Against empirical metformin use in the obese gravida at early pregnancy: A critical appraisal
Dimitrios G. Goulis,
Professor in Reproductive Endocrinology, Unit of Reproductive Endocrinology, First Department of Obstetrics and Gynecology, Medical School, Aristotle University of Thessaloniki, Greece
Obesity in pregnancy confers substantial maternal and fetal risks, including gestational diabetes mellitus (GDM), hypertensive disorders, macrosomia, and adverse neonatal outcomes. The pharmacological potential of metformin, an insulin sensitizer with a well-established safety record in type 2 diabetes mellitus and polycystic ovary syndrome (PCOS), has prompted increasing interest in its prophylactic use in obese pregnant women. However, current evidence does not support its routine antenatal prescription in this population, and several mechanistic, clinical, and ethical concerns warrant critical scrutiny.The landmark EMPOWaR and MOP trials, both double-blind, placebo-controlled randomized studies, demonstrated no statistically significant reduction in primary outcomes, including large-for-gestational-age (LGA) neonates, GDM incidence, gestational weight gain, or composite adverse perinatal outcomes, in obese women receiving metformin from the first trimester. Despite the well-documented transplacental passage of metformin, the implications of fetal exposure remain incompletely characterized across gestation. Notably, metformin-exposed offspring in some cohorts have demonstrated paradoxical increases in childhood adiposity at follow-up assessments, raising concerns about long-term metabolic programming, a finding with considerable epigenetic plausibility, given metformin’s AMPK-mediated interference with mitochondrial function during critical developmental windows.From a pharmacokinetic standpoint, renal clearance of metformin increases substantially during pregnancy due to augmented glomerular filtration, rendering standard dosing regimens potentially subtherapeutic; in contrast, dose escalation introduces gastrointestinal intolerability and a theoretical risk of lactic acidosis in patients with pregnancy-associated physiological hemodynamic shifts. Furthermore, conflating insulin resistance in obesity with established GDM pathophysiology oversimplifies the metabolic heterogeneity of the obese gravid state, in which inflammatory dysregulation, adipokine imbalance, and placental dysfunction contribute independently to adverse outcomes beyond hyperglycemia alone.
Ethical considerations also merit emphasis: Obese pregnant women without confirmed glucose dysregulation or polycystic ovary syndrome represent a population for whom informed consent regarding off-label use, unresolved fetal safety data, and the absence of regulatory approval for this indication must be transparently communicated.
In conclusion, although metformin’s therapeutic role in confirmed GDM and insulin-resistant states remains validated, its empirical prescription in normoglycemic obese pregnant women at early gestation is premature, insufficiently evidence-based, and potentially carries risk to the developing conceptus.
SHOULD PREDICTION MODELS AND NEW BIOMARKERS REPLACE TRADITIONAL RISK FACTORS IN PRETERM BIRTH SCREENING?
Dilly OC Anumba, MBBS FWACS FRCOG MD LL.M (Medical Law)
Chair of Obstetrics and Gynaecology, University of Sheffield
Honorary Consultant Obstetrician/Gynaecologist, Subspecialist in Maternal and Fetal Medicine, Sheffield Teaching Hospitals
NIHR Senior Investigator
Honorary Professor, University of Cape Town, South Africa
Chief Investigator, ECCLIPPx Research Programme
Division of Clinical Medicine
School of Medicine & Population Health, Faculty of Health
The University of Sheffield, 4th Floor, Jessop Wing, Tree Root Walk, Sheffield S10 2SF, UK
Preterm birth remains a leading cause of neonatal morbidity and mortality worldwide, yet our approaches to risk stratification in routine antenatal care continue to rely heavily on traditional clinical risk factors such as prior preterm birth, cervical surgery, smoking, multiple gestation, and short cervical length. In parallel, rapid advances in data science, multi-omics technologies, and biomarker discovery have generated a new generation of prediction models and candidate biomarkers that promise earlier, more precise identification of women at risk. This lecture critically examines whether these emerging tools should replace — or instead complement — established risk-based screening strategies. We will first review the strengths and limitations of traditional risk factors, including their accessibility, low cost, and integration into existing care pathways, but also their modest predictive performance at the individual level. We will then explore contemporary prediction models that integrate clinical, imaging, and biochemical data, as well as novel biomarkers derived from proteomics, metabolomics, inflammatory pathways, extracellular vesicles, and cervicovaginal or circulating analytes. Key issues to be addressed include model discrimination and calibration, reproducibility across populations, clinical utility versus statistical significance, cost-effectiveness, health equity, and implementation challenges in diverse care settings. Finally, we will consider whether the central question is not replacement but integration: how to combine traditional risk factors, cervical assessment, and validated biomarkers within dynamic, individualized prediction frameworks. By evaluating the evidence through a lens of clinical impact, feasibility, and ethical responsibility, this session aims to provide a balanced perspective on the future of preterm birth screening and to outline priorities for research and translation into practice.
