Ultrasonographic fetal weight estimation by an integrated computer-assisted system: can each laboratory improve its accuracy?

Estimated fetal weights from 1684 cases with singleton, live infants born within 7 days of an ultrasonographic examinations were compared with their birth weights, which ranged from 520 to 5920 gm. Estimated weight calculated from Shepard's equation produced a linear relationship of birth weight against estimated weight with an intercept and slope of 209.5 gm and 0.929. Overall, 75% of the estimated weights were within 15% of the actual weight. A separate regression model that used the same variables from Shepard's equation failed to improve the comparison between estimated and actual weights. All cases were reevaluated with multiple regression modeling. Various examination-to-birth intervals were analyzed; intervals less than or equal to 3 days appeared optimal. The best equation increased the percent of cases that were within 15% of the actual birth weight to 80%. The plot of birth weight against estimated weight had an intercept and slope of 33.1 gm and 0.994. Unlike Shepard's equation, the best equation was not statistically different from the ideal one-to-one relationship between estimated and actual weight. When the new equation was applied to an additional 339 new cases, equally accurate results were obtained. Customizing a laboratory's ultrasonographic weight estimation equations may be necessary to obtain the best estimate of fetal weight.