Helen Dolk
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Helen Dolk has a total of 484 co-authors. Here are the top co-authors by number of shared publications:
- Maria Loane
Ulster University
27 shared publications - Joan K. Morris
School of Health and Medical Sciences, City St George’s, University of London, London, UK
14 shared publications Ester Garne
Paediatric Department Hospital
13 shared publicationsVera Nelen
Provincial Institute of Hygiene Antwerp Belgium
11 shared publications- Babak Khoshnood
Université Paris Cité and Université Sorbonne Paris Nord, INSERM, INRAE, Centre for Research in Epidemiology and Statistics Paris France
10 shared publications David Tucker
Research, Data & Digital Directorate Public Health Wales Swansea UK
10 shared publications
Latest Publications
- Open Access
- 29 October 2026
- Published by Springer Nature in Drug Safety
- p. 1-13
Disproportionality Analysis Signal Detection Using the New ‘Information Component–Bicluster–Bayesian Hierarchical Model’
- 29 October 2026
- Published by Springer Nature in Drug Safety
- p. 1-13
Open AccessDisproportionality Analysis Signal Detection Using the New ‘Information Component–Bicluster–Bayesian Hierarchical Model’
- Open Access
Timing matters: when do primiparous and multiparous pregnant women initiate folic acid supplementation, and who initiates it too late?
Background: Women are recommended to take folic acid supplements (FAS) periconceptionally to prevent neural tube defects, but the low levels of FAS uptake are poorly understood. This study aimed to determine how parity, pregnancy planning and other maternal factors affect the timing of FAS initiation. Method: We used data from a representative sample of 966 pregnant women from the Northern Ireland Baby Hearts study. Questionnaire responses on health behaviours, including FAS, were linked to maternity and area-based deprivation data. Logistic regression assessed determinants of FAS initiation preconception. Time-varying Cox regression assessed associations between initiation postconception and pregnancy-related milestones. Results: Forty-one per cent of women initiated FAS preconception. Lower odds of preconception initiation were associated with multiparity (adjusted OR 0.47 (0.32 – 0.68)), unplanned pregnancy (adjusted OR 0.08 (0.05 – 0.13)), deprivation and consuming a low folate diet. Higher odds were associated with fertility treatment and previous pregnancy loss. Forty-seven per cent of women initiated FAS postconception, mainly after pregnancy recognition (median 5 weeks), including 34.1% of those who planned their pregnancy. For primiparous women with unplanned pregnancies, the first health service contact (median 7 weeks) further increased FAS initiation. Conclusion: Many women, even multiparous women with previous pregnancy experience and women who planned their pregnancies, are initiating FAS after pregnancy recognition. This shows awareness of FAS benefits, but is too late to prevent neural tube defects. Public health interventions need to go beyond awareness to address motivation and opportunity for FAS, and pay more attention to multiparous women.
Open AccessTiming matters: when do primiparous and multiparous pregnant women initiate folic acid supplementation, and who initiates it too late?
Background: Women are recommended to take folic acid supplements (FAS) periconceptionally to prevent neural tube defects, but the low levels of FAS uptake are poorly understood. This study aimed to determine how parity, pregnancy planning and other maternal factors affect the timing of FAS initiation. Method: We used data from a representative sample of 966 pregnant women from the Northern Ireland Baby Hearts study. Questionnaire responses on health behaviours, including FAS, were linked to maternity and area-based deprivation data. Logistic regression assessed determinants of FAS initiation preconception. Time-varying Cox regression assessed associations between initiation postconception and pregnancy-related milestones. Results: Forty-one per cent of women initiated FAS preconception. Lower odds of preconception initiation were associated with multiparity (adjusted OR 0.47 (0.32 – 0.68)), unplanned pregnancy (adjusted OR 0.08 (0.05 – 0.13)), deprivation and consuming a low folate diet. Higher odds were associated with fertility treatment and previous pregnancy loss. Forty-seven per cent of women initiated FAS postconception, mainly after pregnancy recognition (median 5 weeks), including 34.1% of those who planned their pregnancy. For primiparous women with unplanned pregnancies, the first health service contact (median 7 weeks) further increased FAS initiation. Conclusion: Many women, even multiparous women with previous pregnancy experience and women who planned their pregnancies, are initiating FAS after pregnancy recognition. This shows awareness of FAS benefits, but is too late to prevent neural tube defects. Public health interventions need to go beyond awareness to address motivation and opportunity for FAS, and pay more attention to multiparous women.
