Empirical Bayesian Research Articles

Nephrotoxicity of immune checkpoint inhibitor therapy: a pharmacovigilance study.

Immune checkpoint inhibitor (ICI) therapy has demonstrated impressive clinical benefits across cancers. However, adverse drug reactions (ADRs) occur in every organ system, often due to autoimmune syndromes. We sought to investigate the association between ICI therapy and nephrotoxicity using a pharmacovigilance database, hypothesizing that inflammatory nephrotoxic syndromes would be reported more frequently in association with ICIs. We analyzed VigiBase, the World Health Organization pharmacovigilance database, to identify renal ADRs (rADRs), such as nephritis, nephropathy and vascular disorders, reported in association with ICI therapy. We performed a disproportionality analysis to explore if rADRs were reported at a different rate with one of the ICI drugs compared with rADRs in the entire database, using an empirical Bayes estimator as a significance screen and defining the effect size with a reporting odds ratio (ROR). We found 2341 rADR for all examined ICI drugs, with a disproportionality signal solely for nephritis [ROR = 3.67, 95% confidence interval (CI) 3.34-4.04]. Examining the different drugs separately, pembrolizumab, nivolumab and ipilimumab + nivolumab combination therapy had significantly higher reporting odds of nephritis than the other ICI drugs (ROR = 4.54, 95% CI 3.81-5.4; ROR = 3.94, 95% CI 3.40-4.56; ROR 3.59, 95% CI 2.71-4.76, respectively). Using a pharmacovigilance method, we found increased odds of nephritis when examining rADRs associated with ICI therapy. Pembrolizumab, nivolumab and a combination of ipilimumab + nivolumab showed the highest odds. Clinicians should consider these findings and be aware of the increased risk of nephritis, especially in patients treated with pembrolizumab, when administering ICI therapy.

Fast screening of covariates in population models empowered by machine learning

One of the objectives of Pharmacometry (PMX) population modeling is the identification of significant and clinically relevant relationships between parameters and covariates. Here, we demonstrate how this complex selection task could benefit from supervised learning algorithms using importance scores. We compare various classical methods with three machine learning (ML) methods applied to NONMEM empirical Bayes estimates: random forest, neural networks (NNs), and support vector regression (SVR). The performance of the ML models is assessed using receiver operating characteristic (ROC) curves. The F1 score, which measures test accuracy, is used to compare ML and PMX approaches. Methods are applied to different scenarios of covariate influence based on simulated pharmacokinetics data. ML achieved similar or better F1 scores than stepwise covariate modeling (SCM) and conditional sampling for stepwise approach based on correlation tests (COSSAC). Correlations between covariates and the number of false covariates does not affect the performance of any method, but effect size has an impact. Methods are not equivalent with respect to computational speed; SCM is 30 and 100-times slower than NN and SVR, respectively. The results are validated in an additional scenario involving 100 covariates. Taken together, the results indicate that ML methods can greatly increase the efficiency of population covariate model building in the case of large datasets or complex models that require long run-times. This can provide fast initial covariate screening, which can be followed by more conventional PMX approaches to assess the clinical relevance of selected covariates and build the final model.

Modeling In-Match Sports Dynamics Using the Evolving Probability Method

The prediction of sport event results has always drawn attention from a vast variety of different groups of people, such as club managers, coaches, betting companies, and the general population. The specific nature of each sport has an important role in the adaption of various predictive techniques founded on different mathematical and statistical models. In this paper, a common approach of modeling sports with a strongly defined structure and a rigid scoring system that relies on an assumption of independent and identical point distributions is challenged. It is demonstrated that such models can be improved by introducing dynamics into the match models in the form of sport momentums. Formal mathematical models for implementing these momentums based on conditional probability and empirical Bayes estimation are proposed, which are ultimately combined through a unifying hybrid approach based on the Monte Carlo simulation. Finally, the method is applied to real-life volleyball data demonstrating noticeable improvements over the previous approaches when it comes to predicting match outcomes. The method can be implemented into an expert system to obtain insight into the performance of players at different stages of the match or to study field scenarios that may arise under different circumstances.

