
  Maternal height ≤ 155 cm0.0431.241.01–1.52
  Maternal weight ≤ 52 kg≤ 0.0011.691.36–2.10
  Cigarette smoker≤ 0.0011.821.55–2.13
  History of abortion0.0191.211.03–1.43
  History of preterm delivery≤ 0.0013.272.52–4.26
  Prolonged MCI≤ 0.0011.861.59–2.18

Overall, the results of the logistic regression confirm those of the univariate analysis (Table 2), but highlight differences in the various parity and gestational age groups. Young, short or underweight mothers were more likely to have preterm babies regardless of parity. Maternal obesity also increased prematurity rates, but in nulliparae only. Smoking was associated with increased preterm births in multiparae but not in nulliparae. The non-European group had increased preterm rates for both nulliparae and multiparae; analysis within ethnic subgroups showed that significance was reached in the Indian/Pakistani group for multiparae only (P= 0.013, adjusted OR 1.39,95% CI 1.07–1.81), while in the Afro-Caribbean group significance was reached for nulliparae only (P= 0.046, adjusted OR 1. 61, 95% CI 1.01–2.56). Preterm delivery in a previous pregnancy had the strongest association with preterm delivery in the index pregnancy, especially for births before 34 weeks (adjusted OR 5.07, CI 3.40–7.57). Moderate alcohol intake, on the other hand, appeared to be associated with a reduced rate of preterm delivery (Table 2), and this reached significance for nulliparae only in the logistic regression analysis (P= 0.025, adjusted OR 0.81, 95% CI 0.68–0.97).

A prolonged menstruation to conception interval was significantly associated with preterm birth in each gesta-tional age and parity category. Despite the strong association of this variable with other factors which themselves are related to prematurity, such as young age, short height, smoking and prior history of preterm delivery (Table 3), the multivariate analysis showed that it was also an independent risk factor for preterm birth. In contrast, a shortened menstruation to conception interval showed no such associations.

In Table 6, the significant determinants for preterm delivery are compared according to their aetiologic fraction (i.e. the population attributable risk)15.

Table 6.  Population attributable risk (PAR) for variables associated with preterm delivery. MCI = menstruation-conception interval, based on the difference in gestational age between menstrual dates and ultrasound dates.
 Prevalence (%)Adjusted odds ratioPAR (%)*
  1. * PAR = 100 × (AOR - 1) I [AOR + (100/Pr) - 1] (where AOR = adjusted odds ratio and Pr = prevalence).

Early preterm (< 34 weeks) Nulliparae   
  Maternal weight ≥ 80 kg9.081.786.61
  Prolonged MCI22.761.459.26
  Cigarette smoker24.001.4710.14
  History of abortion21.971.5510.72
  History of preterm delivery2.125.077.93
  Prolonged MCI22.991.9117.27
  All preterm (< 37 weeks) Nulliparae   
  Age < 20 years15.011.344.79
  Weight ≤ 52 kg10.891.404.21
  Weight ≥ 80 kg9.081.312.77
  Prolonged MCI22.761.5210.66
  Age < 20 years2.361.541.27
  Height ≤ 155 cm12.631.242.92
  Weight ≤ 52 kg9.151.695.91
  Cigarette smoker25.681.8217.38
  History of abortion26.431.215.35
  History of preterm delivery3.663.277.69
  Prolonged MCI23.171.8616.57


  1. Top of page
  2. Abstract
  7. References

To our knowledge this is the first study to investigate early pregnancy risk factors for preterm birth as defined by estimated expected date based on ultrasound scan. Scan dates are more reliable for deriving the actual gestation and predicting the expected date of confinement, and their use refines the accuracy of the preterm category. Ultrasound dates increase the apparent rate of preterm delivery compared with menstrual dates, presumably because a proportion of menstrual dates overestimate the true gestational age.

