Comparative Safety of In Utero Exposure to Buprenorphine Combined With Naloxone vs Buprenorphine Alone.

Importance: Buprenorphine combined with naloxone is commonly used to treat opioid use disorders outside of pregnancy. In pregnancy, buprenorphine alone is generally recommended because of limited perinatal safety data on the combination product. Objective: To compare perinatal outcomes following prenatal exposure to buprenorphine with naloxone vs buprenorphine alone. Design, Settings, and Participants: Population-based cohort study using health care utilization data from Medicaid-insured beneficiaries in the US from 2000 to 2018. The cohort was restricted to pregnant individuals linked to their liveborn infants, with maternal Medicaid enrollment from 3 months before pregnancy to 1 month after delivery and infant enrollment for the first 3 months after birth, unless they died sooner. Exposure: Use of buprenorphine with naloxone vs buprenorphine alone during the first trimester based on outpatient dispensings. Main Outcomes and Measures: Outcomes included major congenital malformations, low birth weight, neonatal abstinence syndrome, neonatal intensive care unit admission, preterm birth, respiratory symptoms, small for gestational age, cesarean delivery, and maternal morbidity. Confounder-adjusted risk ratios were calculated using propensity score overlap weights. Results: This study identified 3369 pregnant individuals exposed to buprenorphine with naloxone during the first trimester (mean [SD] age, 28.8 [4.6] years) and 5326 exposed to buprenorphine alone or who switched from the combination to buprenorphine alone by the end of the first trimester (mean [SD] age, 28.3 [4.5] years). When comparing buprenorphine combined with naloxone with buprenorphine alone, a lower risk for neonatal abstinence syndrome (absolute risk, 37.4% vs 55.8%; weighted relative risk, 0.77 [95% CI, 0.70-0.84]) and a modestly lower risk for neonatal intensive care unit admission (absolute risk, 30.6% vs 34.9%; weighted relative risk, 0.91 [95% CI, 0.85-0.98]) and small for gestational age (absolute risk, 10.0% vs 12.4%; weighted relative risk, 0.86 [95% CI, 0.75-0.98]) was observed. For maternal morbidity, the comparative rates were 2.6% vs 2.9%, respectively, and the weighted relative risk was 0.90 (95% CI, 0.68-1.19). No differences were observed with respect to major congenital malformations overall, low birth weight, preterm birth, respiratory symptoms, or cesarean delivery. Results were consistent across sensitivity analyses. Conclusions and Relevance: There were similar and, in some instances, more favorable neonatal and maternal outcomes for pregnancies exposed to buprenorphine combined with naloxone compared with buprenorphine alone. For the outcomes assessed, compared with buprenorphine alone, buprenorphine with naloxone during pregnancy appears to be a safe treatment option. This supports the view that both formulations are reasonable options for the treatment of opioid use disorder in pregnancy, affirming flexibility in collaborative treatment decision-making.

Hierarchical Clustering Analysis to Inform Classification of Congenital Malformations for Surveillance of Medication Safety in Pregnancy.

There is growing interest in the secondary use of healthcare data to evaluate medication safety in pregnancy. Tree-based scan statistics (TBSS) offer an innovative approach to help identify potential safety signals. TBSS utilize hierarchically organized outcomes, generally based on existing clinical coding systems that group outcomes by organ system. When assessing teratogenicity, such groupings often lack a sound embryologic basis given the etiologic heterogeneity of congenital malformations. The study objective was to enhance the grouping of congenital malformations to be used in scanning approaches through implementation of hierarchical clustering analysis (HCA) and to pilot test an HCA-enhanced TBSS approach for medication safety surveillance in pregnancy in two test cases using >4.2 million mother-child dyads from two US-nationwide databases. HCA identified (1) malformation combinations belonging to the same organ system already grouped in existing classifications, (2) known combinations across different organ systems not previously grouped, (3) unknown combinations not previously grouped, and (4) malformations seemingly standing on their own. Testing the approach with valproate and topiramate identified expected signals, and a signal for an HCA-cluster missed by traditional classification. Augmenting existing classifications with clusters identified through large data exploration may be promising when defining phenotypes for surveillance and causal inference studies.

Deep survival analysis for interpretable time-varying prediction of preeclampsia risk.

