Development of individual growth curve standards for estimated fetal weight: I. Weight estimation procedure.

In this investigation the weight estimation procedure of Rossavik was reassessed with particular emphasis on parameter estimation and performance over a wide weight range. Using a cross-sectional data set (193 patients), a longitudinal data set (20 patients), and an iterative procedure, parameter estimates were obtained based solely on regression analysis. Comparison of weight estimates obtained using a function based on these parameter values with actual birth weights indicated virtually no systematic errors over a 250-g to 4750-g weight range and random errors (+/- 1 SD) of 10% to 13% below 200 g and 6% to 8% above 2000 g. The weights of small- and large-for-gestational age fetuses were systematically overestimated (4.1%) and underestimated (-3.0%), respectively, but systematic errors were not found in average-for-gestational age fetuses. No differences in random errors were seen in these three growth categories. Comparison with other weight estimation procedures indicated that the Rossavik procedure gives weight estimates that are at least as accurate as those obtained with other methods over a wide range of weight and growth categories. However, the Rossavik procedure can be used to generate individual growth curve standards for weight estimates, a characteristic not shared with other weight estimation procedures.