Comparison of the efficacy of vaginal micronised progesterone tablet and gel for in vitro fertilisation

BACKGROUND

Luteal phase support (LPS) with progesterone is a generally accepted practice after controlled ovarian stimulation, although the best protocols for LPS have been debated. We aimed to compare the efficacy of vaginal micronised progesterone tablets and 8% vaginal progesterone gel for LPS using real-world data.

METHODS

This retrospective study included 459 in vitro fertilisation/intracytoplasmic sperm injection cycles performed at a university hospital from 2005 to 2019. All cycles were followed by fresh day 3 embryo transfer (ET). Either progesterone tablets or gel was used for LPS. To control the conditional probability of progesterone tablets or gel use, doubly robust inverse probability weighting composed of inverse-probability-of-treatment weighting (IPTW) and regression adjustment (RA). IPTW was performed based on the covariate balancing propensity score (CBPS).

RESULTS

Progesterone tablets were administered in 65 cycles, and progesterone gel was administered in 394 cycles. Women who used progesterone tablets were more likely to be older (36 vs. 34 years), have primary infertility (78.5% vs. 61.4%), use gonadotropin-releasing hormone antagonist (60.0% vs. 43.2%), and have fewer retrieved oocytes (seven vs. nine) and transferred embryos (two vs. three) than participants who used progesterone gel. After IPTW-CBPS and RA analysis for the above covariates, the adjusted odds for clinical pregnancy in women who used progesterone tablets were 1.10 times compared with women who used progesterone gel; however, the 95% confidence interval did not reach statistical significance (0.96-1.26).

CONCLUSIONS

Clinical pregnancy was comparable between vaginal micronised progesterone tablets and vaginal progesterone gel for LPS in fresh day 3 ET cycles.

Prediction of protein biophysical traits from limited data: a case study on nanobody thermostability through NanoMelt

In-silico prediction of protein biophysical traits is often hindered by the limited availability of experimental data and their heterogeneity. Training on limited data can lead to overfitting and poor generalizability to sequences distant from those in the training set. Additionally, inadequate use of scarce and disparate data can introduce biases during evaluation, leading to unreliable model performances being reported. Here, we present a comprehensive study exploring various approaches for protein fitness prediction from limited data, leveraging pre-trained embeddings, repeated stratified nested cross-validation, and ensemble learning to ensure an unbiased assessment of the performances. We applied our framework to introduce NanoMelt, a predictor of nanobody thermostability trained with a dataset of 640 measurements of apparent melting temperature, obtained by integrating data from the literature with 129 new measurements from this study. We find that an ensemble model stacking multiple regression using diverse sequence embeddings achieves state-of-the-art accuracy in predicting nanobody thermostability. We further demonstrate NanoMelt's potential to streamline nanobody development by guiding the selection of highly stable nanobodies. We make the curated dataset of nanobody thermostability freely available and NanoMelt accessible as a downloadable software and webserver.

Archaea methanogens are associated with cognitive performance through the shaping of gut microbiota, butyrate and histidine metabolism

The relationship between bacteria, cognitive function and obesity is well established, yet the role of archaeal species remains underexplored. We used shotgun metagenomics and neuropsychological tests to identify microbial species associated with cognition in a discovery cohort (IRONMET, n = 125). Interestingly, methanogen archaeas exhibited the strongest positive associations with cognition, particularly Methanobrevibacter smithii (M. smithii). Stratifying individuals by median-centered log ratios (CLR) of M. smithii (low and high M. smithii groups: LMs and HMs) revealed that HMs exhibited better cognition and distinct gut bacterial profiles (PERMANOVA p = 0.001), characterized by increased levels of Verrucomicrobia, Synergistetes and Lentisphaerae species and reduced levels of Bacteroidetes and Proteobacteria. Several of these species were linked to the cognitive test scores. These findings were replicated in a large-scale validation cohort (Aging Imageomics, n = 942). Functional analyses revealed an enrichment of energy, butyrate, and bile acid metabolism in HMs in both cohorts. Global plasma metabolomics by CIL LC-MS in IRONMET identified an enrichment of methylhistidine, phenylacetate, alpha-linolenic and linoleic acid, and secondary bile acid metabolism associated with increased levels of 3-methylhistidine, phenylacetylgluamine, adrenic acid, and isolithocholic acid in the HMs group. Phenylacetate and linoleic acid metabolism also emerged in the Aging Imageomics cohort performing untargeted HPLC-ESI-MS/MS metabolic profiling, while a targeted bile acid profiling identified again isolithocholic acid as one of the most significant bile acid increased in the HMs. 3-Methylhistidine levels were also associated with intense physical activity in a second validation cohort (IRONMET-CGM, n = 116). Finally, FMT from HMs donors improved cognitive flexibility, reduced weight, and altered SCFAs, histidine-, linoleic acid- and phenylalanine-related metabolites in the dorsal striatum of recipient mice. M. smithii seems to interact with the bacterial ecosystem affecting butyrate, histidine, phenylalanine, and linoleic acid metabolism with a positive impact on cognition, constituting a promising therapeutic target to enhance cognitive performance, especially in subjects with obesity.

Fecal gut microbiota and amino acids as noninvasive diagnostic biomarkers of Pediatric inflammatory bowel disease

BACKGROUND AND AIMS

Fecal calprotectin (FCP) has limited specificity as diagnostic biomarker of pediatric inflammatory bowel disease (IBD), leading to unnecessary invasive endoscopies. This study aimed to develop and validate a fecal microbiota and amino acid (AA)-based diagnostic model.

METHODS

Fecal samples from a discovery cohort (de novo IBD and healthy controls [HC]) were used to develop the diagnostic model. This model was applied in a validation cohort (de novo IBD and controls with gastrointestinal symptoms [CGI]). Microbiota and AAs were analyzed using interspace profiling and liquid chromatography-mass spectrometry techniques, respectively. Machine learning techniques were used to build the diagnostic model.

RESULTS

In the discovery cohort (58 IBD, 59 hC), two microbial species (Escherichia coli and Alistipes finegoldii) and four AAs (leucine, ornithine, taurine, and alpha-aminoadipic acid [AAD]) combined allowed for discrimination between both subgroups (AUC 0.94, 95% CI [0.89, 0.98]). In the validation cohort (43 IBD, 38 CGI), this panel of six markers could differentiate patients with IBD from CGI with an AUC of 0.84, 95% CI [0.67, 0.95]). Leucine showed the best diagnostic performance (AUC 0.89, 95% CI [0.81, 0.95]).

CONCLUSIONS

Leucine might serve as adjuvant noninvasive biomarker in the diagnostic work-up of pediatric IBD. Future research should investigate whether the combination of leucine with FCP could improve specificity and may help tailor the course of diagnostics.

Predicting high anger intensity using ecological momentary assessment and wearable-derived physiological data in a trauma-affected sample

Background: Digital technologies offer tremendous potential to predict dysregulated mood and behavior within an individual's environment, and in doing so can support the development of new digital health interventions. However, no prediction models have been built in trauma-exposed populations that leverage real-world data.Objective: This project aimed to determine if wearable-derived physiological data can predict anger intensity in trauma-exposed adults.Method: Heart rate variability (i.e. a commercial wearable stress score) was combined with ecological momentary assessment (EMA) data collected over 10 days (n = 84). Five summary measures from stress scores collected 10 min prior to each EMA were selected using factor analysis of 24 candidates.Results: A high area under the receiver operating curve (AUC) was found for a logistic mixed effects model including these measures as predictors, ranging 0.761 (95% CI:0.569-0.921) to 0.899 (95% CI:0.784-0.980) across cross-validation methods.Conclusions: While the predictive performance may be overly optimistic due to the outcome prevalence (13.8%) and requires replication with larger datasets, our promising findings have significant methodological and clinical implications for researchers looking to build novel prediction and treatment approaches to respond to posttraumatic mental health.

Best practices for developing microbiome-based disease diagnostic classifiers through machine learning

The human gut microbiome, crucial in various diseases, can be utilized to develop diagnostic models through machine learning (ML). The specific tools and parameters used in model construction such as data preprocessing, batch effect removal and modeling algorithms can impact model performance and generalizability. To establish an generally applicable workflow, we divided the ML process into three above-mentioned steps and optimized each sequentially using 83 gut microbiome cohorts across 20 diseases. We tested a total of 156 tool-parameter-algorithm combinations and benchmarked them according to internal- and external- AUCs. At the data preprocessing step, we identified four data preprocessing methods that performed well for regression-type algorithms and one method that excelled for non-regression-type algorithms. At the batch effect removal step, we identified the "ComBat" function from the sva R package as an effective batch effect removal method and compared the performance of various algorithms. Finally, at the ML algorithm selection step, we found that Ridge and Random Forest ranked the best. Our optimized work flow performed similarly comparing with previous exhaustive methods for disease-specific optimizations, thus is generally applicable and can provide a comprehensive guideline for constructing diagnostic models for a range of diseases, potentially serving as a powerful tool for future medical diagnostics.

