Objective While customised centiles bundle a set of physiological characteristics together to adjust the standard for birthweight and growth potential, the authors wanted to investigate the role which individual variables have in improving the fit, that is in predicting the expected birthweight.

Method The authors studied a database of 313 285 consecutive Swedish births with complete information about maternal height, weight, parity, ethnicity and sex of the baby, as well as pathological factors including smoking, hypertension and diabetes. Gestational age was controlled for by deriving residuals of birthweight, with goodness of fit expressed as R². To compare the effect of the variables within an optimal (central) and the more pathological (outlying) birthweight ranges, the database was divided into three equal-size tertiles.

Results The RR² is plotted against the number of variables added. The fit (RR²) was best in the middle tertile (T2), rising from 0.28 with adjustment for sex only, to 0.73 with all variables included. In the lowest tertile (T1), adjustment for physiological factors similarly increased the prediction in a stepwise manner, but to a lesser extent. The same was observed in the highest tertile of birthweight (T3), where maternal weight showed a steep increase in RR². In each case, ethnicity added little to the overall fit, as the various ethnic groups in this database contributed less than 10% to the total population.

Conclusion The stepwise improvement of birthweight prediction (RR²) provides further evidence for the advantage of using customised centiles.