Bayesian Logistic Regression Model Research Articles

Preoperative inspiratory muscle weakness as a risk factor of postoperative pulmonary complications in patients with esophageal cancer.

We examined whether preoperative inspiratory muscle weakness (IMW) is a risk factor for postoperative pulmonary complications (PPCs) in patients with esophageal cancer who underwent subtotal esophagectomy. This single-center retrospective cohort study enrolled patients with esophageal cancer who underwent a scheduled subtotal esophagectomy between June 2020 and May 2022. Maximal inspiratory pressure (MIP) was measured as inspiratory muscle strength using a respiratory dynamometer, and we defined IMW as MIP < 80% of the predicted value. Our primary outcome comprised overall PPCs. We investigated the relationship between IMW and PPCs using the Bayesian logistic regression model. After exclusion, 72 patients were included in this study. IMW was identified in 26 patients (36%), and PPCs developed in 28 patients (39%). Among patients with IMW, 15 (58%) developed PPCs. Preoperative IMW was associated with PPCs (mean odds ratio [OR]: 3.58; 95% credible interval [95% CrI]: 1.29, 9.73) in the unweighted model. A similar association was observed in the weighted model adjusted for preoperative and intraoperative contributing factors (mean OR: 4.15; 95% CrI: 2.04, 8.45). Preoperative IMW was associated with PPCs in patients with esophageal cancer who underwent subtotal esophagectomy. This association remained after adjusting for preoperative and intraoperative contributing factors.

Early and late infarct growth rate in ischemic stroke patients after successful endovascular treatment in early time window: correlation of imaging and clinical factors with clinical outcome.

The prospective study assessed infarct growth rate (IGR) in acute ischemic stroke (AIS) with large vessel occlusion (LVO) after recanalization in early time window. Early IGR (EIGR) and late IGR (LIGR) were correlated with imaging and clinical data; we searched for outcome predictors. We included 71 consecutive patients. Subjects underwent computed tomography perfusion (CTP) for ischemic core volume assessment at 99.0 minutes (median) from stroke onset, recanalization was performed at 78.0 minutes (median) from CTP. Final infarct volume (FIV) was measured on 24±2 hours imaging follow-up. EIGR was calculated as the core volume/time between stroke onset and CTP; LIGR was calculated as FIV/time between CTP and imaging follow-up. Twenty-two subjects were assessed as poor outcome, 49 as good outcome. Group differences were tested by Mann-Whitney test and χ2 test. Bayesian logistic regression models were used to predict clinical outcome, Pearson correlations for the log-transformed predictors. Subjects with poor outcome were older, median age 78.0 [interquartile range (IQR): 71.8, 83.8] versus 68.0 (IQR: 57.0, 73.0) years; 95% confidence interval (CI): 6.00 to 16.00; P<0.001. Their stroke severity scale was higher, median 19.0 (IQR: 16.0, 20.0) versus 15.5 (IQR: 10.8, 18.0); 95% CI: 1.00 to 6.00; P<0.001. They had higher EIGR, median 23.9 (IQR: 6.4, 104.0) versus 6.7 (IQR: 1.7, 13.0) mL/h; 95% CI: 3.26 to 53.68; P=0.002; and larger core, median 52.5 (IQR: 13.1, 148.5) versus 10.0 (IQR: 1.4, 20.0) mL; 95% CI: 11.00 to 81.00; P<0.001. In subjects with poor outcome, infarct growth continued after thrombectomy with LIGR 2.0 (IQR: 1.2, 9.7) versus 0.3 (IQR: 0.0, 0.7) mL/h; 95% CI: 1.10 to 6.10; P<0.001; resulting in larger FIV, median 186.5 (IQR: 49.3, 280.8) versus 18.5 (IQR: 8.0, 34.0) mL; 95% CI: 55.30 to 214.00; P<0.001. Strong correlations among predictors were found e.g., core and EIGR (r=0.942), LIGR and FIV (r=0.779), core and FIV (r=0.761). Clinical outcome was best predicted using data from later measurements as FIV and LIGR. Data from later measurements were more predictive, there was no major benefit to use growth over volume data.

Handling Ill-conditioned Omics Data with Deep Probabilistic Models.

The advent of high-throughput technologies has produced an increase in the dimensionality of omics datasets, which limits the application of machine learning methods due to the great unbalance between the number of observations and features. In this scenario, dimensionality reduction is essential to extract the relevant information within these datasets and project it in a low-dimensional space, and probabilistic latent space models are becoming popular given their capability to capture the underlying structure of the data as well as the uncertainty in the information. This article aims to provide a general classification and dimensionality reduction method based on deep latent space models that tackles two of the main problems that arise in omics datasets: the presence of missing data and the limited number of observations against the number of features. We propose a semi-supervised Bayesian latent space model that infers a low-dimensional embedding driven by the target label: the Deep Bayesian Logistic Regression (DBLR) model. During inference, the model also learns a global vector of weights that allows it to make predictions given the low-dimensional embedding of the observations. Since this kind of dataset is prone to overfitting, we introduce an additional probabilistic regularization method based on the semi-supervised nature of the model. We compared the performance of the DBLR against several state-of-the-art methods for dimensionality reduction, both in synthetic and real datasets with different data types. The proposed model provides more informative low-dimensional representations, outperforms the baseline methods in classification, and can naturally handle missing entries.