- Open Access
- 1 July 2026
- Published by Wiley in British Journal of Clinical Pharmacology
- Vol. 92 (10) , 3586-3597
Case‐malformed signal detection and prioritisation using EUROmediCAT data for pharmacovigilance in pregnancy
Aim: Many women take medications during pregnancy. However, the risk to the fetus from most medications is uncertain. Congenital anomalies are one of the leading causes of infant death and contribute to long‐term disability. Signal detection methods can be used to systematically identify possible medication–anomaly associations that require further investigation. Methods: Data on first trimester medication exposures in pregnancies with a congenital anomaly reported to 14 EUROmediCAT registries with a birth year of 2005–2018 were analysed. Case‐malformed disproportionality analysis identified medication–anomaly signals using a disproportionality signal detection method. Identified signals were then compared to previous EUROmediCAT signal detection studies and ranked based on a proxy for the severity of impact at a population level using the number of estimated excess congenital anomaly cases within the exposed population and the average additional time in hospital over the first year of life. Generalized linear mixed models were used to assess confounding by birth year and registry within the top 20 ranked signals. Results: 1611 medication–outcome pairs with at least three observations were analysed, and 153 signals for 63 different medications were identified. Of the top 20 ranked signals, 10 (involving nine medications) were prioritized for further independent investigation. Conclusion: The signals identified here are hypothesis forming only. Independent studies that adequately account for confounding are subsequently needed to evaluate if the identified signals are causal. Further investigation of signals with low population impact, but high potential individual risk is also recommended.
- 1 July 2026
- Published by Wiley in British Journal of Clinical Pharmacology
- Vol. 92 (10) , 3586-3597
Open AccessCase‐malformed signal detection and prioritisation using EUROmediCAT data for pharmacovigilance in pregnancy
Aim: Many women take medications during pregnancy. However, the risk to the fetus from most medications is uncertain. Congenital anomalies are one of the leading causes of infant death and contribute to long‐term disability. Signal detection methods can be used to systematically identify possible medication–anomaly associations that require further investigation. Methods: Data on first trimester medication exposures in pregnancies with a congenital anomaly reported to 14 EUROmediCAT registries with a birth year of 2005–2018 were analysed. Case‐malformed disproportionality analysis identified medication–anomaly signals using a disproportionality signal detection method. Identified signals were then compared to previous EUROmediCAT signal detection studies and ranked based on a proxy for the severity of impact at a population level using the number of estimated excess congenital anomaly cases within the exposed population and the average additional time in hospital over the first year of life. Generalized linear mixed models were used to assess confounding by birth year and registry within the top 20 ranked signals. Results: 1611 medication–outcome pairs with at least three observations were analysed, and 153 signals for 63 different medications were identified. Of the top 20 ranked signals, 10 (involving nine medications) were prioritized for further independent investigation. Conclusion: The signals identified here are hypothesis forming only. Independent studies that adequately account for confounding are subsequently needed to evaluate if the identified signals are causal. Further investigation of signals with low population impact, but high potential individual risk is also recommended.