Evaluating the Approaches of Small Area Estimation Using Poverty Mapping Data

Nowadays, estimation demand in statistics is increased worldwide to seek out an estimate, or approximation, which may be a value which will be used for various purpose, albeit the input data could also be incomplete, uncertain, or unstable. The development of different estimation methods is trying to provide most accurate estimate and estimation theory deals with finding estimates with good properties. The demand of small area estimation (SAE) method has been increasing rapidly around the world because of its reliability compared to the traditional direct estimation methods, especially in the case of small sample size. This paper mainly focuses on the comparison of several indirect small area estimation methods (poststratified synthetic, SSD and EB estimates) with traditional direct estimator based on a renowned data set. Direct estimator is approximately unbiased but SSD and Post-stratified synthetic estimator is extreme biased. To cope up the problem, we conduct another model-based estimation procedure namely Empirical Bayes (EB) estimator, which is unbiased and compare them using their coefficient of variation (CV). To check the model assumption, we used Q-Q plot as well as a Histogram to confirm the normality, bivariate correlation, Akaike information criterion (AIC). JEL classification numbers: C13, C51, C51. Keywords: Small Area Estimation, Direct Estimation, Indirect Estimation, Empirical Bayes Estimator, Poverty Mapping.

Cardiovascular toxicities associated with abiraterone compared to enzalutamide-A pharmacovigilance study.

BackgroundAndrogen deprivation therapy (ADT) is standard-of-care for advanced prostate cancer. Studies have generally found increased cardiovascular risks associated with ADT, but the comparative risk of newer agents is under-characterized. We defined the cardiac risks of abiraterone and enzalutamide, using gonadotropic releasing hormone (GnRH) agonists to establish baseline ADT risk.MethodsWe used VigiBase, the World Health Organization pharmacovigilance database, to identify cardiac adverse drug reactions (ADRs) in a cohort taking GnRH agonists, abiraterone, or enzalutamide therapy for prostate cancer, comparing them to all other patients. To examine the relationship, we used an empirical Bayes estimator to screen for significance, then calculated the reporting odds ratio (ROR), a surrogate measure of association. A lower bound of a 95% confidence interval (CI) of ROR > 1 reflects a disproportionality signal that more ADRs are observed than expected due to chance.FindingsWe identified 2,433 cardiac ADRs, with higher odds for abiraterone compared to all other VigiBase drugs for overall cardiac events (ROR 1•59, 95% CI 1•48—1•71), myocardial infarction (1•35, 1•16—1•58), arrythmia (2•04, 1•82—2•30), and heart failure (3•02, 2•60—3•51), but found no signal for enzalutamide. Patients on GnRH agonists also had increased risk of cardiac events (ROR 1•21, 95% CI 1•12—1•30), myocardial infarction (1•80, 1•61—2•03) and heart failure (2•06, 1•76—2•41).InterpretationWe found higher reported odds of cardiac events for abiraterone but not enzalutamide. Our data may suggest that patients with significant cardiac comorbidities may be better-suited for therapy with enzalutamide over abiraterone.FundingNone

Investigation of crashes at pedestrian hybrid beacons: Results of a large-scale study in Arizona

Introduction: The pedestrian hybrid beacon (PHB) is a traffic control device used at pedestrian crossings. A recent Arizona Department of Transportation research effort investigated changes in crashes for different severity levels and crash types (e.g., rear-end crashes) due to the PHB presence, as well as for crashes involving pedestrians and bicycles. Method: Two types of methodologies were used to evaluate the safety of PHBs: (a) an Empirical Bayes (EB) before-after study, and (b) a long-term cross-sectional observational study. For the EB before-after evaluation, the research team considered three reference groups: unsignalized intersections, signalized intersections, and both unsignalized and signalized intersections combined. Results: For the signalized and combined unsignalized and signalized intersection groups, all crash types considered showed statistically significant reductions in crashes (e.g., total crashes, fatal and injury crashes, rear-end crashes, fatal and injury rear-end crashes, angle crashes, fatal and injury angle crashes, pedestrian-related crashes, and fatal and injury pedestrian-related crashes). A cross-sectional study was conducted with a larger number of PHBs (186) to identify relationships between roadway characteristics and crashes at PHBs, especially with respect to the distance to an adjacent traffic control signal. The distance to an adjacent traffic signal was found to be significant only at the α = 0.1 level, and only for rear-end and fatal and injury rear-end crashes. Conclusions: This analysis represents the largest known study to date on the safety impacts of PHBs, along with a focus on how crossing and geometric characteristics affect crash patterns. The study showed the safety benefits of PHBs for both pedestrians and vehicles. Practical Applications: The findings from this study clearly support the installation of PHBs at midblock or intersection crossings, as well as at crossings on higher-speed roads.