Paying attention to cardiac surgical risk: An interpretable machine learning approach using an uncertainty-aware attentive neural network.

Machine learning (ML) is increasingly applied to predict adverse postoperative outcomes in cardiac surgery. Commonly used ML models fail to translate to clinical practice due to absent model explainability, limited uncertainty quantification, and no flexibility to missing data. We aimed to develop and benchmark a novel ML approach, the uncertainty-aware attention network (UAN), to overcome these common limitations. Two Bayesian uncertainty quantification methods were tested, generalized variational inference (GVI) or a posterior network (PN). The UAN models were compared with an ensemble of XGBoost models and a Bayesian logistic regression model (LR) with imputation. The derivation datasets consisted of 153,932 surgery events from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS) Cardiac Surgery Database. An external validation consisted of 7343 surgery events which were extracted from the Medical Information Mart for Intensive Care (MIMIC) III critical care dataset. The highest performing model on the external validation dataset was a UAN-GVI with an area under the receiver operating characteristic curve (AUC) of 0.78 (0.01). Model performance improved on high confidence samples with an AUC of 0.81 (0.01). Confidence calibration for aleatoric uncertainty was excellent for all models. Calibration for epistemic uncertainty was more variable, with an ensemble of XGBoost models performing the best with an AUC of 0.84 (0.08). Epistemic uncertainty was improved using the PN approach, compared to GVI. UAN is able to use an interpretable and flexible deep learning approach to provide estimates of model uncertainty alongside state-of-the-art predictions. The model has been made freely available as an easy-to-use web application demonstrating that by designing uncertainty-aware models with innately explainable predictions deep learning may become more suitable for routine clinical use.

Integrating the rights-based and capability approaches in the analysis of maternal healthcare utilization in sub-Saharan Africa: A multilevel modelling study.

The rights-based and capability approaches received increased attention relative to maternal health in the aftermath of the 2015 Millennium Development Goals (MDGs). This may be in view of the sub-optimal progress gained in reducing maternal and child mortality, especially in developing countries. Despite the combined potential of these approaches, there are limited empirical studies testing their viability in aiding our understanding of maternal healthcare utilization in developing countries. This is what this study sought to accomplish. We combined several datasets, including the Demographic Health Surveys (DHS), World Development Indicators, the World Governance Indicators and Freedom House. Bayesian multilevel logistic regression models were applied on three indicators of maternal healthcare utilization (antenatal care visits, institutional delivery, and postnatal check-ups) in relation to selected variables representing right-based and capability approaches. After controlling for relevant individual and community-level factors, the results show that living in countries with high freedom status (POR = 1.19) and higher female secondary school enrolments (POR = 1.54) increases the odds of adequate antenatal care. Residence in countries with high freedom status (POR = 1.33) and higher voice and accountability (POR = 1.72) has a positive influence on institutional delivery. Similar results are reported for postnatal care where country freedom status (POR = 1.89), voice and accountability (POR = 1.25) and female school enrolment (POR = 1.41) are significant predictors. The results imply that the rights-based and capability approaches have the potential to enhance maternal healthcare utilization in sub-Saharan Africa. Therefore, policy strategies emphasizing on freedoms, accountability, and individual capability functionings should be encouraged in the pursuit of partly achieving Sustainable Development Goals (SDG) number 3.

Evaluating the likelihood of slope failure in ageing earthworks using Bayesian updating

Earthwork assets, including cut slopes and embankments, are essential components of the infrastructure supporting road and rail transportation networks. Asset owners must assess the stability of these slopes as they deteriorate, to prevent unwanted slope failures. Assessing the stability of individual earthworks within a portfolio using slope stability analyses can be expensive and time consuming. Hence, a Bayesian logistic regression model was developed to evaluate the probability of slope failure, using training data from published case histories of slope failures. The Bayesian model was then used to assess the probability of failure for the more specific case of clay cut slopes within a railway earthwork asset portfolio owned by Network Rail (NR) in the UK. The portfolio includes earthworks at various stages of degraded strength and with different drainage conditions. The results from models with material properties that were equivalent to those for the deteriorated strength of clays compared most closely with clay cut slope failures within the NR data set. Steeper slopes (&gt;35°) had the highest probability of failure, regardless of slope height, and drainage condition. However, for shallower slopes, poorly drained slopes had a ≈20% higher probability of failure than well-drained slopes.

