We have demonstrated a graded relation between fetal size at 10–24 weeks of gestation and pregnancy duration. Our data showed that fetuses with a second trimester size smaller than expected for their duration of gestation tended to have longer pregnancies compared to large fetuses (12 days between z-score -2.5 and +2.5 for AC) (Table 1). The slow growth of the small fetus is combined with an extended pregnancy duration that does not fully compensate for slow growth as shown by their lower birth weight. When our analysis adjusted for the difference between LMP and ultrasound assigned gestational age (using HC at 10–24 weeks), the effect of fetal size on pregnancy duration was reduced to a smaller but still significant difference; for AC the adjusted difference between z-score -2.5 and +2.5 was 6.4 days (Table 1). Such an adjustment may represent an over-correction of the results reducing biological variation in size and growth.
Adjusting for the difference between LMP- and ultrasound-based (HC) gestational age could possibly represent a confounder, but this is not obvious and in fact the effect of fetal size on pregnancy length is even stronger when not controlling for fetal head size in the second trimester. Since clinicians tend to use ultrasound to adjust fetal age, there is a risk of bias when fetuses with larger heads are assigned to a more advanced gestational age and possibly more inductions in the post-term period. Conversely, with the continuous focus that clinicians have on intra-uterine growth-restriction, it is more likely that the normally growing small fetus is at increased risk of induction and caesarean section. When including also such pregnancies in the analysis, an effect of fetal size on pregnancy length was still present (results not shown).
Several investigators have reported that fetuses smaller than expected in the second trimester have an increased risk of adverse obstetrical outcomes such as low birthweight  and premature birth [16, 17]. These studies were population based and included both normal and growth restricted fetuses. A recent study of pregnancies following in vitro fertilisation demonstrated an increased risk for small for gestational age infants and premature birth (birth < 37 weeks of gestation), and a shorter duration of pregnancy when crown-rump length was less than expected at an early ultrasound scan ; however, the study did not take account of iatrogenic reduction of pregnancy duration by caesarean section or induction of labour. In contrast to the mentioned studies, the present study was based on a healthy population of low-risk women who spontaneously went into labour and the overall obstetric outcome was good, with a low incidence of low birth weight and premature births (Table 1). Thus, smaller fetuses in the present study can be regarded to be within normal biological variation, not growth restricted, and interestingly they had longer, not shorter duration of pregnancies.
Our results are supported by a recent study of the heritable component of duration of pregnancy . In that study, the father and mother's own gestational age at birth were associated with the offspring's gestational age, and fathers with higher birth weights had larger offspring with shorter gestational length, which is in line with our results. The results of all the studies mentioned above fit with a U-shaped relationship between fetal growth and duration of pregnancy; rapidly growing fetuses tend to have shorter duration, slower growing fetuses a longer duration of pregnancy while pathologically slow growth have increased rates of birth before 37 weeks of gestation.
Six factors may influence our results: uncertain LMP , variation in ovulation and implantation [6, 20–22], early growth restriction , random error of the ultrasound measurement  and biological variation in fetal size. In the present study the participants knew the exact date of their LMP and had a history of regular menstrual periods (28 ± 4 days) for at least three months before pregnancy. Extended duration of pregnancy could possibly compensate for underestimated age assessment due to delayed ovulation and implantation, but the present study showed that despite the extended duration of their pregnancies, fetuses found to be small in the second trimester also had lower birth weight.
The error of ultrasound measurement of the fetal head can be reduced by repeated measurements. We took the average of three measurements, as a result of which the error, counted in gestational days, is small (95% CI -1.5;1.5) .
According to the study protocol eight participants were excluded due to a discrepancy of more than 14 days between LMP and ultrasound dating. The choice of 14 days as cut off was based on reports that there is an increased risk of growth-restriction and adverse outcome in such pregnancies [24–26] and we aimed to study a low-risk population.
Wilcox et al. found that "the fertile window" starts ~5 days before ovulation and includes the day of ovulation with the highest probability of conception on the last day before ovulation . Although this "fertility window" may occur within a wide range of a regular cycle, the follicular phase in natural cycles is fairly constant, with an SD of 3 days for most fertile women , suggesting that part of the variation (possibly half, i.e. ~6 days) that we see in pregnancy duration is attributable to ovulation variation, the rest being biological variation linking growth to pregnancy duration.
Discrepancies between gestational age assessed by a regular LMP and ultrasound may also be due to delayed implantation. There are data suggesting that later implantation is associated with an increased risk of early pregnancy loss , and that a large LMP-ultrasound discrepancy is associated with growth restriction and premature labour [24–26], possibly due to late implantation . We believe these results represent extreme conditions that lead to abnormal pregnancies. The present results suggest that within physiological ranges fetal growth may be slower and then associated with an extended pregnancy, a phenomenon that is also known from animal studies . Nutritionally restricted pregnancies resulted in slower growth and longer pregnancies.
The ultrasound method is better than certain menstrual history in predicting the date of spontaneous delivery [9, 10], but ultrasound dating disregards biological variation in growth and pregnancy length. For fetuses smaller than the mean, ultrasound shifts this group to an artificially lower gestational age, and vice versa for fetuses larger than the mean. For example, differences of 2.5 days between male and female with regard to gestational age assessment based on ultrasound, have been reported . This systematic shift leads to an artificially higher number of births defined as post-term among boys compared to girls [11, 28], with a correspondingly higher rate of induction of labour in pregnancies with male fetuses. We believe that a similar error occurs when using ultrasound in normal pregnancies with certain information of a regular LMP, the consequence being an underestimation of gestational age and pregnancy duration in those with smaller head size than the mean, and vice versa for fetuses above the mean.