Background: Although many studies have compared birth-weight charts to determine which better identify infants at risk of adverse perinatal outcomes, less attention has been given to the threshold used to define small or large for gestational age (SGA or LGA) infants. Our aim was to explore different thresholds associated with increased risk of adverse perinatal outcomes using population, customised, and Intergrowth centile charts.
Methods and findings: This is a population-based cohort study (Swedish Medical Birth Registry), which included term singleton births between 2006 and 2015 from women with available data on first-trimester screening. Population, customised, and Intergrowth charts were studied. Outcomes included cesarean section, postpartum haemorrhage, severe perineal tear, Apgar score at 5 minutes, neonatal morbidity, and perinatal mortality. Odds for each outcome were assessed in intervals of 5 centiles of birth weight (reference being 40th-60th centiles) using logistic regression. Intervals of 5% of the population were also explored. Sensitivity for fixed false-positive rates (FPRs) was reported for neonatal outcomes. Data from 212,101 births were analysed. Mean age was 33 ± 5 years, 48% of women were nulliparous, and 80% were born in Sweden. Prevalence of SGA (<10th centile) was 10.1%, 10.0%, and 3.1%, and prevalence of LGA (>90th centile) was 10.0%, 8.2%, and 25.1%, assessed using population, customised, and Intergrowth charts, respectively. In small infants, the risk of perinatal mortality was consistently increased below the 15th, 10th, and 35th birth-weight centiles for the respective charts (odds ratio [OR] 1.59, 95% confidence interval [CI] 1.05-2.39, p = 0.03 for 10th-15th population centile; OR 2.54, 95% CI 1.74-3.71, p < 0.001 for 5th-10th customised centile; OR 1.81, 95% CI 1.07-3.04, p = 0.03 for 30th-35th Intergrowth centile). The strength of association with adverse perinatal outcomes was different between infants below the 5th birth-weight centile for each chart (OR 4.47, 95% CI 3.30-6.04, p < 0.001 for the population chart; OR 5.78, 95% CI 4.22-7.91, p < 0.001 for the customised chart; OR 10.74, 95% CI 7.32-15.77, p < 0.001 for the Intergrowth chart) but similar in the smallest 5% of the population (OR 4.34, 95% CI 3.22-5.86, p < 0.001 for the population chart; OR 5.23, 95% CI 3.85-7.11, p < 0.001 for the customised chart; OR 4.69, 95% CI 3.47-6.34, p < 0.001 for the Intergrowth chart). For a fixed FPR of 10%, different thresholds for each chart achieved similar sensitivity for perinatal mortality in small infants (29% for all charts). Similar behaviour of different thresholds and similar risk/sensitivity for fixed FPR were observed in relation to other outcomes and for LGA infants. Limitations of this study include the relative homogeneity of the Swedish population, which limits generalisability to other populations; customised centiles may perform differently in populations with increased heterogeneity of ethnic background.
Conclusions: The risk of adverse outcomes was consistent across proportions of the population but did not reflect fixed thresholds, such as the 10th or 90th centiles, across different growth charts. Chart-specific thresholds for the population should be considered in clinical practice.