Objective: To apply two different statistical methods to the live births of Lazio region in the years 2000-2003 and calculate updated estimate of the curves of percentiles of birth weight for gestational age.
Design and setting: The data analysed, including clinical-epidemiological and obstetric-neonatal information, have been extracted from birth certificates for the Italian region of Lazio for the period 2000-2003. Two methods of analysis, previously used in two Italian studies, have been applied. In the first method the observed distribution for each gestational week has been decomposed as a mixture of two normal distributions with homoscedastic components. The percentiles obtained, relative to the estimate of the component with the greater weight, have been then used in a third degree polynomial regression with gestational age as covariate, to estimate the final smoothed values. The second method assumes that, for a fixed gestational age, the position of the percentiles can be obtained as a polynomial function of Normal scores. Initially, for both sexes, the percentiles for every gestational age have been estimated using quantile regressions restricted to windows of three weeks. These have been subsequently smoothed by a polynomial model using the logarithm of the birth weight as the dependent variable and gestational age (up to the third degree), the z-score corresponding to the percentiles observed in the three week windows (up to the second degree), sex and the interaction between sex, gestational age and z-score as independent variables.
Results: data relating to 157,335 single live births to Italian women resident in the Lazio region have been analysed. The raw curves of the birth weight percentiles for the preterm babies are very irregular for the more extreme percentiles because of the very low number of deliveries of live pre-term babies. The two methods have produced similar results. However, the first method was less influenced by extreme and potentially anomalous values. The second method tended to estimate percentiles beyond 41a week of gestation inferior to those of the previous weeks. A comparison has been made with the curves obtained in the two previous Italian studies and the new estimated percentiles tend to be inferior for the very preterm live births.
Conclusions: this study has produced new curves of birth weight for gestational age. The analysis has been based on two different techniques, which have produced substantially similar results. For the tenth percentile, our estimates are lower than those obtained in previous studies suggesting a reduction in the late foetal mortality of babies who are not only pre-term but also of low weight. These results could be used in order to define new standards for birth weights in Italy.