Small samples-oriented intrinsically explainable machine learning using Variational Bayesian Logistic Regression: An intensive care unit readmission prediction case for liver transplantation patients

Recently, machine learning has revolutionized medical diagnosis and prediction, rivaling human experts in accuracy. However, the limited interpretability poses challenges in meeting stringent medical decision standards. To facilitate interpretable predictive inference, this article proposes a small samples-oriented intrinsically explainable machine learning model, an Optimal Variational Bayesian Logistic Regression (OVBLR) model with a salient feature estimation strategy (OVBLR-SFE). OVBLR-SFE sequentially learns from limited labeled samples, mining disease-related risk factors to provide intrinsic interpretability to its predictive inference automatically. The model achieves this by reasonably approximating the posterior probability through the proposed OVBLR model to estimate the feature weight and influential effect on model outputs to facilitate salient feature selection. The effectiveness and superiority of OVBLR-SFE are demonstrated through extensive experiments on four benchmark medical datasets from the UCI repository, including the Bone Marrow Transplant: Children Data Set, Breast Cancer Wisconsin (Original) Data Set, SPECT Heart Data Set, and Heart Disease Data Set, as well as a real-world case of intensive care unit readmission prediction. Experimental results reveal that the OVBLR-SFE can achieve high accuracies of 92.02%, 97.21%, 86.71%, and 84.47%, respectively, with an average classification accuracy of 90.10% stably on the four UCI datasets. Especially, OVBLR-SFE can accurately infer significant features that align with that obtained by the post-hoc interpretation approaches: SHAP and AcME, demonstrating remarkable interpretability. In the real application, the proposed method achieved a readmission prediction accuracy of 88.11% in predicting the readmission probabilities of liver transplantation patients. Notably, the inferred salient features, such as total Intensive Care Unit length of stay (ICUTime), N-terminal pro-brain natriuretic peptide(NTproBNP), Operating time of surgery(OperationTime_h), and intraoperative red blood cell transfusion (RBCT), emerges as crucial in readmission prediction, aligning with the assessments of clinical experts. The source code of this study can be accessed at https://github.com/xqw42/OVBLR-SFE.

EVCA Classifier: A MCMC-Based Classifier for Analyzing High-Dimensional Big Data

In this work, we introduce an innovative Markov Chain Monte Carlo (MCMC) classifier, a synergistic combination of Bayesian machine learning and Apache Spark, highlighting the novel use of this methodology in the spectrum of big data management and environmental analysis. By employing a large dataset of air pollutant concentrations in Madrid from 2001 to 2018, we developed a Bayesian Logistic Regression model, capable of accurately classifying the Air Quality Index (AQI) as safe or hazardous. This mathematical formulation adeptly synthesizes prior beliefs and observed data into robust posterior distributions, enabling superior management of overfitting, enhancing the predictive accuracy, and demonstrating a scalable approach for large-scale data processing. Notably, the proposed model achieved a maximum accuracy of 87.91% and an exceptional recall value of 99.58% at a decision threshold of 0.505, reflecting its proficiency in accurately identifying true negatives and mitigating misclassification, even though it slightly underperformed in comparison to the traditional Frequentist Logistic Regression in terms of accuracy and the AUC score. Ultimately, this research underscores the efficacy of Bayesian machine learning for big data management and environmental analysis, while signifying the pivotal role of the first-ever MCMC Classifier and Apache Spark in dealing with the challenges posed by large datasets and high-dimensional data with broader implications not only in sectors such as statistics, mathematics, physics but also in practical, real-world applications.

Thirty years of Mexico’ state elections. What explains subnational alternations?

ABSTRACT Drawing on classic literature on democracy transitions and subnational authoritarian regimes, this article explores some causes of alternations at the subnational level, an understudied topic in the literature. The study uses data on electoral alternations at the state level in Mexico over 30 years. A Bayesian logistic regression model is employed to identify conditions that are most likely to lead to gubernatorial alternations. Elite divisions and coalition building are the main drivers of these changes, suggesting that subnational political context is an important factor in state-level alternations, even after controlling for various economic and social indicators.

Evaluation of the performance of the IFN-γ release assay in bovine tuberculosis free herds from five European countries