Gut microbiome-driven regulation of sex hormone homeostasis: a potential neuroendocrine connection

The gut microbiome is known to have a bidirectional relationship with sex hormone homeostasis; however, its role in mediating interactions between the primary regulatory axes of sex hormones and their productions is yet to be fully understood. We utilized both conventionally raised and gnotobiotic mouse models to investigate the regulatory role of the gut microbiome on the hypothalamic-pituitary-gonadal (HPG) axis. Male and female conventionally raised mice underwent surgical modifications as follows: (1) hormonally intact controls; (2) gonadectomized males and females; (3) gonadectomized males and females supplemented with testosterone and estrogen, respectively. Fecal samples from these mice were used to colonize sex-matched, intact, germ-free recipient mice through fecal microbiota transplant (FMT). Serum gonadotropins, gonadal sex hormones, cecal microbiota, and the serum global metabolome were assessed. FMT recipients of gonadectomized-associated microbiota showed lower circulating gonadotropin levels than recipients of intact-associated microbiota, opposite to that of FMT donors. FMT recipients of gonadectomized-associated microbiota also had greater testicular weights compared to recipients of intact-associated microbiota. The gut microbiota composition of recipient mice differed significantly based on the FMT received, with the male microbiota having a more concerted impact in response to changes in the HPG axis. Network analyses showed that multiple metabolically unrelated pathways may be involved in driving differences in serum metabolites due to sex and microbiome received in the recipient mice. In sum, our findings indicate that the gut microbiome responds to the HPG axis and subsequently modulates its feedback mechanisms. A deeper understanding of interactions between the gut microbiota and the neuroendocrine-gonadal system may contribute to the development of therapies for sexually dimorphic diseases.

Effects of angiotensin-neprilysin inhibition in women vs men: Insights from PARAGLIDE-HF

BACKGROUND

Sub-analyses of key trials suggest a preferential benefit for specific heart failure with preserved ejection fraction (HFpEF) therapies in women. This work investigated treatment effects between women and men in the PARAGLIDE-HF (Prospective comparison of ARNI with ARB Given following stabiLization In DEcompensated HFpEF) trial.

METHODS

In this prespecified subgroup analysis, we examined outcomes according to sex in the PARAGLIDE-HF trial. The primary endpoint was time-average proportional change in amino terminal pro-B type natriuretic peptide (NT-proBNP) from baseline through Weeks 4 and 8. We also examined secondary outcomes and tolerability.

RESULTS

Overall, 242 women (52%) and 224 men (48%) were randomized. Women had significantly higher LVEF, worse renal function, and less comorbidities than men. In the overall population, the time-averaged reduction in NT-proBNP was significantly greater for sacubitril/valsartan (sac/val) than valsartan (ratio of change 0.85, 95% CI, 0.73-0.999). When examined according to sex, the time-averaged reduction in NT-proBNP was numerically greater with sac/val in both women (ratio of change = 0.86, 95% CI, 0.69-1.070) and men (ratio of change 0.84, 95% CI, 0.67-1.05) with no differential treatment effect (P interaction = .91). Similarly, the secondary hierarchical endpoint favored sac/val over valsartan in both women and men but was not statistically significant. Study drug dosage levels were similar across women and men and there were no sex-specific differences in the incidence of adverse events.

CONCLUSIONS

In patients with mildly reduced or preserved EF >40% and a recent worsening HF event, the efficacy, safety and tolerability of sac/val vs valsartan were similar in both women and men, suggesting consistent effects across appropriately selected patients regardless of sex. Future prospective studies are needed to further evaluate sex-specific differences in treatment response of HFpEF therapies.

TRIAL REGISTRATION

Prospective comparison of ARNI with ARB Given following stabiLization In DEcompensated HFpEF; NCT03988634; https://www.

CLINICALTRIALS

gov/study/NCT03988634.

A non-parametric U-statistic testing approach for multi-arm clinical trials with multivariate longitudinal data

Randomized clinical trials (RCTs) often involve multiple longitudinal primary outcomes to comprehensively assess treatment efficacy. The Longitudinal Rank-Sum Test (LRST) [17], a robust U-statistics-based, non-parametric, rank-based method, effectively controls Type I error and enhances statistical power by leveraging the temporal structure of the data without relying on distributional assumptions. However, the LRST is limited to two-arm comparisons. To address the need for comparing multiple doses against a control group in many RCTs, we extend the LRST to a multi-arm setting. This novel multi-arm LRST provides a flexible and powerful approach for evaluating treatment efficacy across multiple arms and outcomes, with a strong capability for detecting the most effective dose in multi-arm trials. Extensive simulations demonstrate that this method maintains excellent Type I error control while providing greater power compared to the two-arm LRST with multiplicity adjustments. Application to the Bapineuzumab (Bapi) 301 trial further validates the multi-arm LRST's practical utility and robustness, confirming its efficacy in complex clinical trial analyses.

Relationships between substance use treatment facilities and alcohol-attributable mortality across U.S. counties

BACKGROUND

Formal substance use treatment is a key resource for recovery among people with alcohol use disorders. Limited county-level availability of substance use treatment facilities may restrict access to care and ultimately contribute to worsening health outcomes and mortality. However, it is unknown whether the availability of such facilities is associated with county-level alcohol-attributable mortality risk.

METHODS

We used Bayesian hierarchical Poisson spatial regression models to assess the relationship between population-weighted county-level treatment facility availability and rates of (1) fully chronic alcohol-attributable mortality, (2) alcohol poisonings, and (3) suicides by exposure to alcohol in 2019-2020. Localized treatment facility availability was calculated using a weighted method incorporating Census block group-level population counts. We adjusted for county-level demographic and socioeconomic factors, hospital density, population density, overall mortality rate, densities of mental health practitioner offices, U.S. Census region, year, and season.

RESULTS

There was county-level heterogeneity in the availability of substance use treatment facilities, with northeastern county treatment facility densities at least twice as high as other regions. Higher county-level densities of treatment facilities were related to increased county-level risk for chronic fully alcohol-attributable deaths and alcohol poisonings but not suicides by exposure to alcohol.

CONCLUSIONS

Availability of substance use treatment facilities and the services they offer is heterogeneous across U.S. counties. The positive relationship between population-weighted county-level densities of treatment facilities and chronic fully alcohol-attributable mortality and alcohol poisonings may suggest that treatment facilities are placed in areas of greatest demand; yet, population-level needs may not fully met by these facilities.

Learning double balancing representation for heterogeneous dose-response curve estimation

Estimating the individuals' potential response to varying treatment doses is crucial for decision-making in areas such as precision medicine and management science. Most recent studies predict counterfactual outcomes by learning a covariate representation that is independent of the treatment variable. However, such independence constraints neglect much of the covariate information that is useful for counterfactual prediction, especially when the treatment variables are continuous. To tackle the above issue, in this paper, we first theoretically demonstrate the importance of the balancing and prognostic representations for unbiased estimation of the heterogeneous dose-response curves, that is, the learned representations are constrained to satisfy the conditional independence between the covariates and both of the treatment variables and the potential responses. Based on this, we propose an end-to-end Contrastive balancing Representation learning Network (CRNet) and a three-stage Weighted Double Balancing Network (WDBN) using a partial distance measure, for estimating the heterogeneous dose-response curves without losing the continuity of treatments. Extensive experiments are conducted on synthetic and real-world datasets demonstrating that our proposal significantly outperforms previous methods. Code is available at: https://github.com/euzmin/Contrastive-Balancing-Representation-Network-CRNet.

Towards a better evaluation of out-of-domain generalization

The objective of Domain Generalization (DG) is to devise algorithms capable of achieving high performance on previously unseen test distributions. In the pursuit of this objective, average measure has been employed as the prevalent measure for comparing algorithms in the existing DG studies. Despite its significance, a comprehensive exploration of the average measure has been lacking and its suitability in approximating the true domain generalization performance has been questionable. In this study, we carefully investigate the limitations inherent in the average measure and propose worst+gap measure as a robust alternative. We establish theoretical grounds of the proposed measure by deriving two theorems starting from two different assumptions. Despite the independence in the two assumptions, we will show that both theorems lead to a common insight. We conduct extensive experimental investigations to compare the proposed worst+gap measure with the conventional average measure. Given the indispensable need to access the true DG performance for studying measures, we modify five existing datasets to come up with SR-CMNIST, C-Cats&Dogs, L-CIFAR10, PACS-corrupted, and VLCS-corrupted datasets. The experiment results unveil an inferior performance of the average measure in approximating the true DG performance and confirm the robustness of the theoretically supported worst+gap measure.

Dietary and lifestyle determinants of vitamin D status in the UK Biobank Cohort study for predictive modeling

Vitamin D (VD) is involved in a wide variety of physiological processes. The high prevalence of VD deficiency in the population requires stronger preventive measures. The aim was to characterize the dietary and lifestyle determinants of VD levels in blood and of VD deficiency to further develop predictive models of these two outcomes. A total of 63,759 participants from the UK Biobank study with available data on dietary intake of VD, assessed via 24-hour recalls, and with measurements of serum 25(OH)D levels were included. Linear and logistic regression models were applied to identify factors associated with VD levels and VD deficiency outcomes, and to evaluate the influence of covariates on the association between VD in serum and VD in the diet. Predictive models for both VD outcomes were constructed using classical regression models and machine learning methods based on penalized likelihood methods. Approximately 10% of the participants had VD deficiency (VD < 25 nmol/L), and 38.9% were at risk of VD inadequacy (VD 25-49 nmol/L). The dietary intake of VD was significantly lower in the VD deficient group. This latter group showed lower engagement in physical activity (22.1%) compared to the non-deficient group (13.4%; P<.001). Also, overweight and obesity (vs normal weight) were related to a greater likelihood of VD deficiency (OR=1.18 and 1.96, respectively). A similar odds of VD deficiency was observed for abdominal obesity (OR=1.83). A weaker association was observed between dietary VD intake, based on participant reports, and VD levels. With regard to sunlight exposure, darker skin tones (OR dark vs fair skin=3.11), season (OR winter vs autumn=3.76) and less outdoor time activities (OR per 1 h increase=0.96) were also related to VD deficiency. Predictive models for both classical regression and machine learning, showed good accuracy (AUC=0.8-0.9 for VD deficiency). In conclusion, while a rich diet in VD boosts its levels, sun exposure plays a more significant role particularly in populations from the UK or Northern Europe. A predictive model including key determinants could effectively assess VD deficiency.

