Background: There is limited evidence to support the use of customised centile charts to identify those at risk of stillbirth and infant death at term. We sought to determine birth weight thresholds at which mortality and morbidity increased and the predictive ability of noncustomised (accounting for gestational age and sex) and partially customised centiles (additionally accounting for maternal height and parity) to identify fetuses at risk.
Methods: This is a population-based linkage study of 979,912 term singleton pregnancies in Scotland, United Kingdom, between 1992 and 2010. The main exposures were noncustomised and partially customised birth weight centiles. The primary outcomes were infant death, stillbirth, overall mortality (infant and stillbirth), Apgar score <7 at 5 min, and admission to the neonatal unit. Optimal thresholds that predicted outcomes for both non- and partially customised birth weight centiles were calculated. Prediction of mortality between non- and partially customised birth weight centiles was compared using area under the receiver operator characteristic curve (AUROC) and net reclassification index (NRI).
Findings: Birth weight ≤25th centile was associated with higher risk for all mortality and morbidity outcomes. For stillbirth, low Apgar score, and neonatal unit admission, risk also increased from the 85th centile. Similar patterns and magnitude of associations were observed for both non- and partially customised birth weight centiles. Partially customised birth weight centiles did not improve the discrimination of mortality (AUROC 0.61 [95%CI 0.60, 0.62]) compared with noncustomised birth weight centiles (AUROC 0.62 [95%CI 0.60, 0.63]) and slightly underperformed in reclassifying pregnancies to different risk categories for both fatal and non-fatal adverse outcomes (NRI -0.027 [95% CI -0.039, -0.016], p < 0.001). We were unable to fully customise centile charts because we lacked data on maternal weight and ethnicity. Additional analyses in an independent UK cohort (n = 10,515) suggested that lack of data on ethnicity in this population (in which national statistics show 98% are white British) and maternal weight would have misclassified ~15% of the large-for-gestation fetuses.
Conclusions: At term, birth weight remains strongly associated with the risk of stillbirth and infant death and neonatal morbidity. Partial customisation does not improve prediction performance. Consideration of early term delivery or closer surveillance for those with a predicted birth weight ≤25th or ≥85th centile may reduce adverse outcomes. Replication of the analysis with fully customised centiles accounting for ethnicity is warranted.