The diagnostic methods for granting and maintenance of the official tuberculosis-free (OTF) status and for intra-Community movement of cattle are the tuberculin skin tests (single or comparative) and the interferon-γ (IFN-γ) release assay (IGRA). However, until now, IGRAs have been primarily applied in infected farms in parallel to the skin test to maximize the number of infected animals detected. Therefore, an evaluation of the performance of IGRAs in OTF herds to assess whether if their specificity is equal to or higher than that of the skin tests is needed. For this, a panel of 4365 plasma samples coming from 84 OTF herds in six European regions (five countries) was assembled and analysed using two IGRA kits, the ID Screen® Ruminant IFN-g (IDvet) and the Bovigam™ TB Kit (Bovigam). Results were evaluated using different cut-offs, and the impact of herd and animal-level factors on the probability of positivity was assessed using hierarchical Bayesian multivariable logistic regression models. The percentage of reactors ranged from 1.7 to 21.0% (IDvet: S/P ≥ 35%), and 2.1–26.3% (Bovigam: ODbovis–ODPBS ≥ 0.1 and ODbovis–ODavium ≥ 0.1) depending on the region, with Bovigam disclosing more reactors in all regions. The results suggest that specificity of IGRAs can be influenced by the production type, age and region of origin of the animals. Changes in the cut-offs could lead to specificity values above 98–99% in certain OTF populations, but no single cut-off yielding a sufficiently high specificity (equal or higher than that of skin tests) in all populations was identified. Therefore, an exploratory analysis of the baseline IFN-γ reactivity in OTF populations could help to assess the usefulness of this technique when applied for the purpose of maintaining OTF status.

Yard Age, Cultivar Susceptibility, and Spring Pruning Practices as Risk Factors for Overwintering of Podosphaera macularis on Hop

Hop powdery mildew, caused by Podosphaera macularis, may be initiated by sexual or asexual spores of the pathogen, but only asexual reproduction has been reported in the Pacific Northwest United States. In this region, overwintering is via bud perennation, leading to "flag shoots" the following season. We expanded on previous research and analyzed 766 location-years of flag shoot data from Oregon hop yards collected from 2014 to 2020, examining potential risk factors of yard age, cultivar susceptibility to powdery mildew, pruning method (mechanical means or not), and pruning thoroughness. Bayesian logistic regression models were fitted to estimate the probability of flag shoot presence or absence. We found little evidence to support yard age as a significant risk factor for flag shoot presence, as the 95% credible interval for the parameter associated with yard age was consistently centered around 0 in each of our models. However, cultivar susceptibility to powdery mildew, spring pruning method, and spring pruning thoroughness were associated with increases in the probability of observing a flag shoot. Although yard age was not an important risk factor, there was an interrelationship between yard age, pruning method, and pruning thoroughness that suggests second-year yards may indirectly be at an elevated risk for flag shoot development because these yards are less often pruned using mechanical methods. Our research further identifies yards that are most at risk for overwintering of P. macularis and reinforces the importance of thorough pruning in yards with an elevated risk for possessing a flag shoot. [Formula: see text] The author(s) have dedicated the work to the public domain under the Creative Commons CC0 “No Rights Reserved” license by waiving all of his or her rights to the work worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law, 2023.

Association of glycemic gap with stroke recurrence in patients with ischemic stroke.

Glycemic gap, as a novel index of acute glycemic excursion, is associated with poor prognosis of different diseases. This study aimed to explore the association of the glycemic gap with long-term stroke recurrence in patients with ischemic stroke. This study included patients with ischemic stroke from the Nanjing Stroke Registry Program. The glycemic gap was calculated by subtracting the estimated average blood glucose from the blood glucose at admission. Multivariable Cox proportional hazards regression analysis was performed to explore the association between the glycemic gap and the risk of stroke recurrence. The Bayesian hierarchical logistic regression model was used to estimate the effects of the glycemic gap on stroke recurrence stratified by diabetes mellitus and atrial fibrillation. Among 2734 enrolled patients, 381 (13.9%) patients experienced stroke recurrence during a median follow-up of 3.02 years. In multivariate analysis, glycemic gap (high group vs. median group) was associated with significantly increased risk for stroke recurrence (adjusted hazard ratio, 1.488; 95% confidence interval, 1.140-1.942; p = .003) and had varying effects on stroke recurrence depending on atrial fibrillation. The restricted cubic spline curve showed a U-shaped relationship between the glycemic gap and stroke recurrence (p = .046 for nonlinearity). Our study found that the glycemic gap was significantly associated with stroke recurrence in patients with ischemic stroke. The glycemic gap was consistently associated with stroke recurrence across subgroups and had varying effects depending on atrial fibrillation.

A phase Ib trial of combined PKC and MEK inhibition with sotrastaurin and binimetinib in patients with metastatic uveal melanoma.