Spatial Patterns in Hospital-Acquired Infections in Portugal (2014-2017).

Background: Hospital-Acquired Infections (HAIs) represent the most frequent adverse event associated with healthcare delivery and result in prolonged hospital stays and deaths worldwide. Aim: To analyze the spatial patterns of HAI incidence from 2014 to 2017 in Portugal. Methods: Data from the Portuguese Discharge Hospital Register were used. We selected episodes of patients with no infection on admission and with any of the following HAI diagnoses: catheter-related bloodstream infections, intestinal infections by Clostridium difficile, nosocomial pneumonia, surgical site infections, and urinary tract infections. We calculated age-standardized hospitalization rates (ASHR) by place of patient residence. We used empirical Bayes estimators to smooth the ASHR. The Moran Index and Local Index of Spatial Autocorrelation (LISA) were calculated to identify spatial clusters. Results: A total of 318,218 HAIs were registered, with men accounting for 49.8% cases. The median length of stay (LOS) was 9.0 days, and 15.7% of patients died during the hospitalization. The peak of HAIs (n = 81,690) occurred in 2015, representing 9.4% of the total hospital admissions. Substantial spatial inequalities were observed, with the center region presenting three times the ASHR of the north. A slight decrease in ASHR was observed after 2015. Pneumonia was the most frequent HAI in all age groups. Conclusion: The incidence of HAI is not randomly distributed in the space; clusters of high risk in the central region were seen over the entire study period. These findings may be useful to support healthcare policymakers and to promote a revision of infection control policies, providing insights for improved implementation.

Epigenome-wide association study of level and change in cognitive abilities from midlife through late life

BackgroundEpigenetic mechanisms are important in aging and may be involved in late-life changes in cognitive abilities. We conducted an epigenome-wide association study of leukocyte DNA methylation in relation to level and change in cognitive abilities, from midlife through late life in 535 Swedish twins.ResultsMethylation levels were measured with the Infinium Human Methylation 450 K or Infinium MethylationEPIC array, and all sites passing quality control on both arrays were selected for analysis (n = 250,816). Empirical Bayes estimates of individual intercept (age 65), linear, and quadratic change were obtained from latent growth curve models of cognitive traits and used as outcomes in linear regression models. Significant sites (p < 2.4 × 10–7) were followed up in between-within twin pair models adjusting for familial confounding and full-growth modeling. We identified six significant associations between DNA methylation and level of cognitive abilities at age 65: cg18064256 (PPP1R13L) with processing speed and spatial ability; cg04549090 (NRXN3) with spatial ability; cg09988380 (POGZ), cg25651129 (-), and cg08011941 (ENTPD8) with working memory. The genes are involved in neuroinflammation, neuropsychiatric disorders, and ATP metabolism. Within-pair associations were approximately half that of between-pair associations across all sites. In full-growth curve models, associations between DNA methylation and cognitive level at age 65 were of small effect sizes, and associations between DNA methylation and longitudinal change in cognitive abilities of very small effect sizes.ConclusionsLeukocyte DNA methylation was associated with level, but not change in cognitive abilities. The associations were substantially attenuated in within-pair analyses, indicating they are influenced in part by genetic factors.

Learning linear modules in a dynamic network using regularized kernel-based methods

In order to identify one system (module) in an interconnected dynamic network, one typically has to solve a Multi-Input-Single-Output (MISO) identification problem that requires identification of all modules in the MISO setup. For application of a parametric identification method this would require estimating a large number of parameters, as well as an appropriate model order selection step for a possibly large scale MISO problem, thereby increasing the computational complexity of the identification algorithm to levels that are beyond feasibility. An alternative identification approach is presented employing regularized kernel-based methods. Keeping a parametric model for the module of interest, we model the impulse response of the remaining modules in the MISO structure as zero mean Gaussian processes (GP) with a covariance matrix (kernel) given by the first-order stable spline kernel, accounting for the noise model affecting the output of the target module and also for possible instability of systems in the MISO setup. Using an Empirical Bayes (EB) approach the target module parameters are estimated through an Expectation–Maximization (EM) algorithm with a substantially reduced computational complexity, while avoiding extensive model structure selection. Numerical simulations illustrate the potentials of the introduced method in comparison with the state-of-the-art techniques for local module identification.

