This article is a commentary on:
AI-based preeclampsia detection and prediction with electrocardiogram data
Lionel Carbillon1,2*
- 1Department of Obstetrics and Gynecology, Hôpital Jean Verdier, Assistance Publique—Hôpitaux de Paris, Bondy, France
- 2Sorbonne North Paris University, Bobigny, France
A Commentary on
AI-based preeclampsia detection and prediction with electrocardiogram data
By Butler L, Gunturkun F, Chinthala L, Karabayir I, Tootooni MS, Bakir-Batu B, Celik T, Akbilgic O, Davis RL. (2024). Front. Cardiovasc. Med. 11:1360238. doi: 10.3389/fcvm.2024.1360238
1 Introduction
Preeclampsia is the final systemic manifestation of poor adaptation of the maternal body to pregnancy (1, 2).
Early gestational adaptation of maternal organism is a consequence and reflection of the first few weeks of pregnancy and quality of placentation. In particular, early adjustments in vascular tone, volume expansion, cardiac output, and blood pressure are crucial, and reflect adaptations of the utero-placental circulation and cardiovascular system caused by the hemochorial nature of placentation in our species (3–5).
In all clinical forms of preeclampsia, the exceeding of the capacities of the circulatory system occurs “at a given moment” in the progression of the underlying pathophysiological process leading towards cardiac output decrease (6, 7), vasoconstrictive tone, hypertension and failure of the maternal organism to the demands of pregnancy, finally leading to preeclamptic syndrome (8, 9). Butler et al. (10) have published an interesting study evaluating a proposed artificial intelligence (AI)-based algorithm for preeclampsia prediction. However, the “moment” (the precise gestational age), at which the clinical emergence of the disease (and even more, severe features of preeclampsia and placenta-related complications) will occur, is extremely difficult to predict with precision, especially without taking into account utero-placental circulation indices. Thus, I believe that the message that “preeclampsia can be identified with high accuracy via application of AI models to electrocardiographic (ECG) data” (10), as claimed by the authors in the conclusion of the abstract, should not be professed. The appropriate message should instead be that “ECG data can help identify pregnant women at high risk of preeclampsia”, as rightly pointed out by the authors at the beginning of the discussion section.
Indeed, defective placentation is the well-recognized primum movens of the underlying pathological process (early “placental” stages of the underlying, pre-clinical process) (1), which can lead to the clinical emergence of preeclampsia when the adaptability of the maternal circulatory system is exceeded (1, 6–8). This also reflects exceeding the immunologic tolerance of the conceptus, after the underlying process of the disease enters its last stages towards a systemic inflammatory response (1, 11, 12), with impaired synthesis of vasorelaxing agents and endogenous anticoagulants, increased production of vasoconstrictors, and the ultimate endothelial dysfunction (13) and its major systemic manifestations. These last stages are associated with inadequate “placental derived factors” in the mother's blood, preceding and promoting the appearance of the complete clinical syndrome. Moreover, the syndrome may occur rapidly or progressively, and also depends on the severity of placental dysfunction, which is not considered by the algorithm.
2 Discussion
Thus, the “10-s 12-lead ECG” data (10), which finely assess gestational adaptation of the heart, provides a very useful tool for assessing the possible maladaptation of the maternal cardiovascular system to pregnancy at a given time. However, potential future users of this device should bear in mind that the potential for progression of the underlying disease at a given moment depends not only on the adaptation of the cardiovascular system, but also on placental function (the very origin of the pathology), fetal growth, and placental derived (pro and anti angiogenic) factors in the mother's blood (14, 15).
Notably, this AI-based model, like others (16), has only a modest predictive performance in women with preeclampsia giving birth at less than 37 weeks of gestation [95% area under the curve confidence interval of 0.76 (0.58–0.95)]. This performance should make potential users very cautious in using its supposed negative predictive value, especially since at these gestational ages (especially before 32–34 weeks of gestation), timely preventive treatment for prematurity complications with betamethasone and magnesium sulfate treatments may be necessary and are time-consuming. Therefore, any delay in diagnosis of preeclampsia can lead to a loss of chance for the newborn, if the diagnosis of preeclampsia is made too late to implement preventive treatments of prematurity complications without endangering the mother.
