Introduction

Preeclampsia, a dangerous hypertensive disorder of third-trimester pregnancy, is a leading cause of maternal and neonatal mortality and morbidity worldwide1. It often presents as hypertension, edema, and severe proteinuria, and can escalate to hypertensive emergency and multisystem end-organ dysfunction1. Besides adverse peripartum outcomes, preeclampsia carries long-term consequences for women’s cardiometabolic health, including a well-documented association with increased risk for cardiovascular disease, worse quality of life, and shortened life expectancy2,3. In view of this growing body of evidence, improved preeclampsia risk prediction can help prevent serious peripartum and long-term complications in women.

Current risk prediction tools for preeclampsia rely on subjective clinical judgment about maternal history and in-office visits. However, prediction based on subjective clinical judgment alone is often less accurate than when combined with AI-based risk models trained with robust data4. An AI-based tool may be able to systematically identify patterns that clinicians may not currently recognize or have the time to assess in the prediction of preeclampsia. Thus, developing non-invasive AI-supported strategies for detecting early stages of preeclampsia could offer up cost-effective advances in preventative maternal health and long-term cardiovascular health in women.

The potential of retinal vascular imaging

Retinal vascular imaging has recently emerged as a notable non-invasive strategy across a wide range of disease screening efforts (e.g., for neurodegenerative, cardiovascular, and metabolic conditions)5,6,7. Wu et al.’s research builds on research documenting the changes in retinal microvascular throughout pregnancy between women who developed preeclampsia and those who had a normotensive pregnancy8,9. Due to systemic hypertension, patients with preeclampsia developed retinal vessel spasms, tortuous alterations, and at times even retinal detachment9. In their findings, Wu et al. document that these retinal vascular changes occur in early pregnancy and precede the onset of preeclampsia signs/symptoms, suggesting that retinal imaging could serve as a powerful, non-invasive screening tool before the onset of clinical hypertension or organ dysfunction8.

A clinical prediction tool utilizing machine-learning-based algorithms to analyze retinal vascular features could accelerate the assessment of preeclampsia, which would enable more efficient and timely diagnosis of this pregnancy complication. In this vein, Wu et al.’s study introduces PROMPT (Preeclampsia Risk factor + Ophthalmic data + Mean arterial pressure Prediction Test), an AI-based model using retinal photography for preeclampsia prediction in the first 14 weeks of gestation, as this timeframe represents an early feasible timeframe to evaluate women’s risk for preeclampsia10. Wu et al. demonstrate that the PROMPT model could improve the detection of severe adverse pregnancy outcomes from 35% to 41%8. From a women’s public health perspective, retinal screening for preeclampsia in primary care or OBGYN offices during the early first trimester allows for both risk stratification as well as timely prophylactic interventions, such as blood pressure monitoring and control, and aspirin prescription to prevent preeclampsia11,12.

Consequences of early screening for health and health financing

Economically, early-detection using PROMPT was estimated to avert 1809 preeclampsia cases and save over $50 million per 100,000 screenings8. Wu et al. derived this estimate based on data collected from this study, real-world clinical protocols, and a literature search of preeclampsia prevalence8. However, their analysis does not account for potential long-term cost savings related to women’s midlife cardiovascular health and life expectancy. This omission is significant, given the well-established association between hypertensive disorders of pregnancy and increased risk of midlife hypertension12.

In addition to health benefits and immediate economic benefits of early-detection, early screening could reduce healthcare costs associated with midlife cardiovascular morbidity. For example, a 2018 analysis of a nationally representative database found that individuals with hypertension incurred nearly $2000 more in annual healthcare expenses than those without13. Identifying and addressing perinatal risk factors for chronic hypertension such as preeclampsia, could mitigate such long-term expenditures.

While the precise pathophysiology underlying preeclampsia is not yet well-understood, current theories suggest that an imbalance of endothelial growth factors triggers chronic vascular inflammation and dysfunction14,15,16. The risks are further compounded in women with recurrent preeclampsia, where associations with later cardiovascular disease are substantially stronger than women with only one episode of a hypertensive disorder of pregnancy17,18. Additionally, early surveillance and intervention following a hypertensive disorder of pregnancy may serve as an important preventive measure for midlife cardiovascular health12,18. Thus, early detection and timely post-natal intervention may protect women’s health across both peripartum and perimenopausal periods.

Conclusion

Wu et al.’s study provides promising insight into early, non-invasive predictive testing for the diagnosis of preeclampsia. Notably, their AI-based model improves the sensitivity of early preeclampsia detection, which may enable healthcare providers and patients to initiate preventive measures and plan for postpartum management of cardiovascular risk factors. However, larger and more robust studies are necessary to validate and extend these findings. Furthermore, as additional early screening tools emerge, the potential long-term cost savings and health benefits associated with reducing maternal cardiovascular risks warrant further investigation.