Agnostic identification of plasma biomarkers for postpartum hemorrhage risk

Background Postpartum hemorrhage (PPH) is difficult to predict, is associated with significant maternal morbidity, and is the leading cause of maternal mortality worldwide. The identification of maternal biomarkers that can predict increased PPH risk would enhance clinical care and may uncover mechanisms that lead to PPH. Objective This retrospective case-control study employed agnostic proteomic profiling of maternal plasma samples to identify differentially abundant proteins in controls and PPH cases. Study Design Maternal plasma samples were procured from a cohort of >60,000 participants in a single institution’s perinatal repository. PPH was defined as a decrease in hematocrit of ≥10% or receipt of transfusion within 24 hours of delivery. PPH cases (N=30) were matched by maternal age and delivery mode (vaginal or cesarean) with controls (N=56). Mass spectrometry was used to identify differentially abundant proteins using integrated peptide peak areas. Statistically significant differences between groups were defined by a p-value of <0.05 after controlling for multiple comparisons. Results By study design, cases and controls did not differ in race, ethnicity, gestational age at delivery, blood type, or pre-delivery platelet count. Cases had slightly, but significantly lower pre- and post-delivery hematocrit and hemoglobin. Mass spectrometry detected 1140 proteins, including 77 proteins for which relative abundance differed significantly between cases and controls (fold change >1.15, P<0.05). Of these differentially abundant plasma proteins, most had likely liver or placental origins. Gene ontology term analysis mapped to protein clusters involved in responses to wound healing, stress response, and host immune defense. Significantly differentially abundant proteins with the highest fold change (prostaglandin D2 synthase, periostin, and several serine protease inhibitors) did not correlate with pre-delivery hematocrit or hemoglobin, but identified PPH cases with logistic regression modeling revealing good-to-excellent area under the operator receiver characteristic curves (AUROC 0.802-0.874). Incorporating pre-delivery hemoglobin with these candidate proteins further improved identification of PPH cases. Conclusion Agnostic analysis of maternal plasma samples identified differentially abundant proteins in controls and PPH cases. Several of these proteins are known to participate in biologically plausible pathways for PPH risk and have potential value for predicting PPH. These findings identify candidate protein biomarkers for future validation and mechanistic studies.