Debate: Tubectomy at cesarean section for ovarian cancer risk reduction? CON
Nikolaos Thomakos, Greece
This presentation critically addresses whether bilateral salpingectomy at the time of cesarean section should be adopted as a routine strategy for ovarian cancer risk reduction. Although the tubal origin hypothesis of many high-grade serous carcinomas provides a strong biological rationale for opportunistic salpingectomy, cesarean delivery represents a distinct obstetric surgical setting in which maternal safety, hemostasis, operative exposure, and informed consent must remain the primary priorities.
The presentation does not argue against salpingectomy itself, but against its unselected, routine use during all cesarean deliveries. Current evidence supports feasibility and short-term safety in selected cases, but important uncertainties remain regarding long-term oncologic benefit, ovarian function, menopause-related outcomes, and real-world implementation. Therefore, a selective, patient-centered approach is proposed, based on antenatal counseling, completed childbearing, documented informed consent, favorable intraoperative conditions, and a clear safety-based decision to abandon the procedure when necessary.
The key message is that salpingectomy at cesarean section may be appropriate for selected women, but it should not become an automatic add-on to cesarean delivery. Prevention should be offered with clinical judgment, ethical clarity, and appropriate surgical safeguards.
Epigenetics and reprogramming
David Monk, Biomedical Research Centre, School of Biological Sciences, University of East Anglia, Norwich Research Park, Norwich, UK
Studies have demonstrated an increased risk for adverse perinatal outcomes, including fetal growth restriction, preeclampsia, and preterm birth following ART. These reports strongly suggest variability in epigenetic regulation, which is further supported by the increased incidence of rare imprinting disorders following assisted conception. It remains to be determined if altered epigenetic profiles are gamete-derived or acquired post-fertilization; either way, they could influence embryo, fetal and placenta growth. However, there is little in the way of follow-up studies to determine the long-term adverse outcomes associated with these modest epigenetic changes. In my presentation I will discuss the current standing of research addressing the epigenetic state of human pre-implantation embryos, including the development of low input and single-cell multiomic techniques to quantify DNA methylation reprograming and embryonic genome activation. Using genomic imprinting as a model system of maintained germline-derived epigenetic state, I will describe how methylation in the placenta is associated with intrauterine growth. Finally, I will explain how research in the fields of imprinting disorders and infertility converge to clarify the function of the subcortical maternal complex and the oocyte cytoplasmic lattice. These structures are encoded by key maternal-effect genes, which lead to a spectrum of phenotypes when mutated, ranging from fertilization failure through to multi-locus imprinting disturbances.
Tubectomy at cesarean section for ovarian cancer risk reduction: The case for opportunistic salpingectomy
Rinat Gabbay-Benziv, MD, Department of Obstetrics and Gynecology, Hillel Yaffe Medical Center, Hadera, Israel;
Faculty of Medicine, Technion – Israel Institute of Technology, Haifa, Israel
For women undergoing cesarean delivery who have completed childbearing and have chosen permanent contraception, opportunistic bilateral salpingectomy offers a unique opportunity to combine definitive contraception with potential ovarian cancer risk reduction.
The rationale is biologically compelling. Increasing evidence supports the distal fallopian tube as the site of origin for a substantial proportion of high-grade serous ovarian carcinomas. Removal of the entire fallopian tube therefore provides a theoretical advantage over traditional tubal interruption, which leaves most tubal epithelium in situ. Population-based studies suggest that salpingectomy is associated with a greater reduction in future ovarian cancer risk than tubal ligation.
