Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Study Protocol

The use of artificial intelligence and machine learning methods in early pregnancy pre-eclampsia screening: A systematic review protocol

  • Paula L. Hedley,

    Roles Conceptualization, Methodology, Writing – original draft

    Affiliations Department for Congenital Disorders, Statens Serum Institut, Copenhagen, Denmark, Brazen Bio, Los Angeles, California, United States of America

    ⨯
  • Christian M. Hagen,

    Roles Investigation, Validation, Writing – review & editing

    Affiliation Department for Congenital Disorders, Statens Serum Institut, Copenhagen, Denmark

    ⨯
  • Casper Wilstrup,

    Roles Conceptualization, Methodology, Supervision, Writing – review & editing

    Affiliations Department for Congenital Disorders, Statens Serum Institut, Copenhagen, Denmark, Abzu, Copenhagen, Denmark

    ⨯
  • Michael Christiansen

    Roles Conceptualization, Investigation, Methodology, Project administration, Writing – original draft

    mic@ssi.dk

    Affiliations Department for Congenital Disorders, Statens Serum Institut, Copenhagen, Denmark, Department of Biomedical Science, University of Copenhagen, Copenhagen, Denmark

    ⨯

Abstract

Introduction

Pre-eclampsia (PE) is a leading cause of perinatal morbidity and mortality worldwide. Low-dose aspirin can prevent PE in high risk pregnancies if started early. However, despite intense research into the area, early pregnancy screening for PE risk is still not a routine part of pregnancy care. Several studies have described the application of artificial intelligence (AI) and machine learning (ML) in risk prediction of PE and its subtypes. A systematic review of available literature is necessary to catalogue the current applications of AI/ML methods in early pregnancy screening for PE, in order to better inform the development of clinically relevant risk prediction algorithms which will enable timely intervention and the development of new treatment strategies. The aim of this systematic review is to identify and assess studies regarding the application of AI/ML methods in early pregnancy screening for PE.

Methods

A systematic review of peer-reviewed as well as the pre-published cohort, case-control, or cross-sectional studies will be conducted. Relevant information will be accessed from the following databases; PubMed, Google Scholar, Scopus, Embase, Web of Science, Cochrane Library, Arxiv, BioRxiv, and MedRxiv. The studies will be evaluated by two reviewers in a parallel, blind assessment of the literature, a third reviewer will assess any studies in which the first two reviewers did not agree. The free online tool Rayyan, will be used in this literature assessment stage. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 checklist will be used to guide the review process and the methods of the studies will be assessed using the Newcastle-Ottawa scale. Narrative synthesis will be conducted for all included studies. Meta-analysis will also be conducted where data quality and availability allow.

Ethics and dissemination

The review will not require ethical approval and the findings will be published in a peer-reviewed journal using the PRISMA guidelines.

Trial registration

Trial registration: The protocol for this systematic review has been registered in PROSPERO [CRD42022345786]. https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42022345786.

Citation: Hedley PL, Hagen CM, Wilstrup C, Christiansen M (2023) The use of artificial intelligence and machine learning methods in early pregnancy pre-eclampsia screening: A systematic review protocol. PLoS ONE 18(4): e0272465. https://doi.org/10.1371/journal.pone.0272465

Editor: Ruxandra Stoean, University of Craiova, ROMANIA

Received: July 17, 2022; Accepted: March 22, 2023; Published: April 20, 2023

Copyright: © 2023 Hedley et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: No datasets will be generated during the current study. All relevant data from this study will be made available upon study completion.

Funding: The authors received no specific funding for this work.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Casper Wilstrup is founder and CEO of Abzu, a company developing AI/ML based research tools. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Abbreviations: AI, artificial intelligence; ML, machine learning; NOS, Newcastle-Ottawa Quality Assessment Scale; PE, pre-eclampsia; PRISMA, preferred reporting items for systematic reviews and meta-analyses; PRISMA-P, PRISMA-protocols; PROSPERO, the international prospective register of systematic reviews