A prompt regularization approach to enhance few-shot class-incremental learning with Two-Stage Classifier

With a limited number of labeled samples, Few-Shot Class-Incremental Learning (FSCIL) seeks to efficiently train and update models without forgetting previously learned tasks. Because pre-trained models can learn extensive feature representations from big existing datasets, they offer strong knowledge foundations and transferability, which makes them useful in both few-shot and incremental learning scenarios. Additionally, Prompt Learning improves pre-trained deep learning models' performance on downstream tasks, particularly in large-scale language or vision models. In this paper, we propose a novel Prompt Regularization (PrRe) approach to maximize the fusion of prompts by embedding two different prompts, the Task Prompt and the Global Prompt, inside a pre-trained Vision Transformer (ViT). In the classification phase, we propose a Two-Stage Classifier (TSC), utilizing K-Nearest Neighbors for base session and a Prototype Classifier for incremental sessions, integrated with a global self-attention module. Through experiments on multiple benchmark tests, we demonstrate the effectiveness and superiority of our method. The code is available at https://github.com/gyzzzzzzzz/PrRe.

An information-theoretic approach for heterogeneous differentiable causal discovery

With the advancement of deep learning, a variety of differential causal discovery methods have emerged, inevitably attracting more attention for their excellent scalability and interpretability. However, these methods often struggle with complex heterogeneous datasets that exhibit environmental diversity and are characterized by shifts in noise distribution. To this end, we introduce a novel information-theoretic approach designed to enhance the robustness of differential causal discovery methods. Specifically, we integrate Minimum Error Entropy (MEE) as an adaptive error regulator into the structure learning framework. MEE effectively reduces error variability across diverse samples, enabling our model to adapt dynamically to varying levels of complexity and noise. This adjustment significantly improves the precision and stability of the model. Extensive experiments on both synthetic and real-world datasets have demonstrated significant performance enhancements over existing methods, affirming the effectiveness of our approach. The code is available at https://github.com/ElleZWQ/MHCD.

Molecular networking: An efficient tool for discovering and identifying natural products

Natural products (NPs), play a crucial role in drug development. However, the discovery of NPs is accidental, and conventional identification methods lack accuracy. To overcome these challenges, an increasing number of researchers are directing their attention to Molecular networking (MN). MN based on secondary mass spectrometry has become an important tool for the separation, purification and structural identification of NPs. However, most new tools are not well known. This review started with the most basic MN tool and explains it from the principle, workflow, and application. Then introduce the principles and workflows of the remaining eight new types of MN tools. The reliability of various MNs is mainly verified based on the application of phytochemistry and metabolomics. Subsequently, the principles and applications of 12 structural annotation tools are introduced. For the first time, the scope of 9 kinds of MN tools is compared horizontally, and 12 kinds of structured annotation tools are classified from the type of compound structure suitable for identification. The advantages and disadvantages of various tools are summarized, and make suggestions for future application directions and the development of computing tools in this review. MN tools are expected to enhance the efficiency of the discovery and identification in NPs.

Deep Huber quantile regression networks

Typical machine learning regression applications aim to report the mean or the median of the predictive probability distribution, via training with a squared or an absolute error scoring function. The importance of issuing predictions of more functionals of the predictive probability distribution (quantiles and expectiles) has been recognized as a means to quantify the uncertainty of the prediction. In deep learning (DL) applications, that is possible through quantile and expectile regression neural networks (QRNN and ERNN respectively). Here we introduce deep Huber quantile regression networks (DHQRN) that nest QRNN and ERNN as edge cases. DHQRN can predict Huber quantiles, which are more general functionals in the sense that they nest quantiles and expectiles as limiting cases. The main idea is to train a DL algorithm with the Huber quantile scoring function, which is consistent for the Huber quantile functional. As a proof of concept, DHQRN are applied to predict house prices in Melbourne, Australia and Boston, United States (US). In this context, predictive performances of three DL architectures are discussed along with evidential interpretation of results from two economic case studies. Additional simulation experiments and applications to real-world case studies using open datasets demonstrate a satisfactory absolute performance of DHQRN.

Establishment and validation of a clinical prediction model for colorectal adenoma risk factors

Colorectal adenomas are benign tumors of the colorectal mucosal epithelium that have malignant potential and are regarded as precancerous lesions of colorectal cancer, for which the specific risk factors are unclear. The present study aimed to identify independent risk factors for colorectal adenoma to develop a prediction model and test its predictive value. A retrospective analysis was performed using data from patients who underwent electronic colonoscopy at the Department of Proctology (Affiliated Hospital of Shandong University of Traditional Chinese Medicine; Jinan, China) from January 2013 to December 2023 and had polyps removed during colonoscopy. Patients with colorectal adenoma were included in the case group, whilst those with no visible abnormalities on endoscopy or with non-adenomatous polyps were included as a control group. The patients were randomly divided into a training and validation group in a 7:3 ratio. Variables were screened using single-component analysis and the filtered variables were employed in multivariate logistic regression to create a clinical prediction model. Finally, the model was internally and externally validated. A total of 730 patients were included in the present study, with 286 assigned to the case group and 444 to the control group. After the initial screening of 39 variables, 12 continued to the next round, resulting in four potential predictors including age, daily number of bowel movements, thrombin time and the number of polyps. A prediction model was created based on these variables. Regarding internal validation, the C-index was 0.7054 [95% confidence interval (CI), 0.6596-0.7512] and the prediction probability in the calibration curve was close to the diagonal line of the calibration graph, indicating that the prediction probability of the model was reasonable. Regarding external validation, the C-index in the validation cohort was 0.6306 (95% CI, 0.5560-0.7053) and the calibration curve also demonstrated good identification capabilities. The Hosmer-Lemeshow test revealed that the model had a reasonable calibration degree, with χ2=9.7893, degree of freedom=8 and P=0.28. The receiver operating characteristic curve and decision curve analysis for the training and validation cohorts demonstrated good efficacy and an ideal application value. In conclusion, the model constructed in the present study demonstrated moderate predictive accuracy for colorectal adenoma risk, laying the groundwork for early detection of colorectal adenoma and secondary prevention of colorectal cancer.

Decade-long variation in the development of complications after cardiac surgery across the United States

BACKGROUND

Interhospital variation in complications after cardiac surgery may reflect differences in quality across centers. Although these differences have been observed for years, changes in variation over time have not been quantified. In the present study, we assessed trends in center-level variation of major adverse events after cardiac surgery and associations with hospital volume.

METHODS

Adult (≥18 years) hospitalizations entailing elective cardiac surgery (coronary artery bypass grafting, valve repair/replacement) were identified in the 2012-2021 National Inpatient Sample. Hospitals were stratified into quartiles using annual cardiac case volume. Major adverse events was defined as in-hospital mortality or any complication (cardiac, respiratory, stroke, infectious, thromboembolic, intraoperative). A hierarchical logistic regression model evaluated the variation in major adverse events attributable to hospital effects.

RESULTS

Of 1,816,755 patients undergoing cardiac surgery, 28.5% experienced major adverse events. Rates of major adverse events decreased over the study period from 33.3% (2012) to 23.3% (2021). Variation of major adverse events attributable to hospital effects was 7.3%. Major adverse event rates decreased at centers in the greatest-volume quartile (33.5 to 22.8%) and the lowest (37.5 to 36.0%) from 2012 to 2021 (both nptrend <0.001). The overall spread in unadjusted major adverse events rate by center decreased from a standard deviation of 15.7% in 2012 to 13.0% in 2021 (nptrend <0.001). After multivariable risk adjustment, the decrease in center-level variation of major adverse events persisted (standard deviation 6.9% in 2012 to 5.6% in 2021; nptrend <0.001).

CONCLUSION

Interhospital variation in major adverse events declined significantly over the decade. Future efforts should focus on underperforming centers to improve consistency in cardiac surgery outcomes.

Evaluating Semi-Markov Processes and Other Epidemiological Time-to-Event Models by Computing Disease Sojourn Density as Partial Differential Equations

IntroductionEpidemiological models benefit from incorporating detailed time-to-event data to understand how disease risk evolves. For example, decompensation risk in liver cirrhosis depends on sojourn time spent with cirrhosis. Semi-Markov and related models capture these details by modeling time-to-event distributions based on published survival data. However, implementations of semi-Markov processes rely on Monte Carlo sampling methods, which increase computational requirements and introduce stochastic variability. Explicitly calculating the evolving transition likelihood can avoid these issues and provide fast, reliable estimates.MethodsWe present the sojourn time density framework for computing semi-Markov and related models by calculating the evolving sojourn time probability density as a system of partial differential equations. The framework is parametrized by commonly used hazard and models the distribution of current disease state and sojourn time. We describe the mathematical background, a numerical method for computation, and an example model of liver disease.ResultsModels developed with the sojourn time density framework can directly incorporate time-to-event data and serial events in a deterministic system. This increases the level of potential model detail over Markov-type models, improves parameter identifiability, and reduces computational burden and stochastic uncertainty compared with Monte Carlo methods. The example model of liver disease was able to accurately reproduce targets without extensive calibration or fitting and required minimal computational burden.ConclusionsExplicitly modeling sojourn time distribution allows us to represent semi-Markov systems using detailed survival data from epidemiological studies without requiring sampling, avoiding the need for calibration, reducing computational time, and allowing for more robust probabilistic sensitivity analyses.HighlightsTime-inhomogeneous semi-Markov models and other time-to-event-based modeling approaches can capture risks that evolve over time spent with a disease.We describe an approach to computing these models that represents them as partial differential equations representing the evolution of the sojourn time probability density.This sojourn time density framework incorporates complex data sources on competing risks and serial events while minimizing computational complexity.