Importantly, this potential benefit does not appear to come at the cost of substantial perioperative morbidity. Systematic reviews and meta-analyses of salpingectomy at cesarean delivery demonstrate only a modest increase in operative time, without consistent increases in blood loss, transfusion, infection, readmission, reoperation, or organ injury. Large real-world population studies have similarly provided reassuring safety data.
Theoretical concerns regarding compromise of ovarian blood supply have also not translated into convincing evidence of reduced ovarian reserve. In addition, economic models suggest that opportunistic salpingectomy can be a cost-effective strategy when performed at cesarean delivery in women already seeking permanent contraception.
The argument is therefore not for routine salpingectomy in every cesarean delivery. Rather, in an appropriately counseled woman who desires permanent contraception, bilateral salpingectomy should be offered as the preferred option when technically feasible. It avoids a future procedure, provides highly effective permanent contraception, and adds a potentially meaningful opportunity for ovarian cancer prevention.
GLP-1 and gestational diabetes
Rinat Gabbay-Benziv, MD, Department of Obstetrics and Gynecology, Hillel Yaffe Medical Center, Hadera, Israel;
Faculty of Medicine, Technion – Israel Institute of Technology, Haifa, Israel
Glucagon-like peptide-1 (GLP-1) has become central to the treatment of type 2 diabetes and obesity, increasing the likelihood that women of reproductive age will encounter GLP-1 receptor agonists before conception or during early pregnancy. This lecture examines the relevance of GLP-1 biology to gestational diabetes mellitus (GDM) and asks whether current evidence supports therapeutic use of GLP-1 receptor agonists during pregnancy.
Pregnancy is characterized by progressive insulin resistance and a compensatory increase in pancreatic β-cell function. GDM develops when β-cell compensation is insufficient for the prevailing metabolic demand. GLP-1 may contribute to this physiology through glucose-dependent stimulation of insulin secretion, glucagon modulation, effects on gastric emptying, and appetite regulation. However, human studies do not support a simple model of GLP-1 deficiency in GDM. Reported findings include reduced GLP-1 secretion relative to glycemia, increased post-load GLP-1 concentrations interpreted as compensation or incretin resistance, and associations between GLP-1 and insulin that do not consistently extend to glucose concentrations.
The biological rationale for GLP-1 receptor agonism is therefore compelling but remains distinct from clinical evidence. Direct trials of intentional GLP-1 receptor agonist treatment for GDM are lacking. Available pregnancy data largely concern inadvertent or periconceptional exposure and are limited by small numbers, observational designs, heterogeneous exposure, and confounding by diabetes, obesity, and glycemic control. A recent systematic review found no consistent signal for major congenital malformations but concluded that the evidence is insufficient for firm safety conclusions. Animal, placental, and cellular studies provide reasons for continued caution rather than proof of clinical harm.
The lecture concludes that GLP-1 is a promising pathway for understanding GDM and a priority for future research, but GLP-1 receptor agonists should not currently be initiated to treat GDM. Preconception planning, prompt review of inadvertent exposure, and transition to established pregnancy-compatible therapies remain the appropriate clinical approach.
The central message is promising pathway, unproven pregnancy therapy.
Oral agents for gestational diabetes: should they be first-line pharmacologic therapy? – CON
Rinat Gabbay-Benziv, MD, Department of Obstetrics and Gynecology, Hillel Yaffe Medical Center, Hadera, Israel;
Faculty of Medicine, Technion – Israel Institute of Technology, Haifa, Israel
When lifestyle intervention fails in gestational diabetes mellitus (GDM), the therapeutic goal is straightforward: normalize maternal glycemia while minimizing treatment-related maternal and fetal risk. Insulin remains uniquely suited to this purpose. It can be precisely titrated, has no practical therapeutic ceiling, does not cross the placenta, and has decades of pregnancy safety data. Accordingly, the 2026 American Diabetes Association Standards of Care continue to recommend insulin as the preferred pharmacologic treatment for GDM and specifically state that metformin and glyburide should not be used as first-line agents.