Uveal melanoma is a disease characterized by constitutive activation of the G alpha pathway and downstream signaling of protein kinase C (PKC) and the mitogen-activated protein kinase (MAPK) pathway. While limited clinical activity has been observed in patients with metastatic disease with inhibition of PKC or MEK alone, preclinical data has demonstrated synergistic antitumor effects with concurrent inhibition of PKC and MEK. We conducted a phase Ib study of the PKC inhibitor sotrastaurin in combination with the MEK inhibitor binimetinib in patients with metastatic uveal melanoma using a Bayesian logistic regression model guided by the escalation with overdose control principle (NCT01801358). Serial blood samples and paired tumor samples were collected for pharmacokinetic (PK) and pharmacodynamic analysis. Thirty-eight patients were treated across six dose levels. Eleven patients experienced DLTs across the five highest dose levels tested, most commonly including vomiting (n=3), diarrhea (n=3), nausea (n=2), fatigue (n=2) and rash (n=2). Common treatment related adverse events included diarrhea (94.7%), nausea (78.9%), vomiting (71.1%), fatigue (52.6%), rash (39.5%), and elevated blood creating phosphokinase (36.8%). Two dose combinations satisfying criteria for the maximum tolerated dose (MTD) were identified: (1) sotrastaurin 300 mg and binimetinib 30 mg; and, (2) sotrastaurin 200 mg and binimetinib 45 mg. Exposure to both drugs in combination was consistent with single-agent data for either drug, indicating no PK interaction between sotrastaurin and binimetinib. Stable disease was observed in 60.5% of patients treated. No patient achieved a radiographic response per RECIST v1.1. Concurrent administration of sotrastaurin and binimetinib is feasible but associated with substantial gastrointestinal toxicity. Given the limited clinical activity achieved with this regimen, accrual to the phase II portion of the trial was not initiated.

First in human phase Ⅰ study of BMS-986299 as monotherapy and combined with nivolumab and ipilimumab in advanced solid tumors.

e14584 Background: Immune checkpoint inhibitors (ICIs) face hurdles in overcoming the immunosuppressive tumor microenvironment. BMS-986299 is first-in-class, NOD-, LRR-, and pyrin domain containing 3 (NLRP3) inflammasome agonist enhancing adaptive immune and T cell memory responses. In-vivo studies demonstrated BMS-986299 synergy with ICIs and induction of durable immunological memory. Methods: This multicenter, open-label, phase I [NCT03444753] study assessed safety and tolerability of intra-tumoral (IT) BMS-986299 monotherapy (Part 1A) and in combination (Part 1B) with the programmed death 1 receptor antibody nivolumab (N), and the anti-cytotoxic T lymphocyte associated protein 4 antibody ipilimumab (I) in advanced solid tumors. Eligible patients (pts) regardless of prior ICI with accessible solid tumors (≥ 2 lesions) received IT BMS-986299 administered on day 1, 8, 15, 22, 29 for Cycle 1 (C1) and on days 1, 29 until C2, 8 weeks apart across 7 dose escalation cohorts in Part 1A (75μg-4000μg; Bayesian Logistic Regression Model (BLRM). In Part 1B, BMS-986299 was given across 3 dose escalation cohorts (75μg-2000μg; BLRM Copula) with nivolumab 480 mg q4weeks and ipilimumab 1 mg/kg q8weeks (fixed dosing) from day 1 up to 24 months (m). Cross-over (CS) was allowed from Part 1A to Part 1B. Results: The study was terminated early, unrelated to safety, due to the COVID-19 pandemic and did not complete enrollment. From April 2018 to February 2022, 50 pts were treated (33; Part 1A and 17; Part 1B). Eleven pts crossed over to Part 1B. Median age of pts enrolled was 59 years and with preponderance in female gender (58.0%) and white (80.0%) pt. participation. Forty-two pts (70%) had treatment related adverse events (TRAEs). Grade (G)3/G4 TRAEs included tubulointerstitial nephritis (1; Part 1A), hepatitis (1; CS), colitis, diabetic ketoacidosis, elevated liver enzymes (1 each; Part 1B). No G5 TRAEs or dose limiting toxicities were seen. In Part 1A, 3 pts (melanoma; bladder; anal) achieved stable disease (SD). In CS cohort, 1 triple negative breast cancer (BC) pt had partial response (PR; -60%) for 27.4 m and 3 had SD (breast; cSCC, squamous oropharynx). In Part 1B, 3 pts had PR including 2 pts with hormone positive BC (1 pt with -86%; 3.2m and 1 pt with -39%; 2m) and 1 pt with cutaneous squamous cell cancer (cSCC) (-43%; 1m). Four pts in part 1B had SD (2 breast, 1 lung, 1 chondrosarcoma). All 4 responders received the combination of BMS-986299 with N + I with 3 responders receiving BMS-986299 at 2000μg and one at 300μg. 3 responders (2 breast; 1 cSCC) had prior ICI therapy. Conclusions: BMS-986299 in combination with ICI resulted in modest clinical activity in select tumor types and carried a manageable toxicity profile. However, definitive conclusions could not be drawn as the study was not powered for inference testing. Further exploration is warranted to define the role of this novel agent in the ICI refractory population. Clinical trial information: NCT03444753 .

A phase 0/Ia study of BI 907828 concentrations in brain tissue and a nonrandomized, open-label, dose escalation study of BI 907828 in combination with radiotherapy in patients with newly diagnosed glioblastoma (GBM).