- Open Access
Does previous pregnancy experience improve folic acid supplementation uptake? A systematic review and meta-analysis
Objective: To examine whether multiparous women have more or less folic acid uptake than primiparous women. Outcomes: Preconceptional, periconceptional and postconceptional folic acid use among all pregnant women and high risk pregnant women. Design: Systematic review and meta-analysis. Data sources: MEDLINE-Ovid, CINAHL Ultimate, Scopus and ProQuest Medical. Eligibility criteria: Observational epidemiological studies comparing folic acid use between primiparous and multiparous women, published in English from January 1994 to June 2024. Data extraction and synthesis: Two reviewers screened all papers meeting the eligibility criteria. One reviewer completed data extraction and assessed study quality using an adapted version of the Newcastle-Ottawa Scale. Three other reviewers independently assessed 10% of the studies as a quality check. Random-effects (DerSimonian and Laird) meta-analyses combined results for each outcome. Heterogeneity, risk of publication bias and certainty of evidence were assessed. Results: Of the 10 982 records identified, 81 studies involving 826 855 women were included in the review. 27 studies were conducted in Europe, 22 in Asia, 11 in North America, 7 in Africa, 7 in Australia, 5 in the Middle East and 2 in South America. Multiparous women were consistently less likely to take folic acid before and during pregnancy than primiparous women. For preconceptional use, the odds were 29% lower among multiparous women (adjusted OR (aOR): 0.71; 95% CI 0.64 to 0.78; n=25 studies; I 2 =88.67%), and 58% lower in multiparous high-risk women (aOR: 0.42, 95% CI 0.27 to 0.64; n=3 studies; I 2 =27.28%). For periconceptional use, the odds were 32% lower among multiparous women (aOR: 0.68; 95% CI 0.63 to 0.74; n=28 studies; I 2 =85.46%). Postconception, the odds were 21% lower among multiparous women (aOR: 0.79; 95% CI 0.74 to 0.85; n=33 studies; I 2 =85.91%). By the second trimester, there was no significant difference between the two parity groups (aOR: 0.96; 95% CI 0.87 to 1.05; n=4 studies; I 2 =0.00%). The certainty of evidence was low for preconceptional, periconceptional and postconceptional uptake due to heterogeneity, and moderate for preconceptional uptake among high-risk women. Conclusions: Multiparous women were less likely to take folic acid preconceptional, periconceptional and postconceptionally, despite their previous pregnancy experience. Barriers to folic acid supplement uptake among multiparous women need to be identified, and strategies to address them in preconception, antenatal and interconception care should be investigated. PROSPERO registration number: CRD42024553241.
Open AccessDoes previous pregnancy experience improve folic acid supplementation uptake? A systematic review and meta-analysis
Objective: To examine whether multiparous women have more or less folic acid uptake than primiparous women. Outcomes: Preconceptional, periconceptional and postconceptional folic acid use among all pregnant women and high risk pregnant women. Design: Systematic review and meta-analysis. Data sources: MEDLINE-Ovid, CINAHL Ultimate, Scopus and ProQuest Medical. Eligibility criteria: Observational epidemiological studies comparing folic acid use between primiparous and multiparous women, published in English from January 1994 to June 2024. Data extraction and synthesis: Two reviewers screened all papers meeting the eligibility criteria. One reviewer completed data extraction and assessed study quality using an adapted version of the Newcastle-Ottawa Scale. Three other reviewers independently assessed 10% of the studies as a quality check. Random-effects (DerSimonian and Laird) meta-analyses combined results for each outcome. Heterogeneity, risk of publication bias and certainty of evidence were assessed. Results: Of the 10 982 records identified, 81 studies involving 826 855 women were included in the review. 27 studies were conducted in Europe, 22 in Asia, 11 in North America, 7 in Africa, 7 in Australia, 5 in the Middle East and 2 in South America. Multiparous women were consistently less likely to take folic acid before and during pregnancy than primiparous women. For preconceptional use, the odds were 29% lower among multiparous women (adjusted OR (aOR): 0.71; 95% CI 0.64 to 0.78; n=25 studies; I 2 =88.67%), and 58% lower in multiparous high-risk women (aOR: 0.42, 95% CI 0.27 to 0.64; n=3 studies; I 2 =27.28%). For periconceptional use, the odds were 32% lower among multiparous women (aOR: 0.68; 95% CI 0.63 to 0.74; n=28 studies; I 2 =85.46%). Postconception, the odds were 21% lower among multiparous women (aOR: 0.79; 95% CI 0.74 to 0.85; n=33 studies; I 2 =85.91%). By the second trimester, there was no significant difference between the two parity groups (aOR: 0.96; 95% CI 0.87 to 1.05; n=4 studies; I 2 =0.00%). The certainty of evidence was low for preconceptional, periconceptional and postconceptional uptake due to heterogeneity, and moderate for preconceptional uptake among high-risk women. Conclusions: Multiparous women were less likely to take folic acid preconceptional, periconceptional and postconceptionally, despite their previous pregnancy experience. Barriers to folic acid supplement uptake among multiparous women need to be identified, and strategies to address them in preconception, antenatal and interconception care should be investigated. PROSPERO registration number: CRD42024553241.