Maternal thyroid dysfunction and depressive symptoms during pregnancy and child behavioral and emotional problems - an ECHO multi-cohort investigation

BACKGROUND

Maternal thyroid dysfunction and maternal depression during pregnancy may increase the risk of child behavioral and emotional problems. We sought to investigate the independent and interactive associations of these two risk factors with child behavior problems.

METHODS

We combined data from four cohorts in the Environmental influences on Child Health Outcomes (ECHO) program (N = 949). Maternal thyroid function (thyroid-stimulating hormone [TSH], free thyroxine [fT4], thyroid peroxidase autoantibodies [TPO-Ab], fT4/TSH ratio) was measured predominantly during the first half of pregnancy. We harmonized maternal depression into a continuous measure of antepartum depressive symptomatology and a dichotomous measure reflecting (history of) clinical depression. Child internalizing and externalizing problems were harmonized to the T-score metric of the Child Behavior Checklist. We used multiple linear regression and random effects meta-analysis to assess the average relationship between each predictor and outcome, and the variability in these relationships across cohorts.

RESULTS

Across cohorts, antepartum depressive symptomatology was positively associated with both internalizing (meta B = 2.879, 95 % CI 1.87-3.89, p < .001) and externalizing problems (meta B = 1.683, 95 % CI 0.67-2.69, p = .001). None of the indicators of maternal thyroid function was associated with child behavior problems across cohorts. TPO-Ab concentrations were positively associated with child externalizing problems only in offspring of depressed mothers (meta B = 3.063, 95 % CI 0.73-5.40, p = .010).

CONCLUSIONS

This study supports the importance of maternal antepartum mental health for child behavior across diverse populations. However, we found little empirical evidence for an association between maternal thyroid function within the normal range during pregnancy and child behavioral problems.

Mothers Inspiring Healthy Actions (MIHA) program: A randomized control trial testing obesity treatment for mother-daughter dyads

BACKGROUND

Obesity and diabetes are highly concordant in Mexican American families. From a family systems perspective, family-level approaches to obesity treatment can improve the adoption and maintenance of weight management behaviors. Hence, this study will conduct a randomized control trial testing the efficacy of an adapted behavioral weight management intervention with counseling on family functioning.

METHODS

Mexican American mothers and adult daughters will be randomly assigned to receive behavioral obesity treatment with (BTR) or without (BT) relationship skills training. Dyads participating in BTR and BT will attend 24 weekly sessions focused on nutrition and physical activity education along with cognitive and behavioral change strategies. Dyads participating in BTR will also receive experiential-based relationship skills training to address culturally influenced relational dynamics that may interfere with dyadic collaboration for lifestyle change. The 12-month trial will consist of an intervention phase (1-6 months) and a maintenance phase (7-12 months). The primary outcome is weight loss. Secondary outcomes include treatment adherence, cardiometabolic risk factors, health behaviors, psychosocial factors, and family functioning.

CONCLUSION

Dyads in the BTR group are expected to achieve greater improvements in primary and secondary outcomes than the BT group.

Biologics for the Treatment of Moderate-to-Severe Plaque Psoriasis: A Systematic Review and Network Meta-analysis

INTRODUCTION

Moderate-to-severe plaque psoriasis is a chronic disease impacting quality of life (QoL). This network meta-analysis (NMA) compared efficacy and safety of all biologics approved for the treatment of moderate-to-severe plaque psoriasis to better inform providers on mid-term outcomes, with a focus on the interleukin-23 p19 inhibitor tildrakizumab.

METHODS

MEDLINE®, Embase, and CENTRAL were searched for randomized clinical trials (RCT) from inception through January 2024. RCTs comparing biologics against placebo or each other reporting Psoriasis Area and Severity Index (PASI), Physician Global Assessment (PGA) 0/1, or Dermatology Life Quality Index (DLQI) 0/1 responses and safety outcomes (adverse events [AEs] or serious AEs [SAEs]) were sought. Bayesian NMAs were performed at week 28 as the primary time point of interest. Analyses were also performed at weeks 12 and 16. Findings were expressed as risk ratios (RR; efficacy outcomes), risk differences (RD; safety outcomes), and numbers needed to treat (NNT) with 95% credible intervals.

RESULTS

Of 7418 publications screened, 187 describing 124 RCTs of 12 biologics were included in the systematic literature review, and 103 RCTs were included for NMA. All treatments demonstrated improved efficacy and QoL vs. placebo at week 28. Tildrakizumab efficacy at week 28 was comparable to risankizumab and guselkumab, respectively, for PASI 75 (RR 8.74 vs. 8.92 and 8.91), PASI 90 (RR 14.09 vs. 14.81 and 14.77), and PGA 0/1 (RR 9.34 vs. 10.29 and 10.23). No biologics exhibited an increased risk of SAEs vs. placebo; tildrakizumab exhibited no increased risk vs. placebo for AEs.

CONCLUSIONS

The investigated biologics demonstrated improved efficacy and QoL relative to placebo at week 28, with no increased risk of SAEs vs. placebo through week 16. At week 28, efficacy of tildrakizumab, risankizumab, and guselkumab was comparable. Limitations include lack of placebo comparators after week 12 or 16, which could affect results.

Identification and external validation of tumor DNA methylation panel for the recurrence risk stratification of stage II colon cancer

BACKGROUND

Tailoring surveillance and treatment strategies for stage II colon cancer (CC) after curative surgery remains challenging, and personalized approaches are lacking. We aimed to identify a gene methylation panel capable of stratifying high-risk stage II CC patients for recurrence beyond traditional clinical variables.

METHODS

Genome-wide tumor tissue DNA methylation data were analyzed from 562 stage II CC patients who underwent surgery in Germany (DACHS study). The cohort was divided into a training set (N = 395) and an internal validation set (N = 131), with external validation performed on 97 stage II CC patients from Spain. DNA methylation markers were primarily selected using the Elastic Net Cox model. The resulting prognostic index (PI), a combination of clinical factors and selected methylation markers, was compared to baseline models using clinical variables or microsatellite instability (MSI), with discrimination and prediction accuracy assessed through time-dependent receiver operating characteristic curves (AUC) and Brier scores.

RESULTS

The final PI incorporated age, sex, tumor stage, location, and 27 DNA methylation markers. The PI consistently outperformed the baseline model including age, sex, and tumor stage in time-dependent AUC across validation cohorts (e.g., 1-year AUC and 95 % confidence interval: internal validation set, PI: 0.66, baseline model: 0.52; external validation set, PI: 0.72, baseline model: 0.64). In internal validation, the PI also showed a consistently improved time-dependent AUC compared with a combination of MSI and tumor stage only. Nevertheless, the PI did not improve the prediction accuracy of CC recurrence compared to the baseline model.

CONCLUSIONS

This study identified 27 tumor tissue DNA methylation biomarkers that improved the discriminative power in classifying recurrence risk among stage II colon cancer patients. While this methylation panel alone lacks sufficient prediction accuracy for clinical application, its discriminative improvement suggests potential value as part of a multimodal risk-stratification tool.

A novel integration of brain structural and functional connectivity for identifying traumatic brain injury induced perturbations

BACKGROUND

The ability of the brain to perform multiple complex tasks with fixed structures has yet to be fully elucidated. Structural connectivity (SC) and functional connectivity (FC) have been increasingly used to understand the structure and function of the brain respectively. However, a limited number of studies have explored the relationship between both entities especially in translational animal models.

NEW METHOD

We proposed an integration of both SC and FC can improve understanding of brain's structure, function, their interplay, and brain's response to neurological conditions such as traumatic brain injury (TBI). We investigated structure-function correlation at multiple scales (small: cortical regions, medium: resting state networks, and large: hemispheric and whole brain), and adapted a Bayesian framework to incorporate SC for constructing structurally-informed FC (siFC) using a translational porcine model.

RESULTS

There is a significantly strong correlation r = 0.277 ± 0.011 between SC and FC in healthy pigs which is consistent across different scales. Further, siFC stability is measured as a Pearson correlation (r = 0.72 ± 0.07) between time-resolved FCs. Subsequent differential degree test analysis using siFC provided more explicit profiling of perturbations caused by TBI.

COMPARING WITH EXISTING METHODS

The siFC is more immune to large, dynamic variability than FC alone. A more accurate profiling of significantly altered connections and affected hubs by TBI is achieved which is consistent with TBI induced structural deformations.

CONCLUSION

Our findings demonstrated that SC-FC integration model improved detection of significant differences in brain connectivity and pinpoints hub regions that had been directly impacted by TBI.

Association of psychosocial risk factors with acute outcomes of elective cancer resection in the United States

BACKGROUND

Psychosocial risk factors, including psychiatric disorders, substance use, limited cognitive comprehension, and low socioeconomic or uninsured status, are increasingly recognized in cancer care. However, their independent effects on acute postoperative outcomes after elective cancer surgery remain unclear. This study evaluates the association between psychosocial risk factors and clinical and financial outcomes in patients undergoing cancer resections.

METHODS

All adult (≥18 years) records entailing elective resection for lung, esophageal, gastric, pancreatic, hepatic, and colon cancer were tabulated from the 2016-2021 Nationwide Readmissions Database. After entropy balancing, multivariable regression models were developed to ascertain the independent association of psychosocial risk factors. with mortality, complications, length of stay, and nonhome discharge.