Model-based approach to sampling optimization in studies of antibacterial drugs for infants and young children.

Clinical trials for pediatric indications and new pediatric drugs face challenges, including the limited blood volume due to the patients’ small bodies. In Japan, the Evaluation Committee on Unapproved or Off‐labeled Drugs with High Medical Needs has discussed the necessity of pediatric indications against the background of a lack of Japanese pediatric data. The limited treatment options regarding antibiotics for pediatric patients are associated with the emergence of antibiotic‐resistant bacteria. Regulatory guidelines promote the use of model‐based drug development to reduce practical and ethical constraints for pediatric patients. Sampling optimization is one of the key study designs for pediatric drug development. In this simulation study, we evaluated the precision of the empirical Bayes estimates of pharmacokinetic (PK) parameters based on the sampling times optimized by published pediatric population PK models. We selected three previous PK studies of cefepime and ciprofloxacin in infants and young children as paradigms. The number of sampling times was reduced from original full sampling times to two to four sampling times based on the Fisher information matrix. We observed that the precision of empirical Bayes estimates of the key PK parameters and the predicted efficacy based on the reduced sampling times were generally comparable to those based on the original full sampling times. The model‐based approach to sampling optimization provided a maximization of PK information with a minimum burden on infants and young children for the future development of pediatric drugs.

MicroRNA biomarkers of type 2 diabetes: A protocol for corroborating evidence by computational genomics and meta-analyses.

BackgroundFew microRNAs were found consistently dysregulated in type 2 diabetes (T2D) that would gain confidence from Big Pharma to develop diagnostic or therapeutic biomarkers. This study aimed to corroborate evidence from eligible microRNAs-T2D association studies according to stringent quality criteria covering both biological and statistical significance in T2D for biomarker development.Methods and analysesControlled microRNA expression profiling studies on human with T2D will be retrieved from PubMed, ScienceDirect, and Embase for selecting the statistically significant microRNAs according to pre-specified search strategies and inclusion criteria. Multiple meta-analyses with restricted maximum-likelihood estimation and empirical Bayes estimation under the random-effects model will be conducted by metafor package in R. Subgroup and sensitivity analyses further examine the microRNA candidates for their disease specificity, tissue specificity, blood fraction specificity, and statistical robustness of evidence. Biologically relevant microRNAs will then be selected through genomic database corroboration. Their association with T2D is further measured by area under the curve (AUC) of receive operating characteristic (ROC). Meta-analysis of AUC of potential biomarkers will also be conducted. Enrichment analysis on potential microRNA biomarkers and their target genes will be performed by iPathwayGuide and clusterProfiler, respectively. The corresponding reporting guidelines will be used to assess the quality of included studies according to their profiling methods (microarray, RT-PCR, and RNA-Seq).Ethics and disseminationNo ethics approval is required since this study does not include identifiable personal patient data.Protocol registration numberCRD42017081659.

Empirical Bayes method using surrounding pixel information for number and brightness analysis

Number and brightness (N&B) analysis is useful for monitoring the spatial distribution of the concentration and oligomeric state of fluorescently labeled proteins in cells. N&B analysis is based on the statistical analysis of fluorescence images by using the method of moments (MoM). Furthermore, N&B analysis can determine the particle number and particle brightness, which indicate the concentration and oligomeric state, respectively. However, the statistical accuracy and precision are limited in actual experiments with fluorescent proteins, owing to low excitation and the limited number of images. In this study, we applied maximum likelihood (ML) estimation and maximum a posteriori (MAP) estimation coupled with the empirical Bayes (EB) method (referred to as EB-MAP). In EB-MAP, we constructed a simple prior distribution for a pixel to utilize the information of the surrounding pixels. To evaluate the accuracy and precision of our method, we conducted simulations and experiments and compared the results of MoM, ML, and EB-MAP. The results showed that MoM estimated the particle number with many outliers. The outliers hampered the visibility of the spatial distribution and cellular structure. In contrast, EB-MAP suppressed the number of outliers and improved the visibility notably. The precision of EB-MAP was better by an order of magnitude in terms of particle number and 1.5 times better in terms of particle brightness compared with those of MoM. The proposed method (EB-MAP-N&B) is applicable to studies on fluorescence imaging and would aid in accurately recognizing changes in the concentration and oligomeric state in cells. Our results hold significant importance because quantifying the concentration and oligomeric state would contribute to the understanding of dynamic processes in molecular mechanism in cells.