- Open Access
- 29 April 2026
- Published by Wiley in Pharmacoepidemiology and Drug Safety
- Vol. 35 (5) , e70360
The EUROmediCAT Network and Databases: A Resource for Pharmacovigilance in Pregnancy
Background: The evidence gap relating to the risk of congenital anomalies (CA) associated with first trimester medication exposure in pregnancy is well recognized. Aims: We describe the EUROmediCAT network and databases, and the methodological approach to pregnancy pharmacovigilance. Material and Methods: Multidisciplinary expertise includes CA diagnosis and epidemiology, pharmacoepidemiology, pharmacology and teratology. The EUROmediCAT central database comprises standardized data from 19 EUROCAT CA registries in 14 countries, including more than 40 000 CA cases 1995–2021 with first trimester medication exposure data recorded, and a population coverage of 14.6 million births, growing by more than 650 000 births per year. The distributed database enables federated data analysis across eight countries which can link data from CA registries to electronic healthcare data, with population coverage of up to 900 000 births per year for linkage to maternal prescriptions, of which 300 000 births per year for linkage also to data on all births. Results: The databases have enabled a variety of study designs: case‐malformed control studies, cohort studies, disease cohort studies, signal detection studies, prevalence and ecological studies, and medication utilization studies. Discussion: A key strength is that studies of CA risk can address accurately the specificity of risk by type of CA. Conclusion: EUROmediCAT presents a unique data and expert resource for tackling the enormous evidence gap regarding the safety of medication during pregnancy.
- 29 April 2026
- Published by Wiley in Pharmacoepidemiology and Drug Safety
- Vol. 35 (5) , e70360
Open AccessThe EUROmediCAT Network and Databases: A Resource for Pharmacovigilance in Pregnancy
Background: The evidence gap relating to the risk of congenital anomalies (CA) associated with first trimester medication exposure in pregnancy is well recognized. Aims: We describe the EUROmediCAT network and databases, and the methodological approach to pregnancy pharmacovigilance. Material and Methods: Multidisciplinary expertise includes CA diagnosis and epidemiology, pharmacoepidemiology, pharmacology and teratology. The EUROmediCAT central database comprises standardized data from 19 EUROCAT CA registries in 14 countries, including more than 40 000 CA cases 1995–2021 with first trimester medication exposure data recorded, and a population coverage of 14.6 million births, growing by more than 650 000 births per year. The distributed database enables federated data analysis across eight countries which can link data from CA registries to electronic healthcare data, with population coverage of up to 900 000 births per year for linkage to maternal prescriptions, of which 300 000 births per year for linkage also to data on all births. Results: The databases have enabled a variety of study designs: case‐malformed control studies, cohort studies, disease cohort studies, signal detection studies, prevalence and ecological studies, and medication utilization studies. Discussion: A key strength is that studies of CA risk can address accurately the specificity of risk by type of CA. Conclusion: EUROmediCAT presents a unique data and expert resource for tackling the enormous evidence gap regarding the safety of medication during pregnancy.