Introduction

Estimated to cause 46–76,000 maternal and 500,000 infant deaths annually [1,2]; pre-eclampsia (PE) is a frequent (3–8%) [3] and important contributor to maternal [4,5] and perinatal [6]—mortality and morbidity. A large proportion of this morbidity and mortality is carried by populations in low and middle income countries (LMICs), where access to pregnancy health care is limited [7]. Thus, using the 2019 data from the Global Burden of Disease (GBD) collaboration [8], it can be calculated that the mortality rate among 15–49 year old women from hypertensive pregnancy disorders is ~ 50 times higher in African low income countries (LICs) than in European high income countries (HICs). In LMICs [7,9], but also in high income countries (HICs) [10], PE associated adverse outcomes are frequently preventable. PE is associated with numerous disorders in mother and child, and particularly cardiovascular morbidity and metabolic syndrome later in life for both mother and child [11,12].

Low dose aspirin treatment reduces the risk of PE, if initiated prior to week 16 [13]. It is therefore important that high-risk patients are identified early, offered preventive treatment, and then carefully monitored throughout pregnancy [6].

Demographic and clinical risk factors, e.g. maternal age, ethnicity, parity, previous pregnancy with PE, maternal metabolic syndrome, along with many first trimester biochemical and ultrasound markers of the development of PE [14–21] have been identified. These factors and markers have been combined in several recommended algorithms [22] with variable performance [23], particularly for late-onset PE. There is thus a need to develop more effective screening methods [2]. An immense body of knowledge has been assembled on markers and potential risk algorithms and it has been suggested that future developments should rely on large-scale prospective studies [24]. However, recent improvements in computing power and cloud storage will make it possible to use clinical phenotype data from patient registers [25] and novel methods of data combination [26] to develop high-performance, robust, clinical screening algorithms for PE [27]. Using big data in this way may also enable the definition of algorithms that can identify pregnancies which will benefit from preventive treatment, as well as support research into the development and implementation of individualised preventive treatments [28].

Artificial intelligence (AI) and machine learning (ML) methods provide the key to unlocking the potential of the many markers and risk factors of PE identified to date. By enabling the combination of clinical phenotype information (extracted from electronic health records) and biomarker information along with environmental exposures, AI/ML methods provide the promise of producing a clinically relevant PE prediction algorithm [25]. Furthermore, AI/ML methods have been applied to the prediction of pregnancy complications, albeit not in a clinical care setting [29].

The objective of this systematic review is to identify and assess studies regarding the application of AI/ML methods in first-trimester screening for PE. The research questions (below) follow a PIO method in which the population of interest (P) are first-trimester pregnancies, the intervention (method of analysis) (I) is any AI/ML method, and the outcome (O) is prediction of PE risk.

Research questions

  • Which AI/ML methods (I) have been used to assess PE risk (O) during early pregnancy (before 16 weeks) (P)?
  • How effective are these algorithms (I) at predicting PE risk (O) during early pregnancy (P)?
  • What risk factors have been associated with PE risk (O) during early pregnancy (P) using AI/ML methods (I)?

Methods

In accordance with the guidelines, our systematic review protocol has been registered in PROSPERO [CRD42022345786].

Study design

A systematic review of both peer-reviewed and pre-published literature with respect to occurrence of PE If possible a meta-analysis, will be performed in accordance with the Preferred Report Items for Systematic Reviews and Meta-analyses (PRISMA) statement [30]. The PRISMA Protocols (PRISMA-P) checklist [31] was used to prepare this protocol (S1 Table).

Eligibility criteria and information sources

We will include original studies (cohort, case–control, cross-sectional, and interventional studies) on PE, performed on samples taken during early pregnancy or performed on data which is otherwise not specific to the stage of pregnancy (e.g. genetic variants). AI/ML methods must have been used to assess the data. Inclusion and exclusion criteria are listed in Table 1. Electronic searches of literature will be carried out using the following databases: PubMed, Scopus, Embase, Web of Science, Google Scholar, Cochrane Library, Arxiv, MedRxiv, and BioRxiv. No date limit will be applied to the searches, with the exception of the pre-print, grey literature sources (Arxiv, MedRxiv, and BioRxiv) which will be limited to articles submitted from the 1st of January 2021 in order to reduce duplication with published sources as well as eliminate publication failures.

thumbnail
Download:
Table 1. Inclusion/exclusion criteria.

