A global reference for fetal-weight and birthweight percentiles

Background Definition of small for gestational age in various populations worldwide remains a challenge. References based on birthweight are deficient for preterm births, those derived from ultrasound estimates might not be applicable to all populations, and the individualised reference can be too complex to use in developing countries. Our aim was to create a generic reference for fetal weight and birthweight that overcame these deficiencies and could be readily adapted to local populations. Methods We used the fetal-weight reference developed by Hadlock and colleagues and the notion of proportionality proposed by Gardosi and colleagues and made the weight reference easily adjustable according to the mean birthweight at 40 weeks of gestation for any local population. For application and validation, we used data from 24 countries in Africa, Latin America, and Asia that participated in the 2004-08 WHO Global Survey on Maternal and Perinatal Health (237025 births). We compared our reference with that of Hadlock and colleagues (non-customised) and with that of Gardosi and colleagues (individualised). For every reference, the odds ratio (OR) of adverse perinatal outcomes (stillbirths, neonatal deaths, referral to higher-level or special care unit, or Apgar score lower than 7 at 5 min) for infants who were small for gestational age versus those who were not was estimated with multilevel logistic regression. Findings OR of adverse outcomes for infants small for gestational age versus those not small for gestational age was 1.59 (95% CI 1.53-1.66) for the non-customised fetal-weight reference compared with 2.87 (2.73-3.01) for our country-specific reference, and 2.84 (2.71-2.99) for the fully individualised reference. Interpretation Our generic reference for fetal-weight and birthweight percentiles can be easily adapted to local populations. It has a better ability to predict adverse perinatal outcomes than has the non-customised fetal-weight reference, and is simpler to use than the individualised reference without loss of predictive ability.