OBJECTIVE: Survival analysis is widely utilized in healthcare to predict the timing of disease onset. Traditional methods of survival analysis are usually based on Cox Proportional Hazards model and assume proportional risk for all subjects. However, this assumption is rarely true for most diseases, as the underlying factors have complex, non-linear, and time-varying relationships. This concern is especially relevant for pregnancy, where the risk for pregnancy-related complications, such as preeclampsia, varies across gestation. Recently, deep learning survival models have shown promise in addressing the limitations of classical models, as the novel models allow for non-proportional risk handling, capturing nonlinear relationships, and navigating complex temporal dynamics. METHODS: We present a methodology to model the temporal risk of preeclampsia during pregnancy and investigate the associated clinical risk factors. We utilized a retrospective dataset including 66,425 pregnant individuals who delivered in two tertiary care centers from 2015 to 2023. We modeled the preeclampsia risk by modifying DeepHit, a deep survival model, which leverages neural network architecture to capture time-varying relationships between covariates in pregnancy. We applied time series k-means clustering to DeepHit's normalized output and investigated interpretability using Shapley values. RESULTS: We demonstrate that DeepHit can effectively handle high-dimensional data and evolving risk hazards over time with performance similar to the Cox Proportional Hazards model, achieving an area under the curve (AUC) of 0.78 for both models. The deep survival model outperformed traditional methodology by identifying time-varied risk trajectories for preeclampsia, providing insights for early and individualized intervention. K-means clustering resulted in patients delineating into low-risk, early-onset, and late-onset preeclampsia groups-notably, each of those has distinct risk factors. CONCLUSION: This work demonstrates a novel application of deep survival analysis in time-varying prediction of preeclampsia risk. Our results highlight the advantage of deep survival models compared to Cox Proportional Hazards models in providing personalized risk trajectory and demonstrating the potential of deep survival models to generate interpretable and meaningful clinical applications in medicine.

Eighteen-Month Outcomes Among Pregnant and Nonpregnant Reproductive-Aged People Hospitalized for Coronavirus Disease 2019.

Background: Physiologic and immunologic adaptations in pregnancy may increase the risk of adverse outcomes from respiratory viral infections. However, data are limited on longer-term outcomes after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in pregnancy prior to widespread vaccine availability. Methods: Using electronic health record data, we retrospectively compared 6-, 12-, and 18-month outcomes including death and rehospitalization between pregnant and nonpregnant reproductive-aged individuals hospitalized for SARS-CoV-2 infection between 2020 and 2021 at 2 academic referral hospitals. Results: There were 190 nonpregnant and 70 pregnant participants. Mean age was 31 years for pregnant and 34 years for nonpregnant participants. For pregnant patients, mean gestational age at coronavirus disease 2019 (COVID-19) diagnosis was 36 weeks, 54% delivered by cesarean, and 97% delivered a live birth. Compared to pregnant participants, nonpregnant participants had a higher prevalence of baseline comorbidities and a higher proportion received mechanical ventilation (84% vs 55%). Index hospitalization complications (31% vs 17%) and mortality (3% vs 0%) were more common in nonpregnant participants. Over 18 months following index hospitalization, 39 (21%) nonpregnant and 5 (7%) pregnant participants were readmitted, most for infection (28/44 [64%]). Most readmissions occurred within 6 months. There were no posthospitalization deaths in the pregnant group. Conclusions: Pregnant people with severe COVID-19 disease had a low rate of severe adverse outcomes after index hospitalization. The low readmission rate is reassuring that pregnant individuals may not be at higher risk for long-term severe adverse health outcomes after COVID-19 compared to the nonpregnant reproductive-aged population, possibly because any increased risk conferred by pregnancy resolves soon after delivery.

Placental senescence pathophysiology is shared between peripartum cardiomyopathy and preeclampsia in mouse and human.

Peripartum cardiomyopathy (PPCM) is an idiopathic form of pregnancy-induced heart failure associated with preeclampsia. Circulating factors in late pregnancy are thought to contribute to both diseases, suggesting a common underlying pathophysiological process. However, what drives this process remains unclear. Using serum proteomics, we identified the senescence-associated secretory phenotype (SASP), a marker of cellular senescence associated with biological aging, as the most highly up-regulated pathway in young women with PPCM or preeclampsia. Placentas from women with preeclampsia displayed multiple markers of amplified senescence and tissue aging, as well as overall increased gene expression of 28 circulating proteins that contributed to SASP pathway enrichment in serum samples from patients with preeclampsia or PPCM. The most highly expressed placental SASP factor, activin A, was associated with cardiac dysfunction or heart failure severity in women with preeclampsia or PPCM. In a murine model of PPCM induced by cardiomyocyte-specific deletion of the gene encoding peroxisome proliferator-activated receptor γ coactivator-1α, inhibiting activin A signaling in the early postpartum period with a monoclonal antibody to the activin type II receptor improved heart function. In addition, attenuating placental senescence with the senolytic compound fisetin in late pregnancy improved cardiac function in these animals. These findings link senescence biology to cardiac dysfunction in pregnancy and help to elucidate the pathogenesis underlying cardiovascular diseases of pregnancy.