Metformin is clearly preferable to glyburide and may be appropriate when insulin is declined or impractical. However, “reasonable alternative” is not equivalent to “preferred first-line therapy.” Metformin freely crosses the placenta, with fetal concentrations approaching or exceeding maternal levels. Treatment failure requiring insulin occurs in approximately 14 – 46% of women. Moreover, a 2026 network meta-analysis of 71 randomized trials including 14,877 participants found biguanides associated with increased low birth weight (OR 2.04, 95% CI 1.04–4.01). The authors noted that fetal metformin exposure may affect mitochondrial oxidative phosphorylation, AMPK signaling, and nutrient-allocation pathways. All are mechanisms with plausible implications for fetal growth and metabolic programming.
Long-term data remains incomplete. In MiG-TOFU, children exposed to metformin had similar total and abdominal fat percentages and metabolic parameters at 7–9 years but were larger in weight and arm and waist circumferences, with uncertain long-term significance.
Recent randomized evidence further argues against an oral-first strategy. In the 2025 SUGAR-DIP trial, sequential metformin followed by glyburide failed to demonstrate noninferiority to insulin for large-for-gestational-age birth (23.9% vs 19.9%), and maternal hypoglycemia was more frequent (20.9% vs 10.9%). Glyburide is even less defensible: the INDAO trial failed to establish noninferiority for perinatal complications and concluded that its use as first-line therapy was not justified.
Oral agents have an important role, particularly when insulin is unacceptable or impractical. But first-line therapy should remain insulin: treat maternal hyperglycemia effectively without unnecessarily exposing the fetus to a pharmacologically active metabolic agent whose long-term effects remain incompletely defined.
Can we accurately predict macrosomia? The use of AI
Giuseppe Rizzo, Università Sapienza Rome Italy
Fetal macrosomia complicates 5–10% of pregnancies and carries a well-recognised burden: prolonged labour, shoulder dystocia, birth trauma, caesarean delivery, postpartum haemorrhage, neonatal hypoglycaemia and long-term metabolic risk for the offspring. Accurate antenatal prediction should allow timely counselling and individualised planning of birth. In practice it remains elusive: estimated fetal weight (EFW) derived from conventional biometry is limited by operator dependence, formula error and the physiological variability of late-pregnancy growth, and models based on fetal biometry and maternal characteristics alone perform poorly.
Two lines of evidence show where the residual predictive signal lies. In a prospective cohort of 2156 singleton pregnancies scanned at 36 weeks, macrosomia complicated 9.8% of pregnancies, and EFW alone yielded an AUC of 0.750, rising to 0.801 once maternal characteristics were added; incorporating umbilical vein blood flow normalised for abdominal circumference (UVBF/AC) increased the AUC to 0.892, with a parallel gain for birthweight >4000 g (0.724 → 0.770 → 0.851). The same haemodynamic marker is informative far earlier: at 11+0 to 13+6 weeks a model integrating parity, BMI and PAPP-A MoM had an AUC of 0.72 for large-for-gestational-age, improving to 0.79 with the addition of the UVBF Z-score and reaching 0.83 for birthweight >4000 g. Growth acceleration is therefore preceded by measurable changes in nutrient and oxygen delivery, and prediction improves when a physiological variable, rather than another anthropometric one, is added
This is precisely the setting in which artificial intelligence can contribute. Machine-learning algorithms integrating maternal characteristics, metabolic parameters and fetal ultrasound measurements have outperformed standard Hadlock-based approaches, with reported specificity of 82% and AUC up to 0.95; ensemble models combining random forest, XGBoost and LightGBM likewise outperform both single classifiers and the traditional formula, while deep learning applied directly to ultrasound images promises automated, operator-independent biometry and Doppler acquisition.
Enthusiasm must be tempered. Most published models are retrospective and single-centre, developed on imbalanced datasets, rarely externally validated, seldom calibrated and almost never tested for their impact on clinical decisions or perinatal outcome; reporting frequently falls short of TRIPOD+AI standards, and performance degrades when models are transported to populations with different macrosomia prevalence and scanning practice.
So, can we accurately predict macrosomia? Better than a decade ago, and best when AI is used not as a substitute for physiology but as a framework to combine maternal, biochemical, biometric and haemodynamic information. The next step is prospective, multicentre validation with maternal and neonatal morbidity, rather than birthweight alone, as the endpoint.