The findings suggest that early pregnancy risk factors for preterm birth vary in their effect according to parity and degree of prematurity, and emphasise that the heterogeneity of prematurity requires stratified analysis. Young, short or underweight mothers have increased rates of preterm delivery but for nulliparae, being overweight also constitutes an increased risk, especially of early prematurity (< 34 weeks). This may reflect known associations between pre-eclampsia and maternal obesity16. Pre-pregnant obesity has also been linked with preterm delivery17. We included an assessment of body mass index in the univariate analysis and found it to show similar results to maternal weight. For the logistic regression, we used height and weight separately as we agree with a previous report that for risk assessment in pregnancy, body mass index offers no advantage over maternal weight alone18.

Comparisons of ethnic group in this population are consistent with previous observations that Indian/Pakistani and Afro-Caribbean women have shorter gestations19. Both these groups had higher rates of prematurity, although the relatively low numbers in the subgroups precluded more detailed assessment. No attempt is made in the routinely collected obstetric database to classify according to social class. This variable may have a separate effect on prematurity, although it has also been suggested that social class is not a significant factor once variables such as maternal size and smoking are taken into account20,21.

Smoking was strongly associated with preterm delivery in multiparae but not in nulliparae. This finding is in agreement with previous observations that the effect of smoking is strongest in multiparae22 and that a negative association in nulliparae may be due to a protective effect of smoking on pre-eclampsia16. Previous reports of the effect of alcohol on preterm delivery have been inconclusive23 or found it to have a mildly protective effect24. Our analysis shows differing effects according to parity: for primiparae only, moderate alcohol consumption appeared to have a significant protective effect. Further work on larger databases is required to explore the validity and nature of this interesting association. It must also be noted that in our population too few women had, or admitted to, regular/heavy alcohol consumption to be able to make any statement on its effects on prematurity.

A history of abortion was associated with preterm delivery, which is in keeping with previous reports21,25. Our data did not differentiate between spontaneous miscarriage and induced abortion, but both have been found to be risk factors to a similar degree26. A past history of preterm delivery was the strongest factor, with an overall OR of 3.1 (Table 2) which is similar to the OR of 3.4 reported in a Norwegian population27.

The use of routine scan dates in preference to specified menstrual dates has allowed the definition of a new variable which expresses the discrepancy between them. The data show a hitherto unrecognised link with preterm delivery which applies in varying degrees across all gestational age and parity categories. An unknown, uncertain or unspecified menstrual history is known to be linked to adverse socioeconomic factors and has been shown to be associated with adverse outcome including premature birth10,28–30. An association between prematurity and unspecified dates is also apparent here, as shown in Table 1, where this category shows a preterm rate of 9.4%. Even after exclusion of these cases, leaving a dataset with specified, ‘certain’ last menstrual periods, an increased discrepancy in either direction between menstrual and scan dates was seen in young mothers and smokers (Table 3). The multivariate analysis however shows an association with preterm delivery only when the error is towards over-estimation of menstrual dates compared with scan dates (i.e. a prolongation of the menstruation to conception interval).

While the ultrasound scan is now well established as a more reliable method of dating pregnancy than the menstrual history9, scan error nevertheless does exist. It is conceivable that an under-estimation of true gestation by scan would artificially increase the prematurity rate as calculated by scan dates, while an over-estimation by scan would appear to reduce the prematurity rate. However, we know from studies of assisted conception pregnancies that the measurement error in routine scan dating is narrow6,8 and our limits of ± 7 days represent at least a 90% confidence interval. We also know that there are no significant differences in second trimester biparietal measurements between singletons and twins, or babies who are subsequently born preterm compared with term31. A discrepancy in dates noted in early pregnancy does not usually affect management and in any case the observed increased prematurity rates applied also to spontaneous onset preterm deliveries. Furthermore, the increased proportion of pregnancies which had a menstrual-scan date discrepancy of > +7 days was seen in each of the preterm gestation categories (Table 3). Finally, the high prematurity rate for these pregnancies (10.2%, Table 2) was elevated regardless of whether the discrepancy between menstrual and scan dates was 1–2 weeks (10.3%), 2–3 weeks (10.6%), 3–4 weeks (9.5%) or over 4 weeks (10.1%). It is therefore unlikely that our observations are artifactual or explained by scan error.