RESULTS

Of ∼655,376 patients, 223,035 (34.2%) were considered to comprise psychosocial risk factors. Relative to others, the psychosocial risk factors cohort was more commonly female (51.9 vs 47.1%, P < .001), and had a greater Elixhauser Comorbidity Index (4 [3-5] vs 3 [2-4], P < .001). Both groups most frequently underwent resection for colectomy (51.9% vs 51.1%%, P < .001), yet those with psychosocial risk factors had greater rates of lobectomy (27.8% vs 25.8%, P < .001), compared with their counterparts. After entropy balancing, psychosocial risk factors were linked to greater odds of mortality (adjusted odds ratio, 1.43, 95% confidence interval, 1.31-1.57), respiratory (adjusted odds ratio, 1.25; 95% confidence interval, 1.21-1.31) and infectious (adjusted odds ratio, 1.17; 95% confidence interval, 1.12-1.21) complications. Furthermore, patients with psychosocial risk factors faced incrementally increased resource use.

CONCLUSION

Psychosocial risk factors independently predict adverse clinical and financial outcomes after elective cancer operations. Systematic screening for psychosocial risk factors may facilitate targeted interventions to improve care for high-risk patients.

Preschoolers Selectively Attend to Speech That They Can Learn More From

We introduce a novel method to test a classic idea in developmental science that children's attention to a stimulus is driven by how much they can learn from it. Preschoolers (4-6 years,M = 4.6 ${\it M}=4.6$ ) watched a video where a distracting animation accompanied static, page-by-page illustrations of a storybook. The audio narration for each storybook page was looped so that children could listen to it up to six times in total. However, the narration automatically ended if the child looked at the distractor for an extended period of time, indicating their loss of attention to the story, and triggering the next page. The complexity of the narration was manipulated between-subjects: The Simple narration largely contained words that should be familiar to preschoolers, while the Complex narration contained many rare, late-acquired words. Children's learning was measured via post-tests of their plot comprehension and ability to generalize the embedded rare words. Consistent with the hypothesis that children's attention was driven at least partly by their ability to learn from the speech, we observed a significant interaction between narration complexity and age in predicting children's probability of continuing listening on each page, and the proportion of their visual attention that they devoted to the story illustration, over the animated distractor. That is, while younger children were more likely to continue listening to the Simple speech, older children became increasingly likely to sustain attention to the Complex speech. Our results provide evidence that young children may actively direct their attention toward linguistic input that is most appropriate for their current level of cognitive and linguistic development, which may provide the best learning opportunities.

Optical water typing in optically complex waters: A case study of Chesapeake Bay

Optical water typing has been widely used in aquatic research to classify water bodies based on their inherent optical properties as perceived through satellite-based measures of water color. While optical water type (OWT) classifications have primarily been used to better understand water color dynamics and improve satellite-based estimates of water clarity, chlorophyll a, and other optically active constituents, its potential for broader water quality assessment has received less attention. In this study, we examine the relationships between a suite of water quality parameters, including nutrient concentrations, and OWTs in Chesapeake Bay, an optically complex temperate estuary with an extensive water quality monitoring program. Using machine learning, we grouped Rrs data into ten dominant OWTs; the optimum number of clusters identified from a statistical within-cluster dispersion test. These OWTs ranged from brown to blue/green estuarine waters and emerged with high spatial contiguity. By analyzing synchronously measured discrete water quality variables grouped by corresponding OWTs, unexpected patterns became evident. Notably, total nitrogen concentration emerged as having statistically significant differences between OWTs, suggesting our approach can enhance understanding of nutrient pollution at the scale of a large optically complex estuary, especially in times of reduced fixed sampling routines (e.g., winter). This study aids in the interpretation of Bay-wide water quality trends, can assist in the dynamic selection of water quality retrieval algorithms, and provides high resolution data to identify regions of water quality impairment.

Prior knowledge guided logistic regression model with group lasso penalty for modeling epilepsy disease prediction

"Small sample size, high dimension" data bring tremendous challenges to epilepsy Electroencephalography (EEG) data analysis and seizure onset prediction. Commonly, sparsity technique is introduced to tackle the problem. In this paper, we construct a indicator matrix acting as prior knowledge to assist logistic regression model with group lasso penalty to implement seizure prediction. The proposed method selects the feature at the group level, and it achieves the seizure prediction based on the important feature groups, recognizes the unknown clusters properly and performs well for both synthetic data following Bernoulli distribution and dataset CHB-MIT.

Relationships between numerical score and free text comments in student evaluations of teaching: A sentiment topic analysis reveals the influence of gender and culture

Student evaluations of teaching (SET) have been widely used by university staff to inform decisions on hiring and promotion. In recent years, an increasing body of research has revealed that student evaluations may be systemically affected by students' own conscious or unconscious biases. In this article, we study a data set from an Australian university, where both numerical and text survey responses were available in large quantities. Our study directly linked comments to numerical ratings, we developed approaches to convert text to quantitative data in the form of topics and sentiment scores, and make use of Bayesian ordinal regression techniques to identify drivers of SET scores. Our analysis of text identified 6 teaching dimensions that students discuss in their comments. Our findings suggest that students' SET ratings were correlated primarily with the personal characteristics of the lecturer (such as approachability, and being nice) than measures related to teaching dimensions such as course content and assessment. We found a positive gender effect towards the majority gender in a faculty, possibly reflecting students' gendered expectations. Finally we found that lecturers with a non-English language background were consistently rated lower by the student population, and this effect manifests strongly in local students.

Developmental Behavioural Plasticity and DNA Methylation Patterns in Response to Predation Stress in Trinidadian Guppies

Early-life experiences can predict the environments experienced later in life, giving individuals an opportunity to develop adaptive behaviour appropriate to a likely future environment. Epigenetic mechanisms such as DNA methylation (DNAm) have been implicated in developmental behavioural plasticity; however, studies investigating this possibility are limited in taxonomic breadth and ecological relevance. We investigated the impact of early-life exposure to predation stress on behaviour and DNAm in the brains of Trinidadian guppies (Poecilia reticulata). We exposed guppies throughout development to either an alarm cue (conspecific skin extract), inducing predation stress, or a control cue (water) for 8 weeks and then raised them to adulthood under identical conditions. Then, we conducted two behavioural assays, an open-field and a grouping test, before performing whole-genome bisulfite sequencing on whole brains. Guppies exposed to the alarm cue during development exhibited increased grouping (shoaling) in adulthood compared to those exposed to the control treatment, but there were no detectable impacts on activity, boldness, or exploratory behaviour. We also identified stable shifts in brain DNAm in response to developmental alarm cue exposure in genes involved in behavioural regulation. Some differentially methylated sites were significantly associated with shoaling propensity in both males and females. Additionally, males and females differed in the magnitude of DNAm responses and the genes impacted, suggesting distinct roles for DNAm between the sexes. This study shows how early-life predation stress can induce behavioural changes in adulthood and that shifts in neural DNAm could be an underlying mechanism responsible for these changes.

KRAS mutation status integrated with RNA subtyping improves prognostic modeling in FOLFIRINOX-treated metastatic pancreatic cancer

BACKGROUND

First-line chemotherapy (5-fluorouracil, leucovorin, irinotecan, and oxaliplatin [FOLFIRINOX]) benefits few patients with metastatic pancreatic ductal adenocarcinoma (mPDAC). Prognostic markers for treatment-related survival are needed. This study validated the added benefit of whole-genome sequencing (WGS) to transcriptome-based classification in modeling FOLFIRINOX-related survival.

METHODS

Patients with mPDAC planning to start FOLFIRINOX were included in a prospective nationwide cohort. Pretreatment biopsies were submitted to WGS and RNA sequencing. Samples of non-FOLFIRINOX-treated patients were included for exploratory analyses.

FINDINGS

WGS was performed in biopsies from 108 FOLFIRINOX-treated patients and 51 non-FOLFIRINOX-treated patients. 12% of the tumors were KRAS wild type. These tumors had more targetable alterations (42% vs. 17%) and were associated with a longer median overall survival (mOS) than KRAS mutant tumors (7.8 months in KRAS mutant vs. 17.7 months in wild-type tumors, p = 0.0024). Transcriptome-based clustering revealed a tumor subgroup showing low classical and basal-like gene expression, enriched for KRAS wild-type status (p < 0.0001), a so-called "classifier-negative" subtype. The gene expression of these classifier-negative tumors correlated with neural-like signatures. For patients with a homologous recombination-deficient (HRD) tumor, mOS was not increased (8.0 months in homologous recombination-proficient [HRP] vs. 13.3 months in HRD tumors, p = 0.21).

CONCLUSIONS

KRAS wild-type tumors are a distinct PDAC subgroup with a better prognosis. Consequently, KRAS status assessment before transcriptome-based subtyping can stratify patients into three prognostic molecular subgroups (KRAS wild type, KRAS mutant classical, and KRAS mutant basal like). This integrative way of classification should be validated prior to incorporation in diagnostic practice.

FUNDING

ZonMw "Good Use of Medicine" program (848101012).

COVID-19 RT-qPCR-based screening in Austrian schools and incidences in the general population: a Bayesian spatiotemporal analysis

In 2021, the emergence of highly transmissible COVID-19 variants of concern increased susceptibility among younger populations. Despite this risk, face-to-face education remained essential for societal functioning and children's well-being, prompting the Austrian government to implement a nationwide screening program in educational institutions. This study explores the impact of this program on COVID-19 transmission by examining the relationship between incidence rates and factors such as age, vaccination coverage, and RT-qPCR positivity rates among school-aged children across Austrian districts, using a Bayesian spatiotemporal discrete model. Our findings highlight significant effects of vaccination and positivity rates on COVID-19 incidence, with variations in their influence across different age groups and locations. These results underscore the importance of monitoring these variables, particularly when active screening programs are in place.