TPS2081 Background: Murine double minute 2 (MDM2) directly regulates the stability of the tumor suppressor p53, and dysregulation of this pathway can critically disrupt cellular response to genomic stress and lead to cell death. BI 907828 is a highly potent inhibitor of the MDM2-p53 interface that triggers accumulation of p53 and subsequent apoptosis in multiple GBM patient-derived xenografts (PDXs) with wild-type p53. Moreover, the combination of BI 907828 and radiotherapy provides significant survival extension in multiple orthotopic GBM PDXs. Methods: An integrated Phase 0/1 trial design is being used to evaluate the potential for BI 907828 to achieve a therapeutic concentration in brain tumor tissue and to test the safety of the combination with radiation. The co-primary endpoints for the Phase 0 study are i) measured total concentration of BI 907828 and ii) calculated unbound concentration of BI 907828 in contrast and non-contrast enhancing regions of tumor. Key inclusion criteria for the Phase 0 study are patients at least 18 years of age, ECOG performance status of 0 or 1, histologic or radiologic new diagnosis of GBM, and planned surgical resection. The first approximately 6 patients enrolled in the Phase 0 will be treated at the starting dose of 30 mg, and the future patients enrolled will be dosed at higher levels (45 mg or 60 mg) provided there are no safety concerns. If dose-limiting toxicities (DLTs) are encountered, then a Bayesian Logistic Regression Model (BLRM) with overdose control will be used to guide the dose level decisions for the additional patients. Multiple, image-registered tumor samples and intra-operative plasma samples will be analyzed for BI 907828 by LC-MS/MS. Intra-tumoral variation in BI 907828 will be correlated with pharmacodynamic effects of MDM2 inhibition (elevation of p53, p21, MDM2, GDF15, PUMA). Those patients completing the Phase 0 study and who meet additional eligibility criteria can enroll on the Phase 1a portion of the study. The pathologic inclusion criteria are TP53 wild-type, IDH1/2 wild-type, MGMT unmethylated GBM. Co-primary endpoints for the Phase 1a trial are i) occurrence of DLTs during and within 21 days after completion of radiation therapy, and ii) occurrence of adverse events throughout the treatment period. Tolerability of inter-patient dose-escalation of BI 907828 (30, 45, 60 mg) combined with radiation therapy (60 Gy in 30 fractions) will be evaluated with dose-escalation guided by a two-parameter BLRM with overdose control. Patients will remain on BI 907828 monotherapy following completion of radiotherapy until progression. To date, two patients have enrolled on the Phase 0 study, and one patient went on to enroll on the Phase 1a portion of the study. Clinical trial information: NCT05376800 .

First-in-human clinical trial design of a first-in-class theranostic approach with a peptide-based radioligand targeting CA IX-expressing tumors.

TPS3160 Background: Expression of carbonic anhydrase IX (CA IX), a cell surface glycoprotein, can be induced by a state of hypoxia or by mutation of the Von Hippel-Lindau tumor suppressor gene. CAIX expression in tumors is often associated with progressive disease and overall poor outcomes in various solid cancers. Notably, high CA IX expression has been observed in 83%, 40%, and 29% of clear cell renal cell carcinoma (ccRCC), pancreatic ductal adenocarcinoma (PDAC), and colorectal cancer (CRC) specimens, respectively1. CA IX may therefore represent a valuable therapeutic target. DPI-4452 is a first-in-class cyclic peptide that binds with high affinity to CA IX. Radiolabeling DPI-4452 with gallium-68 ([68Ga]Ga-DPI-4452) or lutetium-177 ([177Lu]Lu-DPI-4452) is an innovative theranostic approach for identifying and treating patients (pts) with CA IX-expressing tumors. Pre-clinical data has shown that [177Lu]Lu-DPI-4452 is taken up by tumors and induces dose-dependent growth inhibition in ccRCC and CRC xenograft mouse models1,2. Methods: We present the design of a Phase I/II study (NCT05706129) of [68Ga]Ga-DPI-4452 and [177Lu]Lu-DPI-4452 in pts with unresectable locally advanced or metastatic ccRCC, CRC or PDAC. The study is comprised of three phases: imaging (Part A), dose-escalation (Part B), and expansion (Part C). In Part A, pts will receive a single intravenous (IV) 185 MBq dose of [68Ga]Ga-DPI-4452. Safety, tolerability, pharmacokinetics (PK), dosimetry and imaging characteristics of [68Ga]Ga-DPI-4452 will be evaluated using positron emission tomography/computed tomography (CT) at four timepoints on Day 1. All pts will be followed for seven days post-injection for safety observations. In Part B, based on [68Ga]Ga-DPI-4452 uptake, pts will receive sequential, escalating doses of IV [177Lu]Lu-DPI-4452 on Day 1 of each 28-day cycle for up to eight cycles. During Cycle 1, each pt will undergo up to four whole-body planar imaging and single photon emission CT/CT analyses for dosimetry over seven days. In Part B, the recommended Phase 2 dose of [177Lu]Lu-DPI-4452 will be determined, guided by a Bayesian Logistic Regression model. Safety, PK, and dosimetry will be assessed to minimize any radiation-induced toxicity and to optimize absorbed radiation dose in the tumor. Upon completion of Part B, Part C will enroll up to three cohorts of pts with CA IX-positive ccRCC, CRC or PDAC to assess preliminary efficacy of [177Lu]Lu-DPI-4452, with a primary endpoint of objective response rate, as measured by RECIST v1.1. References Wiedemann N et al. J Nucl Med 2022;63(Suppl 2):4048. Attinger A et al. Eur J Nucl Med Mol Imaging 2022;49(Suppl 1):S1–S751. Clinical trial information: NCT05706129 .