The ratio of over-compared with under-estimation by last menstrual period by more than one week in our sample (23.3 vs 4.8%, Table 2) is similar to previous reports32. Menstrual dates are always more likely to over-estimate rather than under-estimate true gestational age5, which is due to delayed ovulation within the conception cycle33. The increased rate of prolonged menstruation to conception interval in pregnancies ending in preterm birth may be a result of delayed ovulation, or anovulation prior to the conception cycle. In a study of abortion rates of infertile couples monitored with basal body temperature charts, it was observed that spontaneous abortion was twice as likely if conception occurred several days before or after mid-cycle34. The adverse outcome, however, resulted mainly from conceptions which were delayed, as these were four times more frequent than those which were early. The same study also showed a higher rate of spontaneous abortions if the conception occurred after one or more anovulatory cycles, suggesting a further association between early pregnancy loss and prolonged last menstruation to conception interval. It is possible that some cases of preterm birth represent the other end of a clinical spectrum of poor placentation or other underlying pathology.

These considerations are also relevant to a recent report35 which claims that a lower than expected first trimester ultrasound measurement of crown-rump length is associated with higher preterm rates and lower birthweight, because it is an indication of early growth restriction. The distribution of the observed discrepancy was heavily skewed, with as many as 38% of crown rump length measurements being between two and six days lower than expected. An alternative explanation would be that these pregnancies do not in fact have a high rate of early growth restriction, but that the lower than expected crown-rump length is a manifestation of wrong menstrual dates. Delay of ovulation past mid-cycle is common, and would result in the embryo or fetus being younger at the time of the scan than would be predicted by the menstrual dates. These pregnancies would thus have a longer menstruation to conception interval which could, as in the current study, account for their increased likelihood of prematurity and low birthweight.

The effect of known variables on outcome measures such as prematurity can be studied by the aetiology fraction or population attributable risk15, which takes prevalence into account (Table 6). This analysis shows that even though the relative risk of a prolonged menstruation to conception interval is not the highest, it has a high prevalence (23.3%) which makes it the strongest risk factor for preterm delivery in this analysis of variables known at the beginning of pregnancy. Smoking has a similarly high prevalence and a high population attributable risk value, but for multiparae only. A past history of preterm delivery has the highest relative risk for prematurity, but has a low prevalence and hence has a smaller actiological fraction.

One limitation of our data was that it did not contain second trimester alpha-fetoprote in results, and we were therefore unable to include this in the multivariate analysis. Serum tests are usually sent from community clinics to the laboratory before the mother comes to the hospital, and the results are not always transferred to computer. A high second trimester level of alpha-fetoprotein is known to be associated with adverse outcome including preterm delivery36,37, but the result may also be false positive when the gestational age is wrong. An over-estimation of dates (due to prolonged menstruation to conception interval) would however be more likely to lead to the test being false negative, which could be a contributing factor to the observed low predictive value of serum alpha-fetoprotein screening for low birthweight38. Furthermore, the prevalence of positive alpha-fetoprotein tests is relatively low in low risk (1.9%) and high risk populations (3.8%)37. It would, nevertheless, be of interest to study the associations of high level of alpha-fetoprotein and menstrual dates discrepancy on preterm delivery rates.

Another variable which we were not able to consider was vaginal bleeding in early pregnancy, a known risk factor for preterm delivery39,40. A difficulty with this variable is that it is often unreported or unrecorded, which may explain that its apparent prevalence was as low as 1% in an unselected population39. The bleeding episode may be confused with menstrual bleeding41 and this would result in menstrual dates under-estimating gestational age. A negative menstruation to conception interval was however not associated with ultrasound based prematurity in this study (Table 2).

With the advent of routine ultrasound, more pregnancies are accurately dated. There is now ample evidence that scan dates should be used in preference to even ‘certain’ menstrual dates. However, information on the last menses can also be used—not for dating, but for screening, and a positive discrepancy between menstrual dates and scan dates should act as a warning that the pregnancy is at increased risk of premature birth.


  1. Top of page
  2. Abstract
  7. References