Moreover, fetal growth-restricted pregnancies are characterized by a lower cardiac output and higher total vascular resistance index than that expected for gestation (17, 18). In these circumstances, in accordance with the “Guytonean model”, such “hyperdynamic state makes it possible for extremely slight long-term changes in blood fluid volume and cardiac output to raise or lower arterial pressure” (19–21), which can have serious consequences for the mother and fetus.
As Easterling recently recalled (20), “when cardiac output is under the mean for gestational age and/or vascular resistance is elevated, the association with the development of fetal growth restriction is strong”. Without considering fetal growth restriction (FGR) and possible placental dysfunction, which needs a true targeted expertise of the utero-placental circulation and function (14, 15), the danger of the test lies in its overestimation of the negative predictive value. Such overestimation will inherently impact the relevance of the proposed schedule for follow-up of the patient in the last weeks or months of pregnancy, due to the risk of rapid progression of placental dysfunction and the underlying disease if stages 5–6 have been reached (1) with the possibility of rapid onset of serious complications.
Finally, the combination of various pre-eclampsia risk factors known to “exponentially increase” the risk (22) must also be considered, including the history of early or intermediate preeclampsia, preterm preeclampsia, obesity alone or in combination with chronic hypertension and/or ongoing pregnancy with FGR.
Author contributions
LC: Conceptualization, Formal Analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Writing – original draft, Writing – review & editing.
Funding
The author declares that no financial support was received for the research, authorship, and/or publication of this article.
Conflict of interest
The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1. Redman CWG. The six stages of pre-eclampsia. Pregnancy Hypertens. (2014) 4:246. doi: 10.1016/j.preghy.2014.04.020
2. Reijnders IF, Mulders AGM, Koster MPH, Kropman A, Koning A, Willemsen SP, et al. First-trimester maternal haemodynamic adaptation to pregnancy and placental, embryonic and fetal development: the prospective observational Rotterdam periconception cohort. BJOG. (2022) 129:785–95. doi: 10.1111/1471-0528.16979
3. Meah VL, Cockcroft JR, Backx K, Shave R, Stöhr EJ. Cardiac output and related haemodynamics during pregnancy: a series of meta-analyses. Heart. (2016) 102:518–26. doi: 10.1136/heartjnl-2015-308476
4. Slade LJ, Syngelak AI, Wilson M, Mistry HD, Akoleka R, von Dadelszen P, et al. Blood pressure cut-offs at 11–13 weeks’ gestation and risk of preeclampsia. Am J Obstet Gynecol. (2024):S0002-9378(24)00558-1. doi: 10.1016/j.ajog.2024.04.032
5. Carbillon L, Uzan M, Uzan S. Pregnancy, vascular tone, and maternal hemodynamics: a crucial adaptation. Obstet Gynecol Surv. (2000) 55:574–81. doi: 10.1097/00006254-200009000-00023
6. San-Frutos LM, Fernández R, Almagro J, Barbancho C, Salazar F, Pérez-Medina T, et al. Measure of hemodynamic patterns by thoracic electrical bioimpedance in normal pregnancy and in preeclampsia. Eur J Obstet Gynecol Reprod Biol. (2005) 121:149–53. doi: 10.1016/j.ejogrb.2004.12.018
7. Tyldum EV, Backe B, Støylen A, Slørdahl SA. Maternal left ventricular and endothelial functions in preeclampsia. Acta Obstet Gynecol Scand. (2012) 91:566–73. doi: 10.1111/j.1600-0412.2011.01282.x
8. Bernstein IM, Meyer MC, Osol G, Ward K. Intolerance to volume expansion: a theorized mechanism for the development of preeclampsia. Obstet Gynecol. (1998) 92:306–8. doi: 10.1016/s0029-7844(98)00207-
9. Yagel S, Cohen SM, Admati I, Skarbianskis N, Solt I, Zeisel A, et al. Expert review: preeclampsia type I and type II. Am J Obstet Gynecol MFM. (2023) 5:101203. doi: 10.1016/j.ajogmf.2023.101203