Validation and Derivation of miRNA-Based Germline Signatures Predicting Radiation Toxicity in Prostate Cancer

PURPOSE

Although radiotherapy (RT) is one of the primary treatment modalities used in the treatment of cancer, patients often experience toxicity during or after treatment. RT-induced genitourinary (GU) toxicity is a significant survivorship challenge for patients with prostate cancer, but identifying those at risk has been challenging. Herein, we attempt (i) to validate a previously identified biomarker of late RT-induced GU toxicity, PROSTOX, consisting primarily of miRNA-based germline biomarkers (mirSNPs), and (ii) investigate the possibility of temporally and genetically defining other forms of RT-associated GU toxicity.

EXPERIMENTAL DESIGN

We included 148 patients enrolled in Magnetic Resonance Imaging-Guided Stereotactic Body Radiotherapy for Prostate Cancer (MIRAGE; NCT04384770), a trial comparing MRI- versus CT-guided prostate stereotactic body RT. Linear regression was used to evaluate the association between PROSTOX score and late GU grade toxicity. Machine learning approaches were used to develop predictive models for acute toxicity and chronic GU toxicity, and the accuracy of all models was assessed using AUC metrics. A comparative Gene Ontology analysis was performed.

RESULTS

PROSTOX accurately predicts late GU toxicity, achieving an AUC of 0.76, and demonstrates strong correlation with GU toxicity grade (p-1.2E-9). mirSNP-based signatures can distinguish acute RT-associated GU toxicity and chronic RT-associated GU toxicity (AUCs of 0.770 and 0.763, respectively). Finally, Gene Ontology analysis identifies unique pathways involved in each form of GU toxicity: acute, chronic, and late.

CONCLUSIONS

These findings provide strong evidence for the continued application of mirSNPs to predict toxicity to RT and act as a path for the continued personalization of RT with improved patient outcomes.

Deep plasma proteomics identifies and validates an eight-protein biomarker panel that separate benign from malignant tumors in ovarian cancer

BACKGROUND

Ovarian cancer has the highest mortality of all gynecological cancers and surgery is commonly used as final diagnostic. Available literature indicates that women with benign tumors could often be conservatively managed, but accurate molecular tests are needed for triaging when gold-standard imaging techniques are inconclusive or lacking.

METHODS

Here, we analyzed 5416 plasma proteins in two independent cohorts (N1 = 171, N2 = 233) with women surgically diagnosed with benign or malignant tumors. Using one cohort as discovery, we compared protein levels of benign tumors with early stage (I-II), late stage (III-IV) or any stage (I-IV) ovarian cancer and trained risk-score reporting multivariate models including a fixed cut-off for malignancy. Associations and model performance was then evaluated in the replication cohort.

RESULTS

We identify 327 biomarker associations, corresponding to 191 unique proteins, and replicate 326 (99.7%). By comparing the 191 proteins with their corresponding tumor gene expression we find that only 11% (21/191) have significant correlation. Through analyzes of protein-protein correlation networks, we find that 62 of the 191 proteins have high correlation with at least one other protein, suggesting that many of the associations are secondary effects. In the replication cohort, our model has areas under the curve (AUC = 0.96) corresponding to 97% sensitivity at 68% specificity. For early-stage tumors, we estimate the sensitivity to 91% at a specificity of 68% as compared to 85% and 54% for CA-125 alone.

CONCLUSIONS

Our results indicates that up to one third of benign cases can be identified by molecular measures thereby reducing the need for diagnostic surgery.

DBSCAN and DBCV application to open medical records heterogeneous data for identifying clinically significant clusters of patients with neuroblastoma

Neuroblastoma is a common pediatric cancer that affects thousands of infants worldwide, especially children under five years of age. Although recovery for patients with neuroblastoma is possible in 80% of cases, only 40% of those with high-risk stage four neuroblastoma survive. Electronic health records of patients with this disease contain valuable data on patients that can be analyzed using computational intelligence and statistical software by biomedical informatics researchers. Unsupervised machine learning methods, in particular, can identify clinically significant subgroups of patients, which can lead to new therapies or medical treatments for future patients belonging to the same subgroups. However, access to these datasets is often restricted, making it difficult to obtain them for independent research projects. In this study, we retrieved three open datasets containing data from patients diagnosed with neuroblastoma: the Genoa dataset and the Shanghai dataset from the Neuroblastoma Electronic Health Records Open Data Repository, and a dataset from the TARGET-NBL renowned program. We analyzed these datasets using several clustering techniques and measured the results with the DBCV (Density-Based Clustering Validation) index. Among these algorithms, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) was the only one that produced meaningful results. We scrutinized the two clusters of patients' profiles identified by DBSCAN in the three datasets and recognized several relevant clinical variables that clearly partitioned the patients into the two clusters that have clinical meaning in the neuroblastoma literature. Our results can have a significant impact on health informatics, because any computational analyst wishing to cluster small data of patients of a rare disease can choose to use DBSCAN and DBCV rather than utilizing more common methods such as k-Means and Silhouette coefficient.

Integrating transcriptomics and hybrid machine learning enables high-accuracy diagnostic modeling for nasopharyngeal carcinoma

BACKGROUND

Nasopharyngeal carcinoma (NPC) lacks biomarkers demonstrating both high specificity and sensitivity for early diagnosis. This study aimed to develop robust machine learning (ML)-driven diagnostic models and identify key biomarkers through integrated analysis of multi-cohort transcriptomic data.

METHODS

Seven NPC transcriptomic datasets (GSE12452, GSE40290, GSE53819, and GSE64634 were merged to form the training cohort, while GSE13597, GSE34573, and GSE61218 served as independent external validation sets) were integrated and preprocessed using ComBat for batch effect correction. Differential expression analysis identified 293 differentially expressed genes (DEGs). Twelve ML algorithms (including Stepglm, glmBoost, and RF) were systematically combined into 113 distinct models to classify NPC versus normal tissues. Top-performing models underwent external validation. Immune infiltration patterns and functional enrichment were analyzed using CIBERSORT and GSEA/GSVA, respectively.

RESULTS

The Stepglm[both]-RF hybrid model demonstrated exceptional performance with AUCs of 0.999 (training set; 95% CI: 0.997-1.000), 1.000 (GSE61218/GSE34573 validation), and 0.960 (GSE13597 validation). The glmBoost-RF model showed comparable efficacy, achieving AUCs of 1.000 (training), 0.950 (GSE61218), 1.000 (GSE34573), and 0.947 (GSE13597). Single-gene analysis identified RCN1 as a promising diagnostic marker (AUC = 0.953), with elevated expression levels correlating with poor prognosis in head and neck squamous cell carcinoma (HNSCC; p < 0.05). Immune profiling revealed significant enrichment of M1 macrophages and concomitant reduction of memory B cells in NPC. Functional enrichment analysis associated RCN1 with cell cycle regulation and immune-related pathways.

CONCLUSION

This study establishes two high-performance ML models (Stepglm[both]-RF and glmBoost-RF) with low variability for NPC diagnosis and identifies RCN1 as a dual-function biomarker with diagnostic and prognostic potential. The findings provide a scalable framework for early NPC detection and novel insights into immune microenvironment dysregulation.

Risk of gestational diabetes mellitus in relation to serum metal mixtures and mediating effects of metabolites in Chinese population: A dual approach using bibliometric and epidemiological analyses

Gestational diabetes mellitus (GDM) is a common complication of pregnancy, posing obstetrical and metabolic risks. While emerging evidence suggests an association between specific metal exposure and GDM, the crucial metabolic changes between serum metal exposure and GDM risk have yet to be fully clarified. Here we employed a combination of bibliometric and epidemiological methodologies to investigate the relationship between metal exposure and GDM and potential mediation role of metabolites in Chinese pregnant women. Our findings revealed an increasing research interest in the association of serum metal levels with GDM. Epidemiologically, metals such as iron (Fe) and molybdenum (Mo) were found to be negatively associated with GDM based on single-metal models with ORs being 0.325 (95 % CI: 0.138, 0.768) and 0.030 (95 % CI: 0.002, 0.439), respectively. A joint negative effect from metal co-exposure on GDM was shown, with Fe and Mo identified as the key metals. 7β-hydroxycholesterol mediated 9.9 % of the association between Fe and GDM risk, while melatonin mediated 7.8 % of the association between Mo and GDM risk. Our study provides new perspectives for understanding the beneficial effect of essential metals on GDM and lays a solid foundation for further validation through multicenter investigations and for exploration of mechanisms through in vivo studies.

Eyes wide open: Object-scene congruency and the pupillary response

The pupil response has long been considered a robust marker of cognitive load. In the context of semantic processing, research has demonstrated that the pupil dilates in response to stimuli which violate contextual expectations (e.g. events presented out of chronological order). However, the scope of this relationship has yet to be fully elucidated. For example, incongruent object-scene relationships, while comprehensively explored by eye-tracking and electrophysiology research, have yet to be investigated via pupillometry. In this study, we measured pupil size in response to an object-scene congruency task. Participants were presented with a photorealistic background scene and instructed to fixate their gaze on a cued point within the scene. Upon recovery of pupil size to baseline, a congruent object (i.e. an object which fit into the overall meaning of the scene) or an incongruent object appeared at the cued fixation point for the remainder of each trial. We hypothesized that incongruent objects would result in greater mean pupil dilation from baseline than congruent objects, due to the increase in cognitive effort required for semantic processing of incongruent objects within a scene. Yet, in opposition to our hypothesis, the results of a time course analysis revealed that pupil size was significantly greater for the congruent condition than the incongruent condition. The resulting implications for understanding pupil dilation as a physiological marker, both independently and in comparison to other markers, for high-level cognitive processes such as semantic integration are discussed.