A Bayesian approach to predictive uncertainty in chemotherapy patients at risk of acute care utilization.

Machine learning (ML) predictions are becoming increasingly integrated into medical practice. One commonly used method, ℓ1-penalised logistic regression (LASSO), can estimate patient risk for disease outcomes but is limited by only providing point estimates. Instead, Bayesian logistic LASSO regression (BLLR) models provide distributions for risk predictions, giving clinicians a better understanding of predictive uncertainty, but they are not commonly implemented. This study evaluates the predictive performance of different BLLRs compared to standard logistic LASSO regression, using real-world, high-dimensional, structured electronic health record (EHR) data from cancer patients initiating chemotherapy at a comprehensive cancer centre. Multiple BLLR models were compared against a LASSO model using an 80-20 random split using 10-fold cross-validation to predict the risk of acute care utilization (ACU) after starting chemotherapy. This study included 8439 patients. The LASSO model predicted ACU with an area under the receiver operating characteristic curve (AUROC) of 0.806 (95% CI: 0.775-0.834). BLLR with a Horseshoe+prior and a posterior approximated by Metropolis-Hastings sampling showed similar performance: 0.807 (95% CI: 0.780-0.834) and offers the advantage of uncertainty estimation for each prediction. In addition, BLLR could identify predictions too uncertain to be automatically classified. BLLR uncertainties were stratified by different patient subgroups, demonstrating that predictive uncertainties significantly differ across race, cancer type, and stage. BLLRs are a promising yet underutilised tool that increases explainability by providing risk estimates while offering a similar level of performance to standard LASSO-based models. Additionally, these models can identify patient subgroups with higher uncertainty, which can augment clinical decision-making. This work was supported in part by the National Library Of Medicine of the National Institutes of Health under Award Number R01LM013362. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Phase I study of MP0317, a FAP-dependent DARPin, for tumor-localized CD40 activation in patients with advanced solid tumors.

2584 Background: CD40 is a master switch for both innate and adaptive immune systems. The clinical development of CD40 agonists has been hampered by dose-limiting toxicity (DLT) due to systemic CD40 activation and peripheral target-mediated drug disposition (TMDD). MP0317 is a CD40-agonistic DARPin (designed ankyrin repeat protein), developed to reduce systemic toxicity. MP0317 is exclusively active in the presence of fibroblast activating protein (FAP) expressed by cancer associated fibroblasts in the tumor microenvironment. This allows reaching serum levels that overcome the TMDD and have the potential to deliver sustained and tumor-localized CD40 activation. Ongoing clinical testing aims at establishing its safety/tolerability profile, PK/PD characteristics, and a recommended dose for combination therapy. Methods: NCT05098405 is a Phase 1, multicenter, open label, dose escalation study followed by a safety expansion part in adult patients with advanced solid tumors. The dose escalation scheme uses an adaptive Bayesian logistic regression model with overdose control. Peripheral blood biomarkers are analyzed by immuno-assays and flow cytometry, and baseline and on-treatment tumor biopsies are characterized by RNA sequencing and immunofluorescence. Eligible patients are enrolled into 9 sequentially escalating dose cohorts of MP0317 (0.03-10 mg/kg), administered IV 3-weekly (Q3W) or 1-weekly (Q1W) until disease progression or unacceptable toxicity. Results: At submission, 23 patients across 6 cohorts completed the study, with no DLT observed. Patient enrolling in higher dose cohorts is ongoing. The most frequent AE was grade 2 infusion related reaction in 5 patients. The 23 patients received ≥2 (range 2-8) doses (range 0.03 – 3 mg/kg) of MP0317 across five Q3W and one Q1W cohorts and completed the 28-day DLT period. PK data confirmed all patients were exposed to MP0317 with broadly dose-dependent Cmax and no sign of accumulation. PD data was derived from patients dosed with 0.03 – 3 mg/kg. Soluble biomarkers (sFAP and sCD40) showed target engagement in periphery with signs of FAP saturation at ≥0.5 mg/kg. Colocalization of MP0317 with FAP and CD40 in the tumor was confirmed in 4 out of 8 evaluable paired tumor biopsies. Whole transcriptome and gene set enrichment analyses showed an upregulation of genes related to B-cell trafficking (CXCL3, CXCR5, CCR6, CCL20) upon treatment. Transient increase of IFNg-induced chemokines (CXCL9, CXCL10) was observed, whereas the pro-inflammatory cytokines (TNFa, IL-2, IL-6, IL-8) were not upregulated. Conclusions: Clinical data and the lack of pro-inflammatory circulating cytokines confirm MP0317 is safe and well-tolerated, while analysis of paired pre- and on-treatment tumor biopsies suggests early evidence of tumor-localized CD40 activation. The current data enables further evaluation in a combination setting. Clinical trial information: NCT05098405 .