Aspiring toward equitable benefits from genomic advances to individuals of ancestrally diverse backgrounds.

Advancements in genomic technologies have shown remarkable promise for improving health trajectories. The Human Genome Project has catalyzed the integration of genomic tools into clinical practice, such as disease risk assessment, prenatal testing and reproductive genomics, cancer diagnostics and prognostication, and therapeutic decision making. Despite the promise of genomic technologies, their full potential remains untapped without including individuals of diverse ancestries and integrating social determinants of health (SDOHs). The NHGRI launched the 2020 Strategic Vision with ten bold predictions by 2030, including "individuals from ancestrally diverse backgrounds will benefit equitably from advances in human genomics." Meeting this goal requires a holistic approach that brings together genomic advancements with careful consideration to healthcare access as well as SDOHs to ensure that translation of genetics research is inclusive, affordable, and accessible and ultimately narrows rather than widens health disparities. With this prediction in mind, this review delves into the two paramount applications of genetic testing-reproductive genomics and precision oncology. When discussing these applications of genomic advancements, we evaluate current accessibility limitations, highlight challenges in achieving representativeness, and propose paths forward to realize the ultimate goal of their equitable applications.

Placental transfer dynamics and durability of maternal COVID-19 vaccine-induced antibodies in infants.

Completion of a COVID-19 vaccination series during pregnancy effectively reduces COVID-19 hospitalization among infants less than 6 months of age. The dynamics of transplacental transfer of maternal vaccine-induced antibodies, and their persistence in infants at 2, 6, 9, and 12 months, have implications for new vaccine development and optimal timing of vaccine administration in pregnancy. We evaluated anti-COVID antibody IgG subclass, Fc-receptor binding profile, and activity against wild-type Spike and RBD plus five variants of concern (VOCs) in 153 serum samples from 100 infants. Maternal IgG1 and IgG3 responses persisted in 2- and 6-month infants to a greater extent than the other IgG subclasses, with high persistence of antibodies binding placental neonatal Fc-receptor and FcγR3A. Lowest persistence was observed against the Omicron RBD-specific region. Maternal vaccine timing, placental Fc-receptor binding capabilities, antibody subclass, fetal sex, and VOC all impact the persistence of antibodies in infants through 12 months of age.

Risk of Autism after Prenatal Topiramate, Valproate, or Lamotrigine Exposure.

BACKGROUND: Maternal use of valproate during pregnancy has been associated with an increased risk of neurodevelopmental disorders in children. Although most studies of other antiseizure medications have not shown increased risks of these disorders, there are limited and conflicting data regarding the risk of autism spectrum disorder associated with maternal topiramate use. METHODS: We identified a population-based cohort of pregnant women and their children within two health care utilization databases in the United States, with data from 2000 through 2020. Exposure to specific antiseizure medications was defined on the basis of prescription fills from gestational week 19 until delivery. Children who had been exposed to topiramate during the second half of pregnancy were compared with those unexposed to any antiseizure medication during pregnancy with respect to the risk of autism spectrum disorder. Valproate was used as a positive control, and lamotrigine was used as a negative control. RESULTS: The estimated cumulative incidence of autism spectrum disorder at 8 years of age was 1.9% for the full population of children who had not been exposed to antiseizure medication (4,199,796 children). With restriction to children born to mothers with epilepsy, the incidence was 4.2% with no exposure to antiseizure medication (8815 children), 6.2% with exposure to topiramate (1030 children), 10.5% with exposure to valproate (800 children), and 4.1% with exposure to lamotrigine (4205 children). Propensity score-adjusted hazard ratios in a comparison with no exposure to antiseizure medication were 0.96 (95% confidence interval [CI], 0.56 to 1.65) for exposure to topiramate, 2.67 (95% CI, 1.69 to 4.20) for exposure to valproate, and 1.00 (95% CI, 0.69 to 1.46) for exposure to lamotrigine. CONCLUSIONS: The incidence of autism spectrum disorder was higher among children prenatally exposed to the studied antiseizure medications than in the general population. However, after adjustment for indication and other confounders, the association was substantially attenuated for topiramate and lamotrigine, whereas an increased risk remained for valproate. (Funded by the National Institute of Mental Health.).

SARS-CoV-2 infection elucidates unique features of pregnancy-specific immunity.