Earlywood vessel characteristics are early indicators of drought-induced decline in ring-porous oak species within the Mediterranean Basin

Heat and drought stress have triggered forest dieback episodes worldwide, affecting oak forests, particularly in hotspots of climate change such as the Mediterranean Basin. However, forecasting dieback is not straightforward. In this study, we used the earlywood anatomy to improve dieback forecasts in five oak species characterized by different drought sensitivity (i.e. from high to low Quercus robur, Q. cerris, Q. frainetto and Q. canariensis, Q. humilis, Q. pubescens) across Italy and Spain. We measured radial growth, expressed as basal area increment (BAI), earlywood hydraulic diameter (Dh) and vessel area of coexisting non-declining (ND) and declining (D) trees in each stand. Then, we calculated the product between the coefficient of variation (CV) of vessel area and a spatial aggregation index (AI). High CV × AI values indicate regularly spaced vessels with variable area of vessels, while low values correspond to clustered vessels with similar area. ND trees showed higher BAI values than D trees from 10 to 40 years before the dieback onset, when ND trees grew 20-50 % more than the D trees. We observed a decline in the vessel area CV several decades prior to dieback in D trees, with the exception of Q. cerris. The AI showed higher values in ND than in D trees. Consequently, the CV × AI product was consistently higher in ND than in D trees. The CV × AI divergence between ND and D trees was pronounced in the wettest sites, specifically for Q. robur and Q. humilis. Time series of CV × AI effectively differentiated trees based on their vigor. Wood anatomy variables could be used to enhance predictions of vulnerability to drought-induced dieback. This study can help identify vulnerable trees before the onset of dieback symptoms, serving as a tool to support the management of forests prone to drought.

Intravenous Ferric Carboxymaltose in Heart Failure With Iron Deficiency: The FAIR-HF2 DZHK05 Randomized Clinical Trial

Importance

Uncertainty remains about the efficacy of intravenous iron in patients with heart failure and iron deficiency.

Objective

To assess the efficacy and safety of ferric carboxymaltose in patients with heart failure and iron deficiency.

Design, Setting, and Participants

This multicenter, randomized clinical trial enrolled 1105 patients with heart failure (defined as having a left ventricular ejection fraction of ≤45%) and iron deficiency (serum ferritin level <100 ng/mL; or if transferrin saturation was <20%, a serum ferritin level between 100 ng/mL and 299 ng/mL) at 70 clinic sites in 6 European countries from March 2017 to November 2023. The median follow-up was 16.6 months (IQR, 7.9-29.9 months).

Intervention

Administration of ferric carboxymaltose (n = 558) initially given at an intravenous dose of up to 2000 mg that was followed by 500 mg every 4 months (unless stopping criteria were met) vs a saline placebo (n = 547).

Main Outcomes and Measures

The primary end point events were (1) time to cardiovascular death or first heart failure hospitalization, (2) total heart failure hospitalizations, and (3) time to cardiovascular death or first heart failure hospitalization in patients with a transferrin saturation less than 20%. All end point events were measured through follow-up. The end points would be considered statistically significant if they fulfilled at least 1 of the following conditions: (1) P ≤ .05 for all 3 of the end point comparisons, (2) P ≤ .025 for 2 of the end point comparisons, or (3) P ≤ .0167 for any of the 3 end point comparisons (Hochberg procedure).

Results

Of the 1105 participants (mean age, 70 years [SD, 12 years]; 33% were women), cardiovascular death or first heart failure hospitalization (first primary outcome) occurred in 141 in the ferric carboxymaltose group vs 166 in the placebo group (hazard ratio, 0.79 [95% CI, 0.63-0.99]; P = .04). The second primary outcome (total heart failure hospitalizations) occurred 264 times in the ferric carboxymaltose group vs 320 times in the placebo group (rate ratio, 0.80 [95% CI, 0.60-1.06]; P = .12). The third primary outcome (cardiovascular death or first heart failure hospitalization in patients with a transferrin saturation <20%) occurred in 103 patients in the ferric carboxymaltose group vs 128 patients in the placebo group (hazard ratio, 0.79 [95% CI, 0.61-1.02], P = .07). A similar amount of patients had at least 1 serious adverse event in the ferric carboxymaltose group (269; 48.2%) vs in the placebo group (273; 49.9%) (P = .61).

Conclusions and Relevance

In patients with heart failure and iron deficiency, ferric carboxymaltose did not significantly reduce the time to first heart failure hospitalization or cardiovascular death in the overall cohort or in patients with a transferrin saturation less than 20%, or reduce the total number of heart failure hospitalizations vs placebo.

Trial Registration

ClinicalTrials.gov Identifier: NCT03036462.

Comparison of total event analysis and first event analysis in relation to heterogeneity in cardiovascular trials

BACKGROUND

In cardiovascular (CV) trials, analyzing the total number of events, rather than just time-to-first event, enhances understanding of participants' health. Adapting Cox models to account for between-subject heterogeneity in multiple events and understanding its impact plays crucial roles in total event analysis.

METHOD

This study compares effect sizes from first event and total event analyses in three cardiovascular trials: ORIGIN (N = 12,537, median follow-up of 6.2 years), COMPASS (N = 18,278, median follow-up of 1.8 years), TRANSCEND (N = 5,926, median follow-up of 1.1 years). It also examines the impact of heterogeneity, measured by the negative binomial overdispersion parameter. Treatment effects were assessed using the Cox model for first events and the negative binomial (NB), Andersen-Gill (AG), Prentice-Williams-Peterson (PWP), Wei-Lin-Weissfeld (WLW), and Lin-Wei-Yang-Ying (LWYY) models for total events. Hazard ratios (HRs) or risk ratios (RRs), 95% confidence intervals (CIs), and CI widths were reported. The risk ratio applies to negative binomial. The first composite was consisted of myocardial infarction (MI), stroke, cardiovascular death. Simulations assessed Type I error, power, and mean squared error across the different approaches.

RESULTS

In ORIGIN, the incidence per 100 years increased from 2.9 to 3.8 for the first composite with a heterogeneity of 2.4. The HR or RR for the first composite was 1.03 (95% CI, 0.94-1.12, CI width = 0.18) using Cox, 1.01 (95% CI, 0.92-1.11, CI width = 0.19) for NB, 1.01 (95% CI, 0.94-1.09, CI width = 0.15) for AG, 1.02 (95% CI, 0.94-1.10, CI width = 0.16) for PWP total, 1.01 (95% CI, 0.94-1.09, CI width = 0.15) for PWP gap, 1.03 (95% CI, 0.94-1.12, CI width = 0.18) for WLW and 1.01 (95% CI, 0.92-1.11, CI width = 0.19) for LWYY. Similar trends were observed in other studies. Our simulation results showed that total event approaches had approximately 5% higher power than the Cox model, though power declined exponentially across all methods with increasing heterogeneity. Among the total event methods, AG, PWP gap, and LWYY demonstrated better power, with AG and LWYY also achieving the smallest mean squared error (MSE).

CONCLUSIONS

High heterogeneity arises when a small number of patients experience a disproportionately large number of events. This effect is more pronounced when the overall event incidence is low and few patients experience any events. The effect size and CI width stayed consistent with low heterogeneity across different approaches. Power decreased with high heterogeneity. The AG and LWYY approaches slightly outperformed the other approaches.

CLINICAL TRIAL REGISTRATION

ORIGIN (NCT00069784), COMPASS (NCT01776424), TRANSCEND (NCT00153101).

The spatial layout of antagonistic brain regions is explicable based on geometric principles

Brain activity emerges in a dynamic landscape of regional increases and decreases that span the cortex. Increases in activity during a cognitive task are often assumed to reflect the processing of task-relevant information, while reductions can be interpreted as suppression of irrelevant activity to facilitate task goals. Here, we explore the relationship between task-induced increases and decreases in activity from a geometric perspective. Using a technique known as kriging, developed in earth sciences, we examined whether the spatial organisation of brain regions showing positive activity could be predicted based on the spatial layout of regions showing activity decreases (and vice versa). Consistent with this hypothesis we established the spatial distribution of regions showing reductions in activity could predict (i) regions showing task-relevant increases in activity in both groups of humans and single individuals; (ii) patterns of neural activity captured by calcium imaging in mice; and, (iii) showed a high degree of generalisability across task contexts. Our analysis, therefore, establishes that antagonistic relationships between brain regions are topographically determined, a spatial analog for the well documented anti-correlation between brain systems over time.

Aging and diet alter the protein ubiquitylation landscape in the mouse brain

Post-translational modifications (PTMs) regulate protein homeostasis, but how aging impacts PTMs remains unclear. Here, we used mass spectrometry to reveal changes in hundreds of protein ubiquitylation, acetylation, and phosphorylation sites in the mouse aging brain. We show that aging has a major impact on protein ubiquitylation. 29% of the quantified ubiquitylation sites were affected independently of protein abundance, indicating altered PTM stoichiometry. Using iPSC-derived neurons, we estimated that 35% of ubiquitylation changes observed in the aged brain can be attributed to reduced proteasome activity. Finally, we tested whether protein ubiquitylation in the brain can be influenced by dietary intervention. We found that one cycle of dietary restriction and re-feeding modifies the brain ubiquitylome, rescuing some but exacerbating other ubiquitylation changes observed in old brains. Our findings reveal an age-dependent ubiquitylation signature modifiable by dietary intervention, providing insights into mechanisms of protein homeostasis impairment and highlighting potential biomarkers of brain aging.

Unlocking sea turtle diving behaviour from low-temporal resolution time-depth recorders

Biologging is a rapidly advancing field providing information on previously unexplored aspects of animal ecology, including the vertical movement dimension. Understanding vertical behaviour through the use of time-depth recorders (TDRs) in marine vertebrates is critical to aid conservation and management decisions. However, using TDRs can be particularly problematic to infer animal behaviour from elusive animals, when tags are difficult to recover and collected data is satellite-relayed at lower temporal frequencies. Here, we present a novel method to process low-resolution TDR data at 5-minute intervals and infer diving behaviour from loggerhead turtles (Caretta caretta) during their elusive pelagic life stage spanning extended periods (> 250 days). Using a Hidden Markov Model (HMM) we identify four behavioural states, associated with resting, foraging, shallow exploration, and deep exploration. Three of the four behavioural states were found to have strong seasonal patterns, corroborating with known sea-turtle biology. The results presented provide a novel way of interpreting low-resolution TDR data and provide a unique insight into sea turtle ecology.

