Size at birth by gestational age and hospital mortality in very preterm infants: Results of the area-based ACTION project

•In very preterm singletons GA-adjusted mortality decreased with increasing birthweight centile.•Lowest values of GA-adjusted mortality were reached between the 50th and the 89th centiles.•Mortality raised again ≥ the 90th centile.•Continuous measures of neonatal size by GA may be more appropriate than the usual 10th centile cut-off.

BackgroundSize at birth is an important predictor of neonatal outcomes, but there are inconsistencies on the definitions and optimal cut-offs.AimsThe aim of this study is to compute birth size percentiles for Italian very preterm singleton infants and assess relationship with hospital mortality.Study designProspective area-based cohort study.SubjectsAll singleton Italian infants with gestational age 22–31 weeks admitted to neonatal care in 6 Italian regions (Friuli Venezia-Giulia, Lombardia, Marche, Tuscany, Lazio and Calabria) (n. 1605).Outcome measureHospital mortality.MethodsAnthropometric reference charts were derived, separately for males and females, using the lambda (λ) mu (μ) and sigma (σ) method (LMS). Logistic regression analysis was used to estimate mortality rates by gestational age and birth weight centile class, adjusting for sex, congenital anomalies and region.ResultsAt any gestational age, mortality decreased as birth weight centile increased, with lowest values observed between the 50th and the 89th centiles interval. Using the 75th–89th centile class as reference, adjusted mortality odds ratios were 7.94 (95% CI 4.18–15.08) below 10th centile; 3.04 (95% CI 1.63–5.65) between the 10th and 24th; 1.96 (95% CI 1.07–3.62) between the 25th and the 49th; 1.25 (95% CI 0.68–2.30) between the 50h and the 74th; and 2.07 (95% CI 1.01–4.25) at the 90th and above.ConclusionsCompared to the reference, we found significantly increasing adjusted risk of death up to the 49th centile, challenging the usual 10th centile criterion as risk indicator. Continuous measures such as the birthweight z-score may be more appropriate to explore the relationship between growth retardation and adverse perinatal outcomes.