https://doi.org/10.1371/journal.pone.0272465.t001

Search strategy

Search terms to be used are listed in Table 2. Population terms related to the early pregnancy will not be used as the time of testing may be unrelated to the stage of pregnancy (e.g. genetic variants, clinical history). As an example; PubMed would be searched using the following search strategy: ("artificial intelligence"[MeSH Terms] OR "machine learning"[MeSH Terms] OR "integrated learning"[Title/Abstract] OR "neural network"[Title/Abstract] OR "sequence learning"[Title/Abstract] OR "deep learning"[Title/Abstract] OR "network analysis"[Title/Abstract] OR "boosting"[Title/Abstract] OR “AdaBoost"[Title/Abstract] OR “XGBoost"[Title/Abstract] OR "symbolic regression"[Title/Abstract] OR "supervised learning"[Title/Abstract] OR "unsupervised learning"[Title/Abstract] OR “Random forest"[Title/Abstract] OR “Decision forest"[Title/Abstract] OR “Decision tree"[Title/Abstract] OR “Support vector machine"[Title/Abstract] OR “Reinforcement learning"[Title/Abstract] OR “Bayesian network"[Title/Abstract] OR “Genetic algorithm"[Title/Abstract] OR “Dimensionality reduction"[Title/Abstract] OR “K-nearest neighbour"[Title/Abstract] OR “K-nearest neighbor"[Title/Abstract]) AND ("preeclampsia"[Title/Abstract] OR "pre-eclampsia"[MeSH Terms]).

thumbnail
Download:
Table 2. List of search terms to be applied covering the AI/ML intervention terms and PE outcome terms.

https://doi.org/10.1371/journal.pone.0272465.t002

Study records

Data management and selection process.

Literature search results will be collected and merged using Endnote (vers 20.3) and, in order to facilitate collaboration among reviewers during the study selection process, transferred to the free online tool Rayyan [32], where duplicates will be removed, and keyword lists based on the inclusion and exclusion criteria will be made in order to facilitate both level 1 and 2 assessment of the literature search results. Prior to the formal screening process, a calibration exercise will be undertaken to pilot and refine the keyword lists. Two reviewers (PLH and MC) will independently assess titles and abstracts against the inclusion and exclusion criteria, as guided by the keyword lists, in order to identify eligible articles. Articles will be rated as ‘included’, ‘excluded’ or ‘maybe’ (i.e. insufficient information in abstract to decide eligibility) and a full-text review will then be performed by both reviewers for all ‘included’ and ‘maybe’ articles. Any discrepancies between the first and second reviewer will be discussed and, in the event, that a consensus cannot be reached a third reviewer (CMH or CW) will make the final decision regarding eligibility. A PRISMA flow diagram will be drawn to demonstrate the stages of the literature selection process and record the reason for excluding studies (Fig 1) [30].

thumbnail
Download:
Fig 1. PRISMA flow diagram.

An illustration of the process of selecting studies for review.

https://doi.org/10.1371/journal.pone.0272465.g001

Critical appraisal of selected studies

The reviewers (PLH and MC) will assess the quality of all eligible studies using the Newcastle-Ottawa Quality Assessment Scale (NOS) for case-control and cohort studies [33] as well as an adapted cross-sectional studies NOS as needed [34]. NOS is a validated quality assessment tool which is divided into three sections focused on selection of participants, comparability of study groups, and ascertainment of outcomes or exposures. Each section contains a number of criteria for which a study can be awarded a star, the maximum number of stars that can be awarded is nine. Any discrepancy in NOS scoring will be settled by a third reviewer (CMH or CW). If a discrepancy persists then the average score will be used. Studies with 7–9 stars will be considered high-quality and their data used in a meta-analysis.

Data collection

One reviewer (PLH) will extract data from the included studies using an excel spreadsheet. Data from multiple reports pertaining to the same study will be linked so as not to report duplicate results. The excel spreadsheet will collect data on the first author, publication year, study period, country of study, type of study, peer-reviewed status, publication status, sample size, maternal age, gestational age at sampling, blood pressure measurements, maternal body mass index, birth weight, risk markers, AI/ML method, all reported outcomes pertaining to PE, and NOS scores. The excel sheet will be adapted following a pilot test on 10 articles, selected to represent various AI/ML methods and risk factors, in order to ensure all relevant data is extracted. The second reviewer (MC) will approve the excel sheet and check the accuracy of the data.