Perbandingan Metode Bootstrap, Jacknife Jiang Dan Area Specific Jacknife Pada Pendugaan Mean Square Error Model Beta-Bernoulli

Small area estimation is defined as a statistical technique for estimating the parameters of a subpopulation with a small sample size. One method of estimating small area parameters is the Empirical Bayes (EB) method. The accuracy of the Empirical Bayes (EB) estimator can be measured by evaluating the Mean Squared Error (MSE). In this study, 3 methods to determine MSE in the EB estimator of the Beta-Bernoulli model will be compared, namely the Bootstrap, Jackknife Jiang and Area-specific Jackknife methods. The study is carried out theoretically and empirically through simulation with R-studio software version 1.2.5033. The simulation results in a number of areas and pairs of prior distribution parameter values, namely Beta, show the effect of sample size and parameter value pairs on the Mean Square Error (MSE) value. The larger the number of areas and the smaller the initial 𝛽, the smaller the MSE value. The area-specific Jackknife method produces the smallest MSE in the number of areas 100 and the Beta parameter value 0.1.

Effect of increasing the freeway posted speed limit on entry ramp speed-change lane crash frequency

The objective of this research is to model the effect of increasing the freeway posted speed limit on entry ramp speed-change lane crash frequency. The entry ramps along I-85, I-77, I-277, and I-485 in Charlotte, North Carolina, United States were considered to collect data, identify critical predictor variables, develop safety performance functions (SPFs) by severity, and validate the SPFs. Data for 76 and 20 entry ramp speed-change lanes (reference and comparison groups) were used for modeling and validation/calibration, respectively. The Negative Binomial log-link distribution-based models were the best-fit as the crash data are over-dispersed. The traffic volume, the speed difference between the freeway and the ramp, and the ramp configuration are significant predictor variables to estimate total, property damage only (PDO), and non-incapacitating and possible injury crash frequency at entry ramp speed-change lanes. The Empirical Bayes (EB) method was then applied to examine the effect of increasing the freeway posted speed limit on 48 entry ramp speed-change lanes (treatment/test group) on I-485. The findings from the EB method-based analysis indicate that there was a decrease in the total, non-incapacitating and possible injury, and PDO crash frequencies at 18, 37, and 18 (38%, 77%, and 38%) of the selected 48 entry ramp speed-change lanes after the increase in the freeway posted speed limit on I-485. An increase or no change in the total, non-incapacitating and possible injury, and PDO crash frequencies was also observed at some entry ramp speed-change lanes after the increase in the freeway posted speed limit on I-485.

Shrinkage estimation of the three-parameter logistic model.

The three-parameter logistic model is widely used to model the responses to a proficiency test when the examinees can guess the correct response, as is the case for multiple-choice items. However, the weak identifiability of the parameters of the model results in large variability of the estimates and in convergence difficulties in the numerical maximization of the likelihood function. To overcome these issues, in this paper we explore various shrinkage estimation methods, following two main approaches. First, a ridge-type penalty on the guessing parameters is introduced in the likelihood function. The tuning parameter is then selected through various approaches: cross-validation, information criteria or using an empirical Bayes method. The second approach explored is based on the methodology developed to reduce the bias of the maximum likelihood estimator through an adjusted score equation. The performance of the methods is investigated through simulation studies and a real data example.

The unique molecular mechanism of diabetic nephropathy: a bioinformatics analysis of over 250microarray datasets.

