Background: Preeclamptic-toxaemia (PET) (new-onset hypertension and proteinuria presenting after 20 weeks gestation) is a serious complication affecting 1 in 10 pregnancies. Annually, PET claims the lives of 50,000 mothers and 500,000 babies, making it one of the world’s deadliest pregnancy complications. Currently, birth of the baby is the only treatment and is associated with a significant risk of long-term infant morbidity and mortality. Accurate diagnosis and risk-stratification would reduce the enormous competing risks for mothers and babies, facilitating early intervention before severe complications have occurred. Therefore, there is an urgent, unmet need to develop accurate risk-stratification tools for PET.

Aims: To develop novel diagnostic and risk-stratification tools for PET.

Methods: Development of AI_PREMie tool.

Ethical approval was granted for AI_PREMie case/control study in three of Ireland’s largest maternity hospitals capturing 50% of all births in Ireland. AI_PREMie patent-pending biomarkers were quantified in plasma of 150 patients and together with extensive patient data were subjected to statistical analysis and machine learning (ML) in SAS^® Viya (Figure 1).

Results: A prototype champion model for AI_PREMie.

Previously, utilizing our platelet-based diagnostics platform, we identified novel PET biomarkers from platelets. Here, these biomarkers were quantified in plasma from an extended patient cohort (n=150). Statistical exploration of the biomarkers and patient data showed presence of complex non-linear relationships between variables and PET diagnosis, highlighting the need for sophisticated ML techniques. Thus, a pipeline containing a variety of different ML models was created. Support Vector Machine, with an Area under ROC of 0.839 was the best performing model (Figure 2) indicating the strong viability of our approach. This blood-based test offers significant advantage over current clinical methods, including an overall improvement in PET detection and eliminates the need for intensive staff training.

Conclusions: Our results demonstrate that platelet-uncovered biomarkers are easily translated into plasma-based tests. Furthermore, these biomarkers can diagnose and predict outcome in PET.

« Back to ISTH 2021 Congress

ISTH Congress Abstracts - https://abstracts.isth.org/abstract/ai_premie-a-novel-risk-stratification-tool-for-preeclamptic-toxaemia/