Data synthesis

Tables will be created to summarise the characteristics of the included studies. A narrative analysis will be presented with reference to the particular AI/ML method and type of data used in the study. We intend to compare the diagnostic efficiency of the different methods, if sufficient studies will be found. If possible, considering the wide range of possible AI/ML methods employed, a meta-analysis will be performed using R package meta. In the event that a meta-analysis cannot be performed, data for similar AI/ML methods will be compiled and reported together. Strength of evidence will be graded by two reviewers (PLH and MC) using the Evidence-based Practice Centre 2015 guidelines [35].

Patient and public involvement

Patients or the public will not be involved in this systematic review.

Ethics and dissemination

This systematic review will not need ethical approval because it will retrieve and synthesise data from publicly available published and pre-published studies. Study results will be disseminated through scientific publications presented at relevant local and international scientific conferences.

Discussion

Studies have been performed which use AI/ML methods to predict complications in pregnancy [29]. However, as most studies focus on second trimester pregnancies an overview of first trimester pregnancy PE prediction using AI/ML methods is needed as identification of PE risk in the first trimester can be used to select pregnancies that may benefit from preventive treatment with Low dose aspirin, which has been documented to reduce the occurrence of PE [13,36].

Many AI/ML methods are currently in use within research and development [26]; this review would allow us to comment on the ability of these methods to support patient participation in decisions with reference to the “explainability” or “interpretability” of the result. These two concepts refer to the ability to explain how the algorithm parameters relates to the result (explainability of black-box models) and how understandable and trustworthy the algorithm results are (interpretability of white-box models) [37]. As risk assessment for PE supports high-risk decisions it is a very important that patients and doctors can understand which underlying factors determine the outcome.

This systematic review will establish the current state of knowledge concerning the prediction of PE during the first trimester through use of AI/ML methods and will provide the highest level of evidence to inform future research as well as the development of early pregnancy, PE screening algorithms. However, the considerable heterogeneity expected among the included literature (i.e. different AI/ML methods used, different sample type or gestational ages at time of sampling) may pose a challenge for consistent and comprehensive data extraction. Consequently, meta-analysis will be limited by the quality and quantity of data available. Any significant amendments to this systematic review protocol will require the approval of all authors and will be documented, with the date of the amendment and the rationale for the amendment, on the PROSPERO record, updated protocol document, and in the final manuscript. The review findings will be published in a peer-reviewed open-access scientific journal at the conclusion of this study.

Supporting information

S1 Table. PRISMA-P 2015 checklist.

https://doi.org/10.1371/journal.pone.0272465.s001

(PDF)

