Evaluation of ultrasound-estimated date of delivery in 17,450 spontaneous singleton births: do we need to modify Naegele's rule?

OBJECTIVE: To compare the size of error in the predicted date of delivery by biparietal diameter (BPD) and last menstrual period (LMP) in different clinical models.
MATERIALS AND METHODS: Predicted dates of delivery were calculated in 14,805 spontaneous deliveries with a reliable date of LMP using two assumptions: average length of pregnancy of 280 and of 282 days. Errors in these calculated dates were tested when used alone or combined, i.e. the LMP-predicted date of delivery was used as first choice unless the discrepancy between gestational age calculated by BPD and LMP exceeded 7, 10 or 14 days.
RESULTS: The average discrepancy (error) between predicted date of delivery from BPD and LMP and date of spontaneous delivery was 7.96 and 8.63 days, respectively (p < 0.0001). Adding 282 instead of 280 days to the first day of the LMP reduced the error of the LMP method from 8.63 to 8.41 days, reduced the percentage of classified post-term deliveries from 7.9 to 5.2% and increased the preterm births from 3.96 to 4.48%. No models of combined use of LMP and BPD were superior to the use of BPD alone.
CONCLUSION: If both BPD and LMP are available, BPD should be used to predict term. If only LMP is available, term should be calculated as the first date of the LMP plus 282.