Comparison with results of previous studies
In a series of previous first and second trimester studies for the prediction of stillbirth we highlighted that the causes of this adverse event are heterogeneous and that the focus of research should be placental dysfunction related stillbirths because they are relatively common and to a great extent potentially preventable17-22. However, a systematic review of 69 previous systematic reviews which aimed to identify variables that could be relevant to the development of a clinical prediction model for stillbirth treated this adverse event as a homogeneous condition.23 The study reported that no marker had useful screening performance, but maternal age, body mass index and history of prior adverse pregnancy outcomes had a more convincing association than the best performing tests, which were pregnancy-associated plasma protein-A (PAPP-A), placental growth factor (PlGF) and UtA-PI.23 Such types of publications that do not recognize the fact that the causes of stillbirth are heterogeneous could not possibly advance the development of strategies for prediction and prevention of stillbirth.
The same group of authors attempted to externally validate previously published prediction models for stillbirth using individual participant data (IPD) meta-analysis from a heterogeneous group of 19 datasets24. A literature search identified 40 stillbirth models, but they could only validate three of these models due to lack of availability of the necessary predictors in their dataset or the model equations in the previous publications; surprisingly for such a study there was no attempt to contact the authors of the models to request details on the equations. The authors reported that the three models showed poor and uncertain predictive performance in their data, they had limited clinical utility and that further research is needed to identify stronger prognostic factors and develop more robust prediction models 33. However, these conclusions are misleading and can have a potential adverse impact on clinical practice and future research, because first, two of the three models they evaluated were based on maternal risk factors only and they overlooked many prediction models based on a combination of maternal risk factors and first or second trimester biomarkers, second, the heterogeneous datasets used for their IPD meta-analysis were not derived from prospective screening for stillbirth and were therefore inadequate for assessing models derived from prospective examination of patients, and third, the authors examined the value of the reported models for prediction of all stillbirths and overlooked the fact that the original publications highlighted that the models provided good prediction of placental dysfunction related stillbirth, particularly those occurring preterm, rather than prediction of all stillbirths.
In our study we have focussed on placental dysfunction related stillbirth, prospectively recorded data from the maternal history and biomarkers shown over the last few decades to be associated with the birth of SGA neonates, developed and validated a model for prediction of SGA and demonstrated that such model can effectively predict a high proportion of stillbirths, especially those that occur preterm. We have previously reported the increased risk for SGA fetuses / neonates is provided by lower maternal weight and height, black, South and East Asian racial origin, medical history of chronic hypertension, diabetes mellitus and systemic lupus erythematosus or antiphospholipid syndrome, conception by in vitro fertilization or ovulation induction and smoking.4 For parous women variables from the last pregnancy that increased the risk for SGA were history of preeclampsia or stillbirth, decreasing birth weight z-score and decreasing gestational age at delivery of the last pregnancy and inter-pregnancy interval <0.5 years.4