Pregnancy is a risk factor for increased severity of SARS-CoV-2 and other respiratory infections. The mechanisms underlying this risk have not been well-established, partly due to a limited understanding of how pregnancy shapes immune responses. To gain insight into the role of pregnancy in modulating immune responses at steady state and upon perturbation, we collected peripheral blood mononuclear cells (PBMC), plasma, and stool from 226 women, including 152 pregnant individuals (n = 96 with SARS-CoV-2 infection and n = 56 healthy controls) and 74 non-pregnant women (n = 55 with SARS-CoV-2 and n = 19 healthy controls). We found that SARS-CoV-2 infection was associated with altered T cell responses in pregnant compared to non-pregnant women. Differences included a lower percentage of memory T cells, a distinct clonal expansion of CD4-expressing CD8 + T cells, and the enhanced expression of T cell exhaustion markers, such as programmed cell death-1 (PD-1) and T cell immunoglobulin and mucin domain-3 (Tim-3), in pregnant women. We identified additional evidence of immune dysfunction in severely and critically ill pregnant women, including a lack of expected elevation in regulatory T cell (Treg) levels, diminished interferon responses, and profound suppression of monocyte function. Consistent with earlier data, we found maternal obesity was also associated with altered immune responses to SARS-CoV-2 infection, including enhanced production of inflammatory cytokines by T cells. Certain gut bacterial species were altered in pregnancy and upon SARS-CoV-2 infection in pregnant individuals compared to non-pregnant women. Shifts in cytokine and chemokine levels were also identified in the sera of pregnant individuals, most notably a robust increase of interleukin-27 (IL-27), a cytokine known to drive T cell exhaustion, in the pregnant uninfected control group compared to all non-pregnant groups. IL-27 levels were also significantly higher in uninfected pregnant controls compared to pregnant SARS-CoV-2-infected individuals. Using two different preclinical mouse models of inflammation-induced fetal demise and respiratory influenza viral infection, we found that enhanced IL-27 protects developing fetuses from maternal inflammation but renders adult female mice vulnerable to viral infection. These combined findings from human and murine studies reveal nuanced pregnancy-associated immune responses, suggesting mechanisms underlying the increased susceptibility of pregnant individuals to viral respiratory infections.

Genetic Associations of Circulating Cardiovascular Proteins With Gestational Hypertension and Preeclampsia.

Importance: Hypertensive disorders of pregnancy (HDPs), including gestational hypertension and preeclampsia, are important contributors to maternal morbidity and mortality worldwide. In addition, women with HDPs face an elevated long-term risk of cardiovascular disease. Objective: To identify proteins in the circulation associated with HDPs. Design, Setting, and Participants: Two-sample mendelian randomization (MR) tested the associations of genetic instruments for cardiovascular disease-related proteins with gestational hypertension and preeclampsia. In downstream analyses, a systematic review of observational data was conducted to evaluate the identified proteins' dynamics across gestation in hypertensive vs normotensive pregnancies, and phenome-wide MR analyses were performed to identify potential non-HDP-related effects associated with the prioritized proteins. Genetic association data for cardiovascular disease-related proteins were obtained from the Systematic and Combined Analysis of Olink Proteins (SCALLOP) consortium. Genetic association data for the HDPs were obtained from recent European-ancestry genome-wide association study meta-analyses for gestational hypertension and preeclampsia. Study data were analyzed October 2022 to October 2023. Exposures: Genetic instruments for 90 candidate proteins implicated in cardiovascular diseases, constructed using cis-protein quantitative trait loci (cis-pQTLs). Main Outcomes and Measures: Gestational hypertension and preeclampsia. Results: Genetic association data for cardiovascular disease-related proteins were obtained from 21 758 participants from the SCALLOP consortium. Genetic association data for the HDPs were obtained from 393 238 female individuals (8636 cases and 384 602 controls) for gestational hypertension and 606 903 female individuals (16 032 cases and 590 871 controls) for preeclampsia. Seventy-five of 90 proteins (83.3%) had at least 1 valid cis-pQTL. Of those, 10 proteins (13.3%) were significantly associated with HDPs. Four were robust to sensitivity analyses for gestational hypertension (cluster of differentiation 40, eosinophil cationic protein [ECP], galectin 3, N-terminal pro-brain natriuretic peptide [NT-proBNP]), and 2 were robust for preeclampsia (cystatin B, heat shock protein 27 [HSP27]). Consistent with the MR findings, observational data revealed that lower NT-proBNP (0.76- to 0.88-fold difference vs no HDPs) and higher HSP27 (2.40-fold difference vs no HDPs) levels during the first trimester of pregnancy were associated with increased risk of HDPs, as were higher levels of ECP (1.60-fold difference vs no HDPs). Phenome-wide MR analyses identified 37 unique non-HDP-related protein-disease associations, suggesting potential on-target effects associated with interventions lowering HDP risk through the identified proteins. Conclusions and Relevance: Study findings suggest genetic associations of 4 cardiovascular disease-related proteins with gestational hypertension and 2 associated with preeclampsia. Future studies are required to test the efficacy of targeting the corresponding pathways to reduce HDP risk.