Bayesian Phylodynamic Inference of Multitype Population Trajectories Using Genomic Data

Phylodynamic methods provide a coherent framework for the inference of population parameters directly from genetic data. They are an important tool for understanding both the spread of epidemics as well as long-term macroevolutionary trends in speciation and extinction. In particular, phylodynamic methods based on multitype birth-death models have been used to infer the evolution of discrete traits, the movement of individuals or pathogens between geographic locations or host types, and the transition of infected individuals between disease stages. In these models, population heterogeneity is treated by assigning individuals to different discrete types. Typically, methods which allow inference of parameters under multitype birth-death models integrate over the possible birth-death trajectories (i.e. the type-specific population size functions) to reduce the computational demands of the inference. As a result, it has not been possible to use these methods to directly infer the dynamics of trait-specific population sizes, infected host counts or other such demographic quantities. In this article, we present a method which infers these multitype trajectories with minimal additional computational cost beyond that of existing methods. We demonstrate the practicality of our approach by applying it to a previously published set of Middle East respiratory syndrome coronavirus genomes, inferring the numbers of human and camel cases through time, together with the number and timing of spillovers from the camel reservoir. This application highlights the multitype population trajectory's ability to elucidate properties of the population which are not directly ancestral to its sampled members.

Increased inflammation as well as decreased endoplasmic reticulum stress and translation differentiate pancreatic islets from donors with pre-symptomatic stage 1 type 1 diabetes and non-diabetic donors

AIMS/HYPOTHESIS

Progression to type 1 diabetes is associated with genetic factors, the presence of autoantibodies and a decline in beta cell insulin secretion in response to glucose. Very little is known regarding the molecular changes that occur in human insulin-secreting beta cells prior to the onset of type 1 diabetes. Herein, we applied an unbiased proteomics approach to identify changes in proteins and potential mechanisms of islet dysfunction in islet-autoantibody-positive organ donors with pre-symptomatic stage 1 type 1 diabetes (HbA1c ≤42 mmol/mol [6.0%]). We aimed to identify pathways in islets that are indicative of beta cell dysfunction.

METHODS

Multiple islet sections were collected through laser microdissection of frozen pancreatic tissues from organ donors positive for single or multiple islet autoantibodies (AAb+, n=5), and age (±2 years)- and sex-matched non-diabetic (ND) control donors ( n=5) obtained from the Network for Pancreatic Organ donors with Diabetes (nPOD). Islet sections were subjected to MS-based proteomics and analysed with label-free quantification followed by pathway and functional annotations.

RESULTS

Analyses resulted in ~4500 proteins identified with low false discovery rate (<1%), with 2165 proteins reliably quantified in every islet sample. We observed large inter-donor variations that presented a challenge for statistical analysis of proteome changes between donor groups. We therefore focused on only the donors with stage 1 type 1 diabetes who were positive for multiple autoantibodies (mAAb+, n=3) and genetic risk compared with their matched ND controls (n=3) for the final statistical analysis. Approximately 10% of the proteins (n=202) were significantly different (unadjusted p<0.025, q<0.15) for mAAb+ vs ND donor islets. The significant alterations clustered around major functions for upregulation in the immune response and glycolysis, and downregulation in endoplasmic reticulum (ER) stress response as well as protein translation and synthesis. The observed proteome changes were further supported by several independent published datasets, including a proteomics dataset from in vitro proinflammatory cytokine-treated human islets and single-cell RNA-seq datasets from AAb+ individuals.

CONCLUSIONS/INTERPRETATION

In situ human islet proteome alterations in stage 1 type 1 diabetes centred around several major functional categories, including an expected increase in immune response genes (elevated antigen presentation/HLA), with decreases in protein synthesis and ER stress response, as well as compensatory metabolic response. The dataset serves as a proteomics resource for future studies on beta cell changes during type 1 diabetes progression and pathogenesis.

DATA AVAILABILITY

The LC-MS raw datasets that support the findings of this study have been deposited in the online repository: MassIVE ( https://massive.ucsd.edu/ProteoSAFe/static/massive.jsp ) with accession no. MSV000090212.

Estimating the diagnostic performance of serological assays for emerging pathogens using a Bayesian approach: Myxoma virus in the Iberian hare (Lepus granatensis)

Validated diagnostic tools are essential when conducting serological surveys. However, reliable tests are scarce and hard to attain for emerging pathogens due to the lack of reference tests or samples. Recently, a recombinant myxoma virus (MYXV), named ha-MYXV, raised alarm in the Iberian Peninsula for its impact on Iberian hare (Lepus granatensis) populations and its detection in wild (Oryctolagus cuniculus) and domestic rabbits. Here, we follow a Bayesian approach to evaluate two serological tools, an indirect ELISA (iELISA) and a competitive ELISA (cELISA), used to monitor this emerging pathogen in Iberian hare populations. We modelled serological data from 227 hares conveniently selected retrospectively for their apparent healthy status. First, we applied finite mixture models to adjust the cut-off thresholds of both tests, which improved the agreement between both tests (initial kappa = 0.42, after threshold adjustment = 0.78). Then, we employed Bayesian latent class models (BLCM) to estimate the assays' specificity (Sp) and sensitivity (Se). The BLCM estimated median Sp of 94.0 % (95 % posterior probability interval (PPI): 85.9-99.4) and 96.1 % (PPI: 87.2-100.0), and Se of 77.7 % (PPI: 61.5-89.5) and 91.7 % (PPI: 78.1-99.9), for the iELISA and the cELISA, respectively. The true seroprevalence estimations show higher values in south-central Spain (ranging from 13.1 % to 70.4 %) and lower in the north (Navarra: 5.5 %). A Bayesian approach allowed to evaluate diagnostic tools for ha-MYXV, an emerging wildlife pathogen, in the absence of reference tests or samples. Future epidemiological studies of myxomatosis in Iberian hares should calculate true seroprevalence based on our estimations.

Modeling growth curve parameters in Peruvian llamas using a Bayesian approach

The objective of this study was to fit four nonlinear models (Brody, von Bertalanffy, Gompertz and Logistic) to realizations of llama weight, using frequentist and Bayesian approaches. Animals from both sexes and types (K'ara and Ch'accu) were observed. Data consisted of 43,332 monthly body weight records, taken from birth to 12 months of age from 3611 llamas, collected from 1998 to 2017 in the Quimsachata Experimental Station of the Instituto Nacional de Innovación Agraria (INIA) in Peru. Parameters for Non-linear models for growth curves were estimated by frequentist and Bayesian procedures. The MCMC method using the Metropolis-Hastings algorithm with noninformative prior distributions was applied in the Bayesian approach. All non-linear functions closely fitted actual body weight measurements, while the Brody function provided the best fit in both frequentist and Bayesian approaches in describing the growth data of llamas. The analysis revealed that female llamas reached higher asymptotic weights than males, and K'ara-type llamas exhibited higher asymptotic weights compared to Ch'accu-type animals. The asymptotic body weight, estimated for all data using the Brody model, was 42 kg at 12 months of age in llamas from Peru. The results of this research highlight the potential of applying nonlinear functions to model the weight-age relationship in llamas using a Bayesian approach. However, limitations include the use of historical data, which may not fully represent current growth patterns, and the reliance on non-informative priors, which could be improved with prior knowledge. Future studies should refine these aspects.

Predicting Placenta Accreta Spectrum Disorder Through Machine Learning Using Metabolomic and Lipidomic Profiling and Clinical Characteristics

OBJECTIVE

To perform metabolomic and lipidomic profiling with plasma samples from patients with placenta accreta spectrum (PAS) to identify possible biomarkers for PAS and to predict PAS with machine learning methods that incorporated clinical characteristics with metabolomic and lipidomic profiles.

METHODS

This was a multicenter case-control study of patients with placenta previa with PAS (case group n=33) and previa alone (control group n=21). Maternal third-trimester plasma samples were collected and stored at -80°C. Untargeted metabolomic and targeted lipidomic assays were measured with flow-injection mass spectrometry. Univariate analysis provided an association of each lipid or metabolite with the outcome. The Benjamini-Hochberg procedure was used to control for the false discovery rate. Elastic net machine learning models were trained on patient characteristics to predict risk, and an integrated elastic net model of lipidome or metabolome with nine clinical features was trained. Performance using the area under the receiver operating characteristic curve (AUC) was determined with Monte Carlo cross-validation. Statistical significance was defined at P<.05.

RESULTS

The mean gestational age at sample collection was 33 3/7 weeks (case group) and 35 5/7 weeks (control group) (P<.01). In total, 786 lipid species and 2,605 metabolite features were evaluated. Univariate analysis revealed 31 lipids and 214 metabolites associated with the outcome (P<.05). After false discovery rate adjustment, these associations no longer remained statistically significant. When the machine learning model was applied, prediction of PAS with only clinical characteristics (AUC 0.685, 95% CI, 0.65-0.72) performed similarly to prediction with the lipidome model (AUC 0.699, 95% CI, 0.60-0.80) and the metabolome model (AUC 0.71, 95% CI, 0.66-0.76). However, integration of metabolome and lipidome with clinical features did not improve the model.

CONCLUSION

Metabolomic and lipidomic profiling performed similarly to, and not better than, clinical risk factors using machine learning to predict PAS among patients with PAS with previa and previa alone.