- Open Access
- 30 October 2025
- Published by JMIR Publications Inc. in JMIR Research Protocols
- Vol. 14, e75270
Artificial Intelligence–Assisted Image Extraction in Neonatal Echocardiography for Congenital Heart Disease Diagnosis in Sub-Saharan Africa: Protocol for Model Development
Background: Sub-Saharan Africa (SSA) bears the highest global burden of under-5 mortality, with congenital heart disease (CHD) as a major contributor. Despite advancements in high-income countries, CHD-related mortality in SSA remains largely unchanged due to limited diagnostic capacity and centralized health care. While pulse oximetry aids early detection, confirmation typically relies on echocardiography, a procedure constrained by a shortage of specialized personnel. Artificial intelligence (AI) offers a promising solution to bridge this diagnostic gap. Objective: This study aims to develop an AI-assisted echocardiography system that enables nonexpert operators, such as nurses, midwives, and medical doctors, to perform basic cardiac ultrasound sweeps on neonates suspected of CHD and extract accurate cardiac images for remote interpretation by a pediatric cardiologist. Methods: The study will use a 2-phase approach to develop a deep learning model for real-time cardiac view detection in neonatal echocardiography, utilizing data from St. Padre Pio Hospital in Cameroon and the Red Cross War Memorial Children’s Hospital in South Africa to ensure demographic diversity. In phase 1, the model will be pretrained on retrospective data from nearly 500 neonates (0-28 days old). Phase 2 will fine-tune the model using prospective data from 1000 neonates, which include background elements absent in the retrospective dataset, enabling adaptation to local clinical environments. The datasets will consist of short and continuous echocardiographic video clips covering 10 standard cardiac views, as defined by the American Society of Echocardiography. The model architecture will leverage convolutional neural networks and convolutional long short-term memory layers, inspired by the interleaved visual memory framework, which integrates fast and slow feature extractors via a shared temporal memory mechanism. Video preprocessing, annotation with predefined cardiac view codes using Labelbox, and training with TensorFlow and PyTorch will be performed. Reinforcement learning will guide the dynamic use of feature extractors during training. Iterative refinement, informed by clinical input, will ensure that the model effectively distinguishes correct from incorrect views in real time, enhancing its usability in resource-limited settings. Results: Retrospective data collection for the project began in September 2024, and to date, data from 308 babies have been collected and labeled. In parallel, the initial model framework has been developed and training initiated using a subset of the labeled data. The project is currently in the intensive execution phase, with all objectives progressing in parallel and final results expected within 10 months. Conclusions: The AI-assisted echocardiography model developed in this project holds promise for improving early CHD diagnosis and care in SSA and other low-resource settings. International Registered Report Identifier (IRRID): DERR1-10.2196/75270
- 30 October 2025
- Published by JMIR Publications Inc. in JMIR Research Protocols
- Vol. 14, e75270
Open AccessArtificial Intelligence–Assisted Image Extraction in Neonatal Echocardiography for Congenital Heart Disease Diagnosis in Sub-Saharan Africa: Protocol for Model Development
Background: Sub-Saharan Africa (SSA) bears the highest global burden of under-5 mortality, with congenital heart disease (CHD) as a major contributor. Despite advancements in high-income countries, CHD-related mortality in SSA remains largely unchanged due to limited diagnostic capacity and centralized health care. While pulse oximetry aids early detection, confirmation typically relies on echocardiography, a procedure constrained by a shortage of specialized personnel. Artificial intelligence (AI) offers a promising solution to bridge this diagnostic gap. Objective: This study aims to develop an AI-assisted echocardiography system that enables nonexpert operators, such as nurses, midwives, and medical doctors, to perform basic cardiac ultrasound sweeps on neonates suspected of CHD and extract accurate cardiac images for remote interpretation by a pediatric cardiologist. Methods: The study will use a 2-phase approach to develop a deep learning model for real-time cardiac view detection in neonatal echocardiography, utilizing data from St. Padre Pio Hospital in Cameroon and the Red Cross War Memorial Children’s Hospital in South Africa to ensure demographic diversity. In phase 1, the model will be pretrained on retrospective data from nearly 500 neonates (0-28 days old). Phase 2 will fine-tune the model using prospective data from 1000 neonates, which include background elements absent in the retrospective dataset, enabling adaptation to local clinical environments. The datasets will consist of short and continuous echocardiographic video clips covering 10 standard cardiac views, as defined by the American Society of Echocardiography. The model architecture will leverage convolutional neural networks and convolutional long short-term memory layers, inspired by the interleaved visual memory framework, which integrates fast and slow feature extractors via a shared temporal memory mechanism. Video preprocessing, annotation with predefined cardiac view codes using Labelbox, and training with TensorFlow and PyTorch will be performed. Reinforcement learning will guide the dynamic use of feature extractors during training. Iterative refinement, informed by clinical input, will ensure that the model effectively distinguishes correct from incorrect views in real time, enhancing its usability in resource-limited settings. Results: Retrospective data collection for the project began in September 2024, and to date, data from 308 babies have been collected and labeled. In parallel, the initial model framework has been developed and training initiated using a subset of the labeled data. The project is currently in the intensive execution phase, with all objectives progressing in parallel and final results expected within 10 months. Conclusions: The AI-assisted echocardiography model developed in this project holds promise for improving early CHD diagnosis and care in SSA and other low-resource settings. International Registered Report Identifier (IRRID): DERR1-10.2196/75270