Prescription Stimulant Use During Pregnancy and Risk of Neurodevelopmental Disorders in Children.

Importance: Use of medications for attention-deficit/hyperactivity disorder (ADHD) during pregnancy is increasing in the US. Whether exposure to these medications in utero impacts the risk of neurodevelopmental disorders in children is uncertain. Objective: To evaluate the association of childhood neurodevelopmental disorders with in utero exposure to stimulant medications for ADHD. Design, Setting, and Participants: This cohort study included health care utilization data from publicly insured (Medicaid data from 2000 to 2018) and commercially insured (MarketScan Commercial Claims Database data from 2003 to 2020) pregnant individuals aged 12 to 55 years in the US with enrollment from 3 months prior to pregnancy through 1 month after delivery, linked to children. Children were monitored from birth until outcome diagnosis, disenrollment, death, or end of the study (December 2018 for Medicaid and December 2020 for MarketScan). Exposures: Dispensing of amphetamine/dextroamphetamine or methylphenidate in the second half of pregnancy. Main Outcomes and Measures: Autism spectrum disorder, ADHD, and a composite of any neurodevelopmental disorder were defined using validated algorithms. Hazard ratios were estimated comparing amphetamine/dextroamphetamine and methylphenidate to no exposure. Results: The publicly insured cohort included 2 496 771 stimulant-unexposed, 4693 amphetamine/dextroamphetamine-exposed, and 786 methylphenidate-exposed pregnancies with a mean (SD) age of 25.2 (6.0) years. The commercially insured cohort included 1 773 501 stimulant-unexposed, 2372 amphetamine/dextroamphetamine-exposed, and 337 methylphenidate-exposed pregnancies with a mean (SD) age of 31.6 (4.6) years. In unadjusted analyses, amphetamine/dextroamphetamine and methylphenidate exposure were associated with a 2- to 3-fold increased risk of the neurodevelopmental outcomes considered. After adjustment for measured confounders, amphetamine/dextroamphetamine exposure was not associated with any outcome (autism spectrum disorder: hazard ratio [HR], 0.80; 95% CI, 0.56-1.14]; ADHD: HR, 1.07; 95% CI, 0.89-1.28; any neurodevelopmental disorder: HR, 0.91; 95% CI, 0.81-1.28). Methylphenidate exposure was associated with an increased risk of ADHD (HR, 1.43; 95% CI, 1.12-1.82]) but not other outcomes after adjustment (autism spectrum disorder: HR, 1.06; 95% CI, 0.62-1.81; any neurodevelopmental disorder: HR, 1.15; 95% CI, 0.97-1.36). The association between methylphenidate and ADHD did not persist in sensitivity analyses with stricter control for confounding by maternal ADHD. Conclusions and Relevance: The findings in this study suggest that amphetamine/dextroamphetamine and methylphenidate exposure in utero are not likely to meaningfully increase the risk of childhood neurodevelopmental disorders.

First Trimester Use of Buprenorphine or Methadone and the Risk of Congenital Malformations.