References

  1. 1. Mol BWJ, Roberts CT, Thangaratinam S, Magee LA, de Groot CJM, Hofmeyr GJ. Pre-eclampsia. Lancet. 2016;387(10022):999–1011. Epub 2015/09/08. pmid:26342729.
  2. 2. Chaemsaithong P, Sahota DS, Poon LC. First trimester preeclampsia screening and prediction. Am J Obstet Gynecol. 2022;226(2S):S1071–S97.e2. Epub 20200716. pmid:32682859.
  3. 3. Abalos E, Cuesta C, Grosso AL, Chou D, Say L. Global and regional estimates of preeclampsia and eclampsia: a systematic review. Eur J Obstet Gynecol Reprod Biol. 2013;170(1):1–7. Epub 2013/06/12. pmid:23746796.
  4. 4. Say L, Chou D, Gemmill A, Tuncalp O, Moller AB, Daniels J, et al. Global causes of maternal death: a WHO systematic analysis. Lancet Glob Health. 2014;2(6):e323–33. Epub 2014/08/12. pmid:25103301.
  5. 5. Zhang J, Meikle S, Trumble A. Severe maternal morbidity associated with hypertensive disorders in pregnancy in the United States. Hypertens Pregnancy. 2003;22(2):203–12. Epub 2003/08/12. pmid:12909005.
  6. 6. von Dadelszen P, Payne B, Li J, Ansermino JM, Broughton Pipkin F, Cote AM, et al. Prediction of adverse maternal outcomes in pre-eclampsia: development and validation of the fullPIERS model. Lancet. 2011;377(9761):219–27. Epub 2010/12/28. pmid:21185591.
  7. 7. Moodley J, Soma-Pillay P, Buchmann E, Pattinson RC. Hypertensive disorders in pregnancy: 2019 National guideline. S Afr Med J. 2019;109(9):12723. Epub 20190913. pmid:31635598.
  8. 8. Diseases GBD, Injuries C. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396(10258):1204–22. pmid:33069326.
  9. 9. Hodgins S, Tielsch J, Rankin K, Robinson A, Kearns A, Caglia J. A New Look at Care in Pregnancy: Simple, Effective Interventions for Neglected Populations. PLoS One. 2016;11(8):e0160562. Epub 20160818. pmid:27537281.
  10. 10. Katsuragi S, Tanaka H, Hasegawa J, Kanayama N, Nakata M, Murakoshi T, et al. Analysis of preventability of malignancy-related maternal death from the nationwide registration system of maternal deaths in Japan. J Matern Fetal Neonatal Med. 2021;34(3):432–8. Epub 20190502. pmid:30999803.
  11. 11. Chappell LC, Cluver CA, Kingdom J, Tong S. Pre-eclampsia. Lancet. 2021;398(10297):341–54. Epub 2021/05/31. pmid:34051884.
  12. 12. Garovic VD, White WM, Vaughan L, Saiki M, Parashuram S, Garcia-Valencia O, et al. Incidence and Long-Term Outcomes of Hypertensive Disorders of Pregnancy. J Am Coll Cardiol. 2020;75(18):2323–34. pmid:32381164.
  13. 13. Rolnik DL, Wright D, Poon LC, O’Gorman N, Syngelaki A, de Paco Matallana C, et al. Aspirin versus Placebo in Pregnancies at High Risk for Preterm Preeclampsia. N Engl J Med. 2017;377(7):613–22. Epub 20170628. pmid:28657417.
  14. 14. Akolekar R, Zaragoza E, Poon LC, Pepes S, Nicolaides KH. Maternal serum placental growth factor at 11 + 0 to 13 + 6 weeks of gestation in the prediction of pre-eclampsia. Ultrasound Obstet Gynecol. 2008;32(6):732–9. Epub 2008/10/29. pmid:18956425.
  15. 15. Christiansen M, Hedley PL, Placing S, Wojdemann KR, Carlsen AL, Jorgensen JM, et al. Maternal Serum Resistin Is Reduced in First Trimester Preeclampsia Pregnancies and Is a Marker of Clinical Severity. Hypertension in Pregnancy. 2015;34(4):422–33. pmid:26636480
  16. 16. Cnossen JS, Morris RK, ter Riet G, Mol BW, van der Post JA, Coomarasamy A, et al. Use of uterine artery Doppler ultrasonography to predict pre-eclampsia and intrauterine growth restriction: a systematic review and bivariable meta-analysis. Can Med Assoc J. 2008;178(6):701–11. Epub 2008/03/12. pmid:18332385.
  17. 17. Grill S, Rusterholz C, Zanetti-Dallenbach R, Tercanli S, Holzgreve W, Hahn S, et al. Potential markers of preeclampsia—a review. Reprod Biol Endocrin. 2009;7:70. Epub 2009/07/16. pmid:19602262.
  18. 18. Hedley PL, Placing S, Wojdemann K, Carlsen AL, Shalmi AC, Sundberg K, et al. Free leptin index and PAPP-A: a first trimester maternal serum screening test for pre-eclampsia. Prenatal Diag. 2010;30(2):103–9. Epub 2009/12/17. pmid:20013873.