- Open Access
- 14 April 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Artificial Intelligence–Assisted Image Extraction in Neonatal Echocardiography for Congenital Heart Disease Diagnosis in Sub-Saharan Africa: Protocol for Model Development (Preprint)
BACKGROUND Sub-Saharan Africa (SSA) bears the highest global burden of under-5 mortality, with congenital heart disease (CHD) as a major contributor. Despite advancements in high-income countries, CHD-related mortality in SSA remains largely unchanged due to limited diagnostic capacity and centralized health care. While pulse oximetry aids early detection, confirmation typically relies on echocardiography, a procedure constrained by a shortage of specialized personnel. Artificial intelligence (AI) offers a promising solution to bridge this diagnostic gap. OBJECTIVE This study aims to develop an AI-assisted echocardiography system that enables nonexpert operators, such as nurses, midwives, and medical doctors, to perform basic cardiac ultrasound sweeps on neonates suspected of CHD and extract accurate cardiac images for remote interpretation by a pediatric cardiologist. METHODS The study will use a 2-phase approach to develop a deep learning model for real-time cardiac view detection in neonatal echocardiography, utilizing data from St. Padre Pio Hospital in Cameroon and the Red Cross War Memorial Children’s Hospital in South Africa to ensure demographic diversity. In phase 1, the model will be pretrained on retrospective data from nearly 500 neonates (0-28 days old). Phase 2 will fine-tune the model using prospective data from 1000 neonates, which include background elements absent in the retrospective dataset, enabling adaptation to local clinical environments. The datasets will consist of short and continuous echocardiographic video clips covering 10 standard cardiac views, as defined by the American Society of Echocardiography. The model architecture will leverage convolutional neural networks and convolutional long short-term memory layers, inspired by the interleaved visual memory framework, which integrates fast and slow feature extractors via a shared temporal memory mechanism. Video preprocessing, annotation with predefined cardiac view codes using Labelbox, and training with TensorFlow and PyTorch will be performed. Reinforcement learning will guide the dynamic use of feature extractors during training. Iterative refinement, informed by clinical input, will ensure that the model effectively distinguishes correct from incorrect views in real time, enhancing its usability in resource-limited settings. RESULTS Retrospective data collection for the project began in September 2024, and to date, data from 308 babies have been collected and labeled. In parallel, the initial model framework has been developed and training initiated using a subset of the labeled data. The project is currently in the intensive execution phase, with all objectives progressing in parallel and final results expected within 10 months. CONCLUSIONS The AI-assisted echocardiography model developed in this project holds promise for improving early CHD diagnosis and care in SSA and other low-resource settings. INTERNATIONAL REGISTERED REPORT DERR1-10.2196/75270
- 14 April 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessArtificial Intelligence–Assisted Image Extraction in Neonatal Echocardiography for Congenital Heart Disease Diagnosis in Sub-Saharan Africa: Protocol for Model Development (Preprint)
BACKGROUND Sub-Saharan Africa (SSA) bears the highest global burden of under-5 mortality, with congenital heart disease (CHD) as a major contributor. Despite advancements in high-income countries, CHD-related mortality in SSA remains largely unchanged due to limited diagnostic capacity and centralized health care. While pulse oximetry aids early detection, confirmation typically relies on echocardiography, a procedure constrained by a shortage of specialized personnel. Artificial intelligence (AI) offers a promising solution to bridge this diagnostic gap. OBJECTIVE This study aims to develop an AI-assisted echocardiography system that enables nonexpert operators, such as nurses, midwives, and medical doctors, to perform basic cardiac ultrasound sweeps on neonates suspected of CHD and extract accurate cardiac images for remote interpretation by a pediatric cardiologist. METHODS The study will use a 2-phase approach to develop a deep learning model for real-time cardiac view detection in neonatal echocardiography, utilizing data from St. Padre Pio Hospital in Cameroon and the Red Cross War Memorial Children’s Hospital in South Africa to ensure demographic diversity. In phase 1, the model will be pretrained on retrospective data from nearly 500 neonates (0-28 days old). Phase 2 will fine-tune the model using prospective data from 1000 neonates, which include background elements absent in the retrospective dataset, enabling