Prediction of Small for Gestational Age Infants in Healthy Nulliparous Women Using Clinical and Ultrasound Risk Factors Combined with Early Pregnancy Biomarkers

Abstract

ObjectiveMost small for gestational age pregnancies are unrecognised before birth, resulting in substantial avoidable perinatal mortality and morbidity. Our objective was to develop multivariable prediction models for small for gestational age combining clinical risk factors and biomarkers at 15±1 weeks’ with ultrasound parameters at 20±1 weeks’ gestation.MethodsData from 5606 participants in the Screening for Pregnancy Endpoints (SCOPE) cohort study were divided into Training (n = 3735) and Validation datasets (n = 1871). The primary outcomes were All-SGA (small for gestational age with birthweight <10th customised centile), Normotensive-SGA (small for gestational age with a normotensive mother) and Hypertensive-SGA (small for gestational age with an hypertensive mother). The comparison group comprised women without the respective small for gestational age phenotype. Multivariable analysis was performed using stepwise logistic regression beginning with clinical variables, and subsequent additions of biomarker and then ultrasound (biometry and Doppler) variables. Model performance was assessed in Training and Validation datasets by calculating area under the curve.Results633 (11.2%) infants were All-SGA, 465(8.2%) Normotensive-SGA and 168 (3%) Hypertensive-SGA. Area under the curve (95% Confidence Intervals) for All-SGA using 15±1 weeks’ clinical variables, 15±1 weeks’ clinical+ biomarker variables and clinical + biomarkers + biometry /Doppler at 20±1 weeks’ were: 0.63 (0.59–0.67), 0.64 (0.60–0.68) and 0.69 (0.66–0.73) respectively in the Validation dataset; Normotensive-SGA results were similar: 0.61 (0.57–0.66), 0.61 (0.56–0.66) and 0.68 (0.64–0.73) with small increases in performance in the Training datasets. Area under the curve (95% Confidence Intervals) for Hypertensive-SGA were: 0.76 (0.70–0.82), 0.80 (0.75–0.86) and 0.84 (0.78–0.89) with minimal change in the Training datasets.ConclusionModels for prediction of small for gestational age, which combine biomarkers, clinical and ultrasound data from a cohort of low-risk nulliparous women achieved modest performance. Incorporation of biomarkers into the models resulted in no improvement in performance of prediction of All-SGA and Normotensive-SGA but a small improvement in prediction of Hypertensive-SGA. Our models currently have insufficient reliability for application in clinical practice however, they have potential utility in two-staged screening tests which include third trimester biomarkers and or fetal biometry.

Fetal Genotype and Maternal Glucose Have Independent and Additive Effects on Birth Weight

“…At present, prediction of birth weight prior to term gestation is difficult, and previous studies modelling clinical risk factors, serum biomarkers and fetal biometry measured by ultrasound to predict small for gestational age (SGA; birth weight less than the 10 th centile) and LGA achieved limited diagnostic performance (18, 19). Our preliminary analyses (unpublished data) suggest the fetal genetic score for birth weight has similar predictive ability to maternal fasting glucose.…”

External validation and clinical usefulness of first‐trimester prediction models for small‐ and large‐for‐gestational‐age infants: a prospective cohort study

“…Another limitation to be mentioned is that we had to exclude women who delivered between 16 +0 and 24 +6 weeks of gestation ( n = 8), as the Dutch population‐based reference curves for birthweight centiles are available from 25 weeks of gestation onwards . Lastly, we had to exclude two prediction models in the selection process, as we did not dispose of routine blood parameters (random glucose, rhesus group) and ultrasound measurements (crown–rump length) …”

“…49 Lastly, we had to exclude two prediction models in the selection process, as we did not dispose of routine blood parameters (random glucose, rhesus group) and ultrasound measurements (crown-rump length). 36,78…”

“…Biomarkers and biophysical tests may improve the accuracy of the model beyond using maternal characteristics alone. Published studies show only a limited contribution of these factors to improved discriminative performance, however . Moreover, most of these more complex predictors are relatively expensive, not readily available in general antenatal settings, and are possibly inconvenient for pregnant women .…”

“…Published studies show only a limited contribution of these factors to improved discriminative performance, however. [29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46] Moreover, most of these more complex predictors are relatively expensive, not readily available in general antenatal settings, and are possibly inconvenient for pregnant women. 28 To our knowledge, no external validation studies of prediction models for SGA or LGA have been published so far.…”