  19. 19. Laigaard J, Sorensen T, Placing S, Holck P, Frohlich C, Wojdemann KR, et al. Reduction of the disintegrin and metalloprotease ADAM12 in preeclampsia. Obstet Gynecol. 2005;106(1):144–9. Epub 2005/07/05. pmid:15994630.
  20. 20. Spencer K, Cowans NJ, Nicolaides KH. Low levels of maternal serum PAPP-A in the first trimester and the risk of pre-eclampsia. Prenatal Diag. 2008;28(1):7–10. Epub 2007/11/15. pmid:18000943.
  21. 21. De Villiers CP, Hedley PL, Placing S, Wojdemann KR, Shalmi AC, Carlsen AL, et al. Placental protein-13 (PP13) in combination with PAPP-A and free leptin index (fLI) in first trimester maternal serum screening for severe and early preeclampsia. Clin Chem Lab Med. 2017;56(1):65–74. Epub 2017/07/14. pmid:28704180.
  22. 22. O’Gorman N, Wright D, Poon LC, Rolnik DL, Syngelaki A, de Alvarado M, et al. Multicenter screening for pre-eclampsia by maternal factors and biomarkers at 11–13 weeks’ gestation: comparison with NICE guidelines and ACOG recommendations. Ultrasound Obstet Gynecol. 2017;49(6):756–60. pmid:28295782.
  23. 23. Tan MY, Syngelaki A, Poon LC, Rolnik DL, O’Gorman N, Delgado JL, et al. Screening for pre-eclampsia by maternal factors and biomarkers at 11–13 weeks’ gestation. Ultrasound Obstet Gynecol. 2018;52(2):186–95. Epub 2018/06/14. pmid:29896812.
  24. 24. Townsend R, Khalil A, Premakumar Y, Allotey J, Snell KIE, Chan C, et al. Prediction of pre-eclampsia: review of reviews. Ultrasound Obstet Gynecol. 2019;54(1):16–27. pmid:30267475.
  25. 25. Myatt L. The prediction of preeclampsia: the way forward. Am J Obstet Gynecol. 2022;226(2S):S1102–S7.e8. Epub 20210327. pmid:33785181.
  26. 26. Christiansen M, Wilstrup C, Hedley PL. Explainable "White-box" machine learning is the way forward in pre-eclampsia screening. Am J Obstet Gynecol. 2022. Epub 20220629. pmid:35779588.
  27. 27. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019;25(1):44–56. Epub 20190107. pmid:30617339.
  28. 28. Bica I, Alaa AM, Lambert C, van der Schaar M. From Real-World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges. Clin Pharmacol Ther. 2021;109(1):87–100. Epub 20200628. pmid:32449163.
  29. 29. Bertini A, Salas R, Chabert S, Sobrevia L, Pardo F. Using Machine Learning to Predict Complications in Pregnancy: A Systematic Review. Front Bioeng Biotechnol. 2021;9:780389. Epub 20220119. pmid:35127665.
  30. 30. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. Epub 2021/03/31. pmid:33782057.
  31. 31. Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev. 2015;4:1. Epub 20150101. pmid:25554246.
  32. 32. Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. Rayyan-a web and mobile app for systematic reviews. Syst Rev. 2016;5(1):210. Epub 2016/12/07. pmid:27919275.
  33. 33. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses Ottawa, Canada: Ottawa Hospital Research Institute; 2019 [cited 2020 20/01/2020]. http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp.
    • 34. Modesti PA, Reboldi G, Cappuccio FP, Agyemang C, Remuzzi G, Rapi S, et al. Panethnic Differences in Blood Pressure in Europe: A Systematic Review and Meta-Analysis. PLoS One. 2016;11(1):e0147601. Epub 20160125. pmid:26808317.
    • 35. Berkman ND, Lohr KN, Ansari MT, Balk EM, Kane R, McDonagh M, et al. Grading the strength of a body of evidence when assessing health care interventions: an EPC update. J Clin Epidemiol. 2015;68(11):1312–24. Epub 20141220. pmid:25721570.
    • 36. ACOG Committee Opinion No. 743: Low-Dose Aspirin Use During Pregnancy. Obstet Gynecol. 2018;132(1):e44–e52. pmid:29939940.
    • 37. Barragan-Montero A, Bibal A, Dastarac MH, Draguet C, Valdes G, Nguyen D, et al. Towards a safe and efficient clinical implementation of machine learning in radiation oncology by exploring model interpretability, explainability and data-model dependency. Phys Med Biol. 2022;67(11). Epub 20220527. pmid:35421855.