Background/AimsDiabetic nephropathy (DN) is one of the main causes of end-stage kidney disease worldwide. Emerging studies have suggested that its pathogenesis is distinct from nondiabetic renal diseases in many aspects. However, it still lacks a comprehensive understanding of the unique molecular mechanism of DN.MethodsA total of 255 Affymetrix U133 microarray datasets (Affymetrix, Santa Calra, CA, USA) of human glomerular and tubulointerstitial tissues were collected. The 22 215 Affymetrix identifiers shared by the Human Genome U133 Plus 2.0 and U133A Array were extracted to facilitate dataset pooling. Next, a linear model was constructed and the empirical Bayes method was used to select the differentially expressed genes (DEGs) of each kidney disease. Based on these DEG sets, the unique DEGs of DN were identified and further analyzed using gene ontology and pathway enrichment analysis. Finally, the protein–protein interaction networks (PINs) were constructed and hub genes were selected to further refine the results.ResultsA total of 129 and 1251 unique DEGs were identified in the diabetic glomerulus (upregulated n = 83 and downregulated n = 203) and the diabetic tubulointerstitium (upregulated n = 399 and downregulated n = 874), respectively. Enrichment analysis revealed that the DEGs in the diabetic glomerulus were significantly associated with the extracellular matrix, cell growth, regulation of blood coagulation, cholesterol homeostasis, intrinsic apoptotic signaling pathway and renal filtration cell differentiation. In the diabetic tubulointerstitium, the significantly enriched biological processes and pathways included metabolism, the advanced glycation end products–receptor for advanced glycation end products signaling pathway in diabetic complications, the epidermal growth factor receptor (EGFR) signaling pathway, the FoxO signaling pathway, autophagy and ferroptosis. By constructing PINs, several nodes, such as AGR2, CSNK2A1, EGFR and HSPD1, were identified as hub genes, which might play key roles in regulating the development of DN.ConclusionsOur study not only reveals the unique molecular mechanism of DN but also provides a valuable resource for biomarker and therapeutic target discovery. Some of our findings are promising and should be explored in future work.

Transformed Fay-Herriot model with measurement error in covariates

Statistical agencies are often asked to produce small area estimates (SAEs) for positively skewed variables. When domain sample sizes are too small to support direct estimators, effects of skewness of the response variable can be large. As such, it is important to appropriately account for the distribution of the response variable given available auxiliary information. Motivated by this issue and in order to stabilize the skewness and achieve normality in the response variable, we propose an area-level log-measurement error model on the response variable. Then, under our proposed modeling framework, we derive an empirical Bayes (EB) predictor of positive small area quantities subject to the covariates containing measurement error. We propose a corresponding mean squared prediction error (MSPE) of EB predictor using both a jackknife and a bootstrap method. We show that the order of the bias is where m is the number of small areas. Finally, we investigate the performance of our methodology using both design-based and model-based simulation studies.

Eutopic endometrium from women with endometriosis and chlamydial endometritis share immunological cell types and DNA repair imbalance: A transcriptome meta-analytical perspective

The aim of this study was to identify the key similarities between the eutopic endometrium of women with endometriosis and chlamydia-induced endometritis taking into account tissue microenvironment heterogeneity, transcript gene profile, and enriched pathways. A meta-analysis of whole transcriptome microarrays was performed using publicly available data, including samples containing both glandular and stromal endometrial components. Control samples were obtained from women without any reported pathological condition. Only samples obtained during the proliferative menstrual phase were included. Cellular tissue heterogeneity was predicted using a method that integrates gene set enrichment and deconvolution approaches. The batch effect was estimated by principal variant component analysis and removed using an empirical Bayes method. Differentially expressed genes were identified using an adjusted p-value < 0.05 and fold change = 1.5. The protein-protein interaction network was built using the STRING database and interaction score over 400. The Molecular Signatures Database was used to analyse the functional enrichment analysis. Both conditions showed similarities in cell types in the microenvironment, particularly CD4+ and CD8+ Tem cells, NKT cells, Th2 cells, basophils, and eosinophils. With regards to the regulation of cellular senescence and DNA integrity/damage checkpoint, which are commonly enriched pathways, 21 genes were down-regulated and directly related to DNA repair. Compared to the endometriosis samples, some chlamydial endometritis samples presented a lack of enriched immune pathways. Our results suggest that both conditions show similar distributions of microenvironment cell types, the downregulation of genes involved in DNA repair and cell cycle control, and pathways involved in immune response evasion.