Importance: Use of buprenorphine or methadone to treat opioid use disorder is recommended in pregnancy; however, their teratogenic potential is largely unknown. Objective: To compare the risk of congenital malformations following in utero exposure to buprenorphine vs methadone. Design, Setting, and Participants: This population-based cohort study used health care utilization data from publicly insured Medicaid beneficiaries in the US from 2000 to 2018. A total of 13 360 pregnancies with enrollment from 90 days prior to pregnancy start through 1 month after delivery and first trimester use of buprenorphine or methadone were included and linked to infants. Data were analyzed from July to December 2022. Exposure: A pharmacy dispensing of buprenorphine or a code for administration of methadone in the first trimester. Main Outcomes and Measures: Primary outcomes included major malformations overall and malformations previously associated with opioids (any cardiac malformations, ventricular septal defect, secundum atrial septal defect/nonprematurity-related patent foramen ovale, neural tube defects, clubfoot, and oral clefts). Secondary outcomes included other organ system-specific malformations. Risk differences and risk ratios (RRs) were estimated comparing buprenorphine with methadone, adjusting for confounders with propensity score overlap weights. Results: The cohort included 9514 pregnancies with first-trimester buprenorphine exposure (mean [SD] maternal age, 28.4 [4.6] years) and 3846 with methadone exposure (mean [SD] maternal age, 28.8 [4.7] years). The risk of malformations overall was 50.9 (95% CI, 46.5-55.3) per 1000 pregnancies for buprenorphine and 60.6 (95% CI, 53.0-68.1) per 1000 pregnancies for methadone. After confounding adjustment, buprenorphine was associated with a lower risk of malformations compared with methadone (RR, 0.82; 95% CI, 0.69-0.97). Risk was lower with buprenorphine for cardiac malformations (RR, 0.63; 95% CI, 0.47-0.85), including both ventricular septal defect (RR, 0.62; 95% CI, 0.39-0.98) and secundum atrial septal defect/nonprematurity-related patent foramen ovale (RR, 0.54; 95% CI, 0.30-0.97), oral clefts (RR, 0.65; 95% CI, 0.35-1.19), and clubfoot (RR, 0.55; 95% CI, 0.32-0.94). Results for neural tube defects were uncertain given low event counts. In secondary analyses, buprenorphine was associated with a decreased risk of central nervous system, urinary, and limb malformations but a greater risk of gastrointestinal malformations compared with methadone. These findings were consistent in sensitivity and bias analyses. Conclusions and Relevance: In this cohort study, the risk of most malformations previously associated with opioid exposure was lower in buprenorphine-exposed infants compared with methadone-exposed infants, independent of measured confounders. Malformation risk is one factor that informs the individualized patient decision regarding medications for opioid use disorder in pregnancy.

Deep Survival Analysis for Interpretable Time-Varying Prediction of Preeclampsia Risk.

Objective: Survival analysis is widely utilized in healthcare to predict the timing of disease onset. Traditional methods of survival analysis are usually based on Cox Proportional Hazards model and assume proportional risk for all subjects. However, this assumption is rarely true for most diseases, as the underlying factors have complex, non-linear, and time-varying relationships. This concern is especially relevant for pregnancy, where the risk for pregnancy-related complications, such as preeclampsia, varies across gestation. Recently, deep learning survival models have shown promise in addressing the limitations of classical models, as the novel models allow for non-proportional risk handling, capturing nonlinear relationships, and navigating complex temporal dynamics. Methods: We present a methodology to model the temporal risk of preeclampsia during pregnancy and investigate the associated clinical risk factors. We utilized a retrospective dataset including 66,425 pregnant individuals who delivered in two tertiary care centers from 2015-2023. We modeled the preeclampsia risk by modifying DeepHit, a deep survival model, which leverages neural network architecture to capture time-varying relationships between covariates in pregnancy. We applied time series k-means clustering to DeepHit's normalized output and investigated interpretability using Shapley values. Results: We demonstrate that DeepHit can effectively handle high-dimensional data and evolving risk hazards over time with performance similar to the Cox Proportional Hazards model, achieving an area under the curve (AUC) of 0.78 for both models. The deep survival model outperformed traditional methodology by identifying time-varied risk trajectories for preeclampsia, providing insights for early and individualized intervention. K-means clustering resulted in patients delineating into low-risk, early-onset, and late-onset preeclampsia groups- notably, each of those has distinct risk factors. Conclusion: This work demonstrates a novel application of deep survival analysis in time-varying prediction of preeclampsia risk. Our results highlight the advantage of deep survival models compared to Cox Proportional Hazards models in providing personalized risk trajectory and demonstrating the potential of deep survival models to generate interpretable and meaningful clinical applications in medicine.

Preeclampsia Prediction Using Machine Learning and Polygenic Risk Scores From Clinical and Genetic Risk Factors in Early and Late Pregnancies.

BACKGROUND: Preeclampsia, a pregnancy-specific condition associated with new-onset hypertension after 20-weeks gestation, is a leading cause of maternal and neonatal morbidity and mortality. Predictive tools to understand which individuals are most at risk are needed. METHODS: We identified a cohort of N=1125 pregnant individuals who delivered between May 2015 and May 2022 at Mass General Brigham Hospitals with available electronic health record data and linked genetic data. Using clinical electronic health record data and systolic blood pressure polygenic risk scores derived from a large genome-wide association study, we developed machine learning (XGBoost) and logistic regression models to predict preeclampsia risk. RESULTS: Pregnant individuals with a systolic blood pressure polygenic risk score in the top quartile had higher blood pressures throughout pregnancy compared with patients within the lowest quartile systolic blood pressure polygenic risk score. In the first trimester, the most predictive model was XGBoost, with an area under the curve of 0.74. In late pregnancy, with data obtained up to the delivery admission, the best-performing model was XGBoost using clinical variables, which achieved an area under the curve of 0.91. Adding the systolic blood pressure polygenic risk score to the models did not improve the performance significantly based on De Long test comparing the area under the curve of models with and without the polygenic score. CONCLUSIONS: Integrating clinical factors into predictive models can inform personalized preeclampsia risk and achieve higher predictive power than the current practice. In the future, personalized tools can be implemented to identify high-risk patients for preventative therapies and timely intervention to improve adverse maternal and neonatal outcomes.

