Efficient growth and development of the placenta is crucial to the success of a pregnancy. Conditions such as preeclampsia (PE) and intrauterine growth restriction (IUGR) are associated with abnormal placentation and as such are associated with high maternal and/or fetal morbidity and mortality. Currently, the only form of treatment for both conditions is delivery, often pre-term, which can severely impact the long-term health of the baby. Additionally, there are no screening tests available to predict at risk pregnancies before symptoms appear nor are there any that can provide guidance to clinicians with respect to timing of delivery. Therefore, the aim of our study is to identify novel predictive biomarkers that can detect pregnancies at risk of PE and IUGR. Serum samples were collected from women 26-41 weeks’ gestation with known outcomes of PE, IUGR, PE/IUGR or healthy pregnancies collected prior to and at term (< or >38 weeks gestation, respectively). We used a nuclear magnetic resonance (NMR) spectroscopy based metabolomics approach to identify metabolites (small molecules) that correlated with each disease. Initial spectral analysis segregated healthy and pathological pregnancies into distinct groups. We subsequently developed predictive models based on 25 metabolites. The resulting predictive model for distinguishing PE from healthy pregnancies had 93.5% probability of correctly identifying a PE pregnancy. The same metabolite panel was proven to provide robust and accurate prediction of PE in an independent validation cohort. This suggests that metabolic screening techniques can identify a metabolite model that could potentially be developed into an early predictive test to identify at risk pregnancies and/or screening tests to aid in clinical management of identified high risk pregnancies.