41250 Machine Learning to Identify Predictors of Iatrogenic Injury Using Empirical Bayes Estimates: A Cohort Study of Pressure Injury Prevention

ABSTRACT IMPACT: A machine learning approach using electronic health records can combine descriptive, population-level factors of pressure injury outcomes. OBJECTIVES/GOALS: Pressure injuries cause 60,000 deaths and cost $26 billion annually in the US, but prevention is laborious. We used clinical data to develop a machine learning algorithm for predicting pressure injury risk and prescribe the timing of intervention to help clinicians balance competing priorities. METHODS/STUDY POPULATION: We obtained 94,745 electronic health records with 7,000 predictors to calibrate a predictive algorithm of pressure injury risk. Machine learning was used to mine features predicting changes in pressure injury risk; random forests outperformed neural networks, boosting and bagging in feature selection. These features were fit to multilevel ordered logistic regression to create an algorithm that generated empirical Bayes estimates informing a decision-rule for follow-up based on individual risk trajectories over time. We used cross-validation to verify predictive validity, and constrained optimization to select a best-fit algorithm that reduced the time required to trigger patient follow-up. RESULTS/ANTICIPATED RESULTS: The algorithm significantly improved prediction of pressure injury risk (p&lt;0.001) with an area under the ROC curve of 0.60 compared to the Braden Scale, a traditional clinician instrument of pressure injury risk. At a specificity of 0.50, the model achieved a sensitivity of 0.63 within 2.5 patient-days. Machine learning identified categorical increases in risk when patients were prescribed vasopressors (OR=16.4, p&lt;0.001), beta-blockers (OR=4.8, p&lt;0.001), erythropoietin stimulating agents (OR=3.0, p&lt;0.001), or were ordered a urinalysis screen (OR=9.1, p&lt;0.001), lipid panel (OR=5.7, p&lt;0.001) or pre-albumin panel (OR=2.0, p&lt;0.001). DISCUSSION/SIGNIFICANCE OF FINDINGS: This algorithm could help hospitals conserve resources within a critical period of patient vulnerability for pressure injury not reimbursed by Medicare. Savings generated by this approach could justify investment in machine learning to develop electronic warning systems for many iatrogenic injuries.

Smoking Modifies Pancreatic Cancer Risk Loci on 2q21.3.

Germline variation and smoking are independently associated with pancreatic ductal adenocarcinoma (PDAC). We conducted genome-wide smoking interaction analysis of PDAC using genotype data from four previous genome-wide association studies in individuals of European ancestry (7,937 cases and 11,774 controls). Examination of expression quantitative trait loci data from the Genotype-Tissue Expression Project followed by colocalization analysis was conducted to determine whether there was support for common SNP(s) underlying the observed associations. Statistical tests were two sided and P < 5 × 10-8 was considered statistically significant. Genome-wide significant evidence of qualitative interaction was identified on chr2q21.3 in intron 5 of the transmembrane protein 163 (TMEM163) and upstream of the cyclin T2 (CCNT2). The most significant SNP using the Empirical Bayes method, in this region that included 45 significantly associated SNPs, was rs1818613 [per allele OR in never smokers 0.87, 95% confidence interval (CI), 0.82-0.93; former smokers 1.00, 95% CI, 0.91-1.07; current smokers 1.25, 95% CI 1.12-1.40, P interaction = 3.08 × 10-9). Examination of the Genotype-Tissue Expression Project data demonstrated an expression quantitative trait locus in this region for TMEM163 and CCNT2 in several tissue types. Colocalization analysis supported a shared SNP, rs842357, in high linkage disequilibrium with rs1818613 (r 2 = 0. 94) driving both the observed interaction and the expression quantitative trait loci signals. Future studies are needed to confirm and understand the differential biologic mechanisms by smoking status that contribute to our PDAC findings. SIGNIFICANCE: This large genome-wide interaction study identifies a susceptibility locus on 2q21.3 that significantly modified PDAC risk by smoking status, providing insight into smoking-associated PDAC, with implications for prevention.