Polycystic Ovary Syndrome Physiologic Pathways Implicated Through Clustering of Genetic Loci.

CONTEXT: Polycystic ovary syndrome (PCOS) is a heterogeneous disorder, with disease loci identified from genome-wide association studies (GWAS) having largely unknown relationships to disease pathogenesis. OBJECTIVE: This work aimed to group PCOS GWAS loci into genetic clusters associated with disease pathophysiology. METHODS: Cluster analysis was performed for 60 PCOS-associated genetic variants and 49 traits using GWAS summary statistics. Cluster-specific PCOS partitioned polygenic scores (pPS) were generated and tested for association with clinical phenotypes in the Mass General Brigham Biobank (MGBB, N = 62 252). Associations with clinical outcomes (type 2 diabetes [T2D], coronary artery disease [CAD], and female reproductive traits) were assessed using both GWAS-based pPS (DIAMANTE, N = 898,130, CARDIOGRAM/UKBB, N = 547 261) and individual-level pPS in MGBB. RESULTS: Four PCOS genetic clusters were identified with top loci indicated as following: (i) cluster 1/obesity/insulin resistance (FTO); (ii) cluster 2/hormonal/menstrual cycle changes (FSHB); (iii) cluster 3/blood markers/inflammation (ATXN2/SH2B3); (iv) cluster 4/metabolic changes (MAF, SLC38A11). Cluster pPS were associated with distinct clinical traits: Cluster 1 with increased body mass index (P = 6.6 × 10-29); cluster 2 with increased age of menarche (P = 1.5 × 10-4); cluster 3 with multiple decreased blood markers, including mean platelet volume (P = 3.1 ×10-5); and cluster 4 with increased alkaline phosphatase (P = .007). PCOS genetic clusters GWAS-pPSs were also associated with disease outcomes: cluster 1 pPS with increased T2D (odds ratio [OR] 1.07; P = 7.3 × 10-50), with replication in MGBB all participants (OR 1.09, P = 2.7 × 10-7) and females only (OR 1.11, 4.8 × 10-5). CONCLUSION: Distinct genetic backgrounds in individuals with PCOS may underlie clinical heterogeneity and disease outcomes.

A Scalable Method for Readable Tree Layouts.

Large tree structures are ubiquitous and real-world relational datasets often have information associated with nodes (e.g., labels or other attributes) and edges (e.g., weights or distances) that need to be communicated to the viewers. Yet, scalable, easy to read tree layouts are difficult to achieve. We consider tree layouts to be readable if they meet some basic requirements: node labels should not overlap, edges should not cross, edge lengths should be preserved, and the output should be compact. There are many algorithms for drawing trees, although very few take node labels or edge lengths into account, and none optimizes all requirements above. With this in mind, we propose a new scalable method for readable tree layouts. The algorithm guarantees that the layout has no edge crossings and no label overlaps, and optimizes one of the remaining aspects: desired edge lengths and compactness. We evaluate the performance of the new algorithm by comparison with related earlier approaches using several real-world datasets, ranging from a few thousand nodes to hundreds of thousands of nodes. Tree layout algorithms can be used to visualize large general graphs, by extracting a hierarchy of progressively larger trees. We illustrate this functionality by presenting several map-like visualizations generated by the new tree layout algorithm.

Prescription medication use during pregnancy in the United States from 2011 to 2020: trends and safety evidence.