Bayesian Logistic Regression Analysis for Spatial Patterns of inter-seasonal Drought Persistence

Drought is one of the disastrous natural hazards with complex seasonal and spatial patterns. Understanding the spatial patterns of drought and predicting the likelihood of inter-seasonal drought persistence can provide substantial operational guidelines for water resource management and agricultural production. This study examines drought persistence by identifying the spatial patterns of seasonal drought frequency and inter-seasonal drought persistence in the northeastern region of Pakistan. The Standardized Precipitation Index (SPI) with a three-month time scale is used to examine meteorological drought. Furthermore, Bayesian logistic regression is used to calculate the probability and odds ratios of drought occurrence in the current season, given the previous season's SPI values. For instance, at Balakot station, for the summer-to-autumn season, the value of the odds ratio is significant (6.78). It shows that one unit increase in SPI of the summer season will cause a 5.78 times to increase in odds of autumn drought occurrence. The average drought frequency varies from 37.3 to 89.1%, whereas the average inter-seasonal drought persistence varies from 21.9 to 91.7% in the study region. Results indicate that some areas in the study region, like Kakul and Garhi Dupatta, are more prone to drought and vulnerable to inter-seasonal drought persistence. Furthermore, the Bayesian logistic regression results reveal a negative relationship between spring drought occurrence and winter SPI, demonstrating that the overall study region is more prone to winter-to-spring drought persistence and less vulnerable to summer-to-autumn drought persistence. Overall study has concluded that the region's seasonal drought forecast is challenging due to uncertain drought persistence patterns. However, the Bayesian logistic regression model provides more accurate and precise regional seasonal drought forecasts. The outcome of the present study provides scientific evidence to develop early warning systems and manage seasonal crops in Pakistan.

Neighborhood Deprivation and Association With Neonatal Intensive Care Unit Mortality and Morbidity for Extremely Premature Infants

Socioeconomic status affects pregnancy and neurodevelopment, but its association with hospital outcomes among premature infants is unknown. The Area Deprivation Index (ADI) is a validated measure of neighborhood disadvantage that uses US Census Bureau data on income, educational level, employment, and housing quality. To determine whether ADI is associated with neonatal intensive care unit (NICU) mortality and morbidity in extremely premature infants. This retrospective cohort study was performed at 4 level IV NICUs in the US Northeast, Mid-Atlantic, Midwest, and South regions. Non-Hispanic White and Black infants with gestational age of less than 29 weeks and born between January 1, 2012, and December 31, 2020, were included in the analysis. Addresses were converted to census blocks, identified by Federal Information Processing Series codes, to link residences to national ADI percentiles. ADI, race, birth weight, sex, and outborn status. In the primary outcome, the association between ADI and NICU mortality was analyzed using bayesian logistic regression adjusted for race, birth weight, outborn status, and sex. Risk factors were considered significant if the 95% credible intervals excluded zero. In the secondary outcome, the association between ADI and NICU morbidities, including late-onset sepsis, necrotizing enterocolitis (NEC), and severe intraventricular hemorrhage (IVH), were also analyzed. A total of 2765 infants with a mean (SD) gestational age of 25.6 (1.7) weeks and mean (SD) birth weight of 805 (241) g were included in the analysis. Of these, 1391 (50.3%) were boys, 1325 (47.9%) reported Black maternal race, 498 (18.0%) died before NICU discharge, 692 (25.0%) developed sepsis or NEC, and 353 (12.8%) had severe IVH. In univariate analysis, higher median ADI was found among Black compared with White infants (77 [IQR, 45-93] vs 57 [IQR, 32-77]; P < .001), those who died before NICU discharge vs survived (71 [IQR, 45-89] vs 64 [IQR, 36-86]), those with late-onset sepsis or NEC vs those without (68 [IQR, 41-88] vs 64 [IQR, 35-86]), and those with severe IVH vs those without (69 [IQR, 44-90] vs 64 [IQR, 36-86]). In a multivariable bayesian logistic regression model, lower birth weight, higher ADI, and male sex were risk factors for mortality (95% credible intervals excluded zero), while Black race and outborn status were not. The ADI was also identified as a risk factor for sepsis or NEC and severe IVH. The findings of this cohort study of extremely preterm infants admitted to 4 NICUs in different US geographic regions suggest that ADI was a risk factor for mortality and morbidity after adjusting for multiple covariates.