10. Butler L, Gunturkun F, Chinthala L, Karabayir I, Tootooni MS, Bakir-Batu B, et al. AI-based preeclampsia detection and prediction with electrocardiogram data. Front Cardiovasc Med. (2024) 11:1360238. doi: 10.3389/fcvm.2024.1360238
11. Redman CWG, Sargent IL. Pre-eclampsia, the placenta and the maternal systemic inflammatory response–a review. Placenta. (2003) 24:S21–7. doi: 10.1053/plac.2002.0930
12. Stewart FM, Freeman DJ, Ramsay JE, Greer IA, Caslake M, Ferrell WR. Longitudinal assessment of maternal endothelial function and markers of inflammation and placental function throughout pregnancy in lean and obese mothers. J Clin Endocrinol Metab. (2007) 92:969–75. doi: 10.1210/jc.2006-2083
13. Roberts JM, Taylor RN, Musci TJ, Rodgers GM, Hubel CA, McLaughlin MK. Preeclampsia: an endothelial cell disorder. Am J Obstet Gynecol. (1989) 161:1200–4. doi: 10.1016/0002-9378(89)90665-0
14. Agrawal S, Parks WT, Zeng HD, Ravichandran A, Ashwal E, Windrim RC, et al. Diagnostic utility of serial circulating placental growth factor levels and uterine artery Doppler waveforms in diagnosing underlying placental diseases in pregnancies at high risk of placental dysfunction. Am J Obstet Gynecol. (2022) 227:618.e1–16. doi: 10.1016/j.ajog.2022.05.043
15. Stepan H, Hund M, Andraczek T. Combining biomarkers to predict pregnancy complications and redefine preeclampsia: the angiogenic-placental syndrome. Hypertension. (2020) 75:918–26. doi: 10.1161/HYPERTENSIONAHA.119.13763
16. Snell KIE, Allotey J, Smuk M, Hooper R, Chan C, Ahmed A, et al. External validation of prognostic models predicting pre-eclampsia: individual participant data meta-analysis. BMC Med. (2020) 18:302. doi: 10.1186/s12916-020-01766-9
17. Melchiorre K, Sutherland GR, Liberati M, Thilaganathan B. Maternal cardiovascular impairment in pregnancies complicated by severe fetal growth restriction. Hypertension. (2012) 60:437–43. doi: 10.1161/HYPERTENSIONAHA.112.194159
18. Easterling TR, Benedetti TJ, Carlson KC, Brateng DA, Wilson J, Schmucker BS. The effect of maternal hemodynamics on fetal growth in hypertensive pregnancies. Am J Obstet Gynecol. (1991) 165:902–6. doi: 10.1016/0002-9378(91)90436-u
19. Easterling TR, Benedetti TJ, Schmucker BC, Millard SP. Maternal hemodynamics in normal and preeclamptic pregnancies: a longitudinal study. Obstet Gynecol. (1990) 76:1061–9.2234714
20. Easterling TR. Individualized management of maternal hemodynamics to prevent preeclampsia: improvement in maternal outcomes without adverse fetal effects. Hypertension. (2021) 77:2054–6. doi: 10.1161/HYPERTENSIONAHA.121.17202
21. Guyton AC. Dominant role of the kidneys and accessory role of whole-body autoregulation in the pathogenesis of hypertension. Am J Hypertension. (1989) 2:575–85. doi: 10.1093/ajh/2.7.575
22. Villa PM, Marttinen P, Gillberg J, Lokki AI, Majander A, Ordén M-R, et al. Cluster analysis to estimate the risk of preeclampsia in the high-risk prediction and prevention of preeclampsia and intrauterine growth restriction (PREDO) study. PLoS One. (2017) 12:e0174399. doi: 10.1371/journal.pone.0174399
Keywords: preeclampsia, gestational cardiovascular adaptation, blood pressure, cardiac output, artificial-intelligence, electrocardiogram, fetal growth restriction
Citation: Carbillon L (2024) Commentary: AI-based preeclampsia detection and prediction with electrocardiogram data. Front. Cardiovasc. Med. 11: 1437369. doi: 10.3389/fcvm.2024.1437369
Received: 24 May 2024; Accepted: 25 June 2024;
Published: 30 July 2024.
Edited by:
Pietro Scicchitano, ASLBari—Azienda Sanitaria Localedella provincia di Bari (ASL BA), ItalyReviewed by:
Xiaoyuan Han, University of the Pacific, United States© 2024 Carbillon. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Lionel Carbillon, lionel.carbillon@aphp.fr