SELECTION BIASES INVALIDATE CURRENT LOW-BIRTH-WEIGHT WEIGHT-FOR-GESTATION STANDARDS

Objective To obtain unbiased estimates of the variation of birthweight with gestation in infants born before 32 weeks of gestation. Setting The former Northern Regional Health Authority. Design Information on b irthweight was collected during a collaborative study of every registe red and unregistered birth at 22 to 31 weeks of gestation in the regio n in 1983 and 1990 to 1991. These birthweights were then related to co mputer-generated Tyneside norms for all registered births at 28 to 42 weeks of gestation between 1984 and 1991. Some local information was a lso collected on fetal weight after termination of pregnancy on social grounds at 16 to 21 weeks of gestation. Results Weight centiles const ructed after excluding infants with a gross, externally visible, malfo rmation and those dying before the onset of labour suggest that previo usly published European standards have overestimated birthweight in in fants < 28 weeks of gestation, some low centiles being 30% in error. F emale and first-born infants weighed 4% less than their male and later -born counterparts at all gestations studied. A single correction fact or can therefore be used to correct for sex and parity, eliminating th e need for separate centile graphs. Twin pregnancy was associated with a 10% reduction in mean birthweight in pregnancies lasting < 37 weeks , and this difference increased progressively in pregnancies lasting l onger than this. Conclusion The small number of low birthweight infant s in previous datasets and the selective exclusion of all nonregistere d births have made previous second trimester weight-for-gestation norm s unreliable.