BACKGROUND: Medication use during pregnancy has increased in the United States despite the lack of safety data for many medications. OBJECTIVE: This study aimed to inform research priorities by examining trends in medication use during pregnancy and identifying gaps in safety information on the most commonly prescribed medications. STUDY DESIGN: We identified population-based cohorts of commercially (MarketScan 2011-2020) and publicly (Medicaid Analytic eXtract/Transformed Medicaid Statistical Information System Analytic Files 2011-2018) insured pregnancies ending in live birth from 2 health care utilization databases. Medication use was based on filled prescriptions between the date of last menstrual period through delivery, as well as the period before the last menstrual period and during specific trimesters. We also included a cross-sectional representative sample of pregnancies ascertained by the National Health and Nutrition Examination Survey (2011-2020), with information on prescription medication use during the preceding month obtained through maternal interviews. Teratogen Information System was used to classify the available evidence on teratogenic risk. RESULTS: Among over 3 million pregnancies, the medications most commonly dispensed at any time during pregnancy were analgesics, antibiotics, and antiemetics. The top medications were ondansetron (16.8%), amoxicillin (13.5%), and azithromycin (12.4%) in MarketScan, nitrofurantoin (22.2%), acetaminophen (21.3%; mostly as part of acetaminophen-hydrocodone products), and ondansetron (19.5%) in Medicaid Analytic eXtract/Transformed Medicaid Statistical Information System Analytic Files, and levothyroxine (5.0%), sertraline (2.9%), and insulin (2.9%) in the National Health and Nutrition Examination Survey group. The most commonly dispensed suspected teratogens during the first trimester were antithyroid medications. The use of antidiabetic and psychotropic medications has continued to increase in the United States during the last decade, opioid dispensation has decreased by half, and antibiotics and antiemetics continue to be common. For one-quarter of medications, there is insufficient evidence available to characterize their safety profile in pregnancy. CONCLUSION: There is a need for more drug research in pregnant patients. Future research should focus on anti-infectives with high utilization and limited level of evidence on safety for use during pregnancy. Although lack of evidence is not evidence of safety concerns, it does not indicate risk either. In many instances, the benefits outweigh the risks when these medications are used clinically, and some of the medications with no proven safety may be necessary to treat patients.

Zero-shot interpretable phenotyping of postpartum hemorrhage using large language models.

Many areas of medicine would benefit from deeper, more accurate phenotyping, but there are limited approaches for phenotyping using clinical notes without substantial annotated data. Large language models (LLMs) have demonstrated immense potential to adapt to novel tasks with no additional training by specifying task-specific instructions. Here we report the performance of a publicly available LLM, Flan-T5, in phenotyping patients with postpartum hemorrhage (PPH) using discharge notes from electronic health records (n = 271,081). The language model achieves strong performance in extracting 24 granular concepts associated with PPH. Identifying these granular concepts accurately allows the development of interpretable, complex phenotypes and subtypes. The Flan-T5 model achieves high fidelity in phenotyping PPH (positive predictive value of 0.95), identifying 47% more patients with this complication compared to the current standard of using claims codes. This LLM pipeline can be used reliably for subtyping PPH and outperforms a claims-based approach on the three most common PPH subtypes associated with uterine atony, abnormal placentation, and obstetric trauma. The advantage of this approach to subtyping is its interpretability, as each concept contributing to the subtype determination can be evaluated. Moreover, as definitions may change over time due to new guidelines, using granular concepts to create complex phenotypes enables prompt and efficient updating of the algorithm. Using this language modelling approach enables rapid phenotyping without the need for any manually annotated training data across multiple clinical use cases.

Impact of the COVID-19 Pandemic on Childhood Lead Testing and Blood Lead Levels.

OBJECTIVE: To evaluate the effect of the COVID-19 pandemic on childhood lead testing and blood lead levels. METHODS: A retrospective analysis of lead tests and results was performed across 3 urban medical centers during the pre-COVID-19 (March 10, 2019-March 9, 2020) and COVID-19 (March 10, 2020-March 10, 2022) periods. Interrupted time series analysis with quasi-Poisson regression was used to evaluate changes in lead testing between study periods. The relationship between sociodemographic features with detectable (≧2 µg/dL) and elevated (≧3.5 µg/dL) blood lead levels (BLLs) was assessed with multivariable logistic regression. RESULTS: Among a total of 16,364 lead tests across 10,362 patients, weekly testing rates significantly decreased during COVID-19 (relative risk (RR) 0.64, 95% (confidence interval) CI 0.53-0.78). Census tracts with the greatest proportion of pre-1950s housing had a stronger association with detectable BLLs during the COVID-19 period (pre-COVID-19 adjusted odds ratio (aOR) 1.73, 95% CI 1.35-2.20; aOR 2.58, 95% CI 2.13-3.12; interaction P value .014). When limited to 1 year following COVID-19 (March 10, 2020-March 10, 2021), the association between both elevated BLLs (pre-COVID-19: aOR 1.49, 95% CI 0.87-2.53; COVID-19: aOR 3.51, 95% CI 1.98-6.25; interaction P value .032) and detectable BLLs with pre-1950s housing were greater during the COVID-19 period (pre-COVID-19: aOR 1.73, 95% CI 1.35-2.20; COVID-19: aOR 2.56, 95% CI 1.95-3.34; interaction P value .034). CONCLUSIONS: The COVID-19 pandemic led to a significant reduction in lead surveillance and magnified the effect of known risk factors for lead exposure. Concerted clinical, public health, and community advocacy are needed to address care gaps and excess cases of lead poisoning.