adaptation to local clinical environments. The datasets will consist of short and continuous echocardiographic video clips covering 10 standard cardiac views, as defined by the American Society of Echocardiography. The model architecture will leverage convolutional neural networks and convolutional long short-term memory layers, inspired by the interleaved visual memory framework, which integrates fast and slow feature extractors via a shared temporal memory mechanism. Video preprocessing, annotation with predefined cardiac view codes using Labelbox, and training with TensorFlow and PyTorch will be performed. Reinforcement learning will guide the dynamic use of feature extractors during training. Iterative refinement, informed by clinical input, will ensure that the model effectively distinguishes correct from incorrect views in real time, enhancing its usability in resource-limited settings. RESULTS Retrospective data collection for the project began in September 2024, and to date, data from 308 babies have been collected and labeled. In parallel, the initial model framework has been developed and training initiated using a subset of the labeled data. The project is currently in the intensive execution phase, with all objectives progressing in parallel and final results expected within 10 months. CONCLUSIONS The AI-assisted echocardiography model developed in this project holds promise for improving early CHD diagnosis and care in SSA and other low-resource settings. INTERNATIONAL REGISTERED REPORT DERR1-10.2196/75270
- Open Access
- 26 February 2025
- Published by Springer Nature in European Journal of Clinical Pharmacology
- Vol. 81 (5) , 697-709
Antiretroviral drug exposure in pregnancy and risk of congenital anomalies: a European case/non-case malformed study
Purpose: Antiretroviral drugs are recommended during pregnancy to achieve HIV viral suppression and reduce mother-to-child transmission. Congenital anomaly signals were reported after fetal exposure to antiretroviral drugs in several studies warranting further investigation. We aimed to evaluate the risk of congenital anomalies after fetal exposure to antiretroviral drugs using the European congenital anomaly registry data. Methods: A case/non-case study was performed, using the EUROmediCAT central database. All the congenital anomalies, exposed to any antiretroviral drugs, were included from 1995 to 2019. We explored each signal identified from the literature for associations between congenital anomalies and specific antiretroviral exposures. We compared antiretroviral exposure between the signal anomalies (cases) and all other malformed registrations (controls). Reporting odds ratio (ROR) and their 95% confidence intervals were estimated and adjusted for registry and maternal age. Results: Between 1995 and 2019, 173 cases of congenital anomalies were observed after any exposure to antiretroviral drugs. The signal previously identified in the literature between congenital heart defects and exposure to zidovudine was confirmed in the main analysis (aROR 3.66 [1.63–8.23]). Other signals identified in the literature were not confirmed, although two cases of hypospadias and two cases of limb defects were reported after zidovudine and atazanavir exposure, respectively. The signal detection analysis did not reveal any new signal after applying the Bonferroni correction. Conclusions: Our study does not reveal new signals but confirms the previously identified signal between congenital heart defects and fetal exposure to zidovudine. The physio-pathological hypothesis induced by zidovudine exposure should be explored in future studies.
- 26 February 2025
- Published by Springer Nature in European Journal of Clinical Pharmacology
- Vol. 81 (5) , 697-709
Open AccessAntiretroviral drug exposure in pregnancy and risk of congenital anomalies: a European case/non-case malformed study
Purpose: Antiretroviral drugs are recommended during pregnancy to achieve HIV viral suppression and reduce mother-to-child transmission. Congenital anomaly signals were reported after fetal exposure to antiretroviral drugs in several studies warranting further investigation. We aimed to evaluate the risk of congenital anomalies after fetal exposure to antiretroviral drugs using the European congenital anomaly registry data. Methods: A case/non-case study was performed, using the EUROmediCAT central database. All the congenital anomalies, exposed to any antiretroviral drugs, were included from 1995 to 2019. We explored each signal identified from the literature for associations between congenital anomalies and specific antiretroviral exposures. We compared antiretroviral exposure between the signal anomalies (cases) and all other malformed registrations (controls). Reporting odds ratio (ROR) and their 95% confidence intervals were estimated and adjusted for registry and maternal age. Results: Between 1995 and 2019, 173 cases of congenital anomalies were observed after any exposure to antiretroviral drugs. The signal previously identified in the literature between congenital heart defects and exposure to zidovudine was confirmed in the main analysis (aROR 3.66 [1.63–8.23]). Other signals identified in the literature were not confirmed, although two cases of hypospadias and two cases of limb defects were reported after zidovudine and atazanavir exposure, respectively. The signal detection analysis did not reveal any new signal after applying the Bonferroni correction. Conclusions: Our study does not reveal new signals but confirms the previously identified signal between congenital heart defects and fetal exposure to zidovudine. The physio-pathological hypothesis induced by zidovudine exposure should be explored in future studies.
- 31 January 2025
- Published by Wiley in Birth Defects Research
- Vol. 117 (2) , e2435
Risk of Congenital Ocular Anomaly After Prenatal Exposure to Medications: A EUROmediCAT Study
Background: In Europe, the prevalence of congenital ocular anomaly (COA) is estimated at 3.7 per 10,000 births. While certain COAs have a genetic origin, the cause for most patients remains unknown. The role of medications administered during pregnancy in COA genesis in humans is unclear.Objective: To investigate any association between fetal exposure in the first trimester of pregnancy to medications and the occurrence of COA.Methods: We conducted a case‐malformed‐control study using data on 298,351 cases registered as having congenital anomalies (CA) from 19 registries and one healthcare database in 13 European countries. Two analyses were performed: (i) A signal confirmation analysis of signals from the literature, examining associations between COA and specific medications (nitrofurantoin, NSAIDs, opioids, alprazolam, antihypertensives, asthma medications, pyridoxine, and hydroxyethylrutoside). (ii) A signal detection analysis of all medications reported in the database.Results: We identified 4185 COA cases and 232,718 nongenetic and 38,409 genetic controls. We confirmed the association between prenatal opioid exposure and COA (aROR: 2.66, 95% CI: 1.18, 6.02, and 3.22, 95% CI: 1.35, 7.69, for nongenetic and genetic controls, respectively). Signal detection analysis revealed consistent associations for antiglaucoma preparations and miotics (p < 0.01) related to COA. Other associations included congenital cataracts and lens anomalies with desloratadine, congenital glaucoma with antiepileptics, and eyelid malformations with dermatological hydrocortisone.Conclusions: This pharmacoepidemiological study in Europe analyzing COA following fetal medication exposure confirmed reported signals regarding opioids and COA and identified new associations. Validation in independent datasets is necessary to consolidate these findings.
- 31 January 2025
- Published by Wiley in Birth Defects Research
- Vol. 117 (2) , e2435
Risk of Congenital Ocular Anomaly After Prenatal Exposure to Medications: A EUROmediCAT Study
Background: In Europe, the prevalence of congenital ocular anomaly (COA) is estimated at 3.7 per 10,000 births. While certain COAs have a genetic origin, the cause for most patients remains unknown. The role of medications administered during pregnancy in COA genesis in humans is unclear.Objective: To investigate any association between fetal exposure in the first trimester of pregnancy to medications and the occurrence of COA.Methods: We conducted a case‐malformed‐control study using data on 298,351 cases registered as having congenital anomalies (CA) from 19 registries and one healthcare database in 13 European countries. Two analyses were performed: (i) A signal confirmation analysis of signals from the literature, examining associations between COA and specific medications (nitrofurantoin, NSAIDs, opioids, alprazolam, antihypertensives, asthma medications, pyridoxine, and hydroxyethylrutoside). (ii) A signal detection analysis of all medications reported in the database.Results: We identified 4185 COA cases and 232,718 nongenetic and 38,409 genetic controls. We confirmed the association between prenatal opioid exposure and COA (aROR: 2.66, 95% CI: 1.18, 6.02, and 3.22, 95% CI: 1.35, 7.69, for nongenetic and genetic controls, respectively). Signal detection analysis revealed consistent associations for antiglaucoma preparations and miotics (p < 0.01) related to COA. Other associations included congenital cataracts and lens anomalies with desloratadine, congenital glaucoma with antiepileptics, and eyelid malformations with dermatological hydrocortisone.Conclusions: This pharmacoepidemiological study in Europe analyzing COA following fetal medication exposure confirmed reported signals regarding opioids and COA and identified new associations. Validation in independent datasets is necessary to consolidate these findings.