An archival perspective on pretraining data.

Alongside an explosion in research and development related to large language models, there has been a concomitant rise in the creation of pretraining datasets-massive collections of text, typically scraped from the web. Drawing on the field of archival studies, we analyze pretraining datasets as informal archives-heterogeneous collections of diverse material that mediate access to knowledge. We use this framework to identify impacts of pretraining data creation and use beyond directly shaping model behavior and reveal how choices about what is included in pretraining data necessarily involve subjective decisions about values. In doing so, the archival perspective helps us identify opportunities for researchers who study the social impacts of technology to contribute to confronting the challenges and trade-offs that arise in creating pretraining datasets at this scale.

Outcomes Following Use of Bovine Pericardium (Xenoprosthetic) Grafts for Reconstruction of Mycotic Aortic Aneurysms and Infected Aortic Grafts: A Systematic Review and Meta-Analysis.

BACKGROUND: Infected aortic grafts and mycotic aneurysms represent one of the most complex challenges faced by vascular surgeons. Treatment has progressed from extra-anatomical bypass to in situ reconstruction. Additionally, bovine pericardium reconstruction (BPR) has increased, due to accessibility and reduced lower limb morbidity. There remains, however, limited evidence for its use. The aim is to pool all known data to understand outcomes following BPR of mycotic aneurysms or infected vascular grafts. METHODS: A systematic review was conducted in November 2021 with subsequent computerized meta-analysis of the pooled results and a final search in March 2022. Three databases, Excerpta Medica dataBASE (EMBASE), Cumulative Index of Nursing and Allied Health Literature (CINAHL), and National Institutes of Health PubMed (PubMed), were searched for the search term "(bovine OR xenoprosthetic) AND (aneurysm)", according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. RESULTS: From 9 studies, there were 133 patients: 67% graft infections and 33% mycotic aneurysms. Fifty-seven percent of reconstructions were in the abdominal aorta and 33% were in the thoracic aorta. One hundred fifty-eight pathogens were identified, including Staphylococcus aureus (23%), Candida albicans (13%), and Escherichia coli (13%). In 12%, no microorganisms were identified. Thirty-day mortality was 19.14% (CI 10.83-28.71), late mortality was 19.08% (confidence interval [CI] 7.76-32.83), and overall mortality was 40.20% (CI 29.82-50.97). One patient died intraoperatively. There were a total of 151 in-hospital complications after 30 days postoperation. Common complications were acute renal failure (17%), pneumonia (14%), delirium (12%), respiratory insufficiency (11%) and renal insufficiency (7%). Lower limb ischemia was low, occurring in 5.66% (CI 0.54-13.82) of patients. Loss of graft patency leading to reintervention occurred in 1.20% (CI 0.00-7.71) of the grafts. Reinfection rate was 0.00% (CI 0.00-1.21). CONCLUSIONS: This meta-analysis highlights low reinfection and high graft patency using BPR with medium-length follow-up; however, there remain limited long-term and comparative data regarding options for aortic reconstruction. As expected in this complex cohort, the complication rate and 30-day mortality remain high.

A systematic review and meta-analysis of the use of the Omniflow II biosynthetic graft for aortic reconstruction.

OBJECTIVE: Despite the improvements in xenogeneic grafts and surgical techniques, management of aortic graft infection has remained challenging. The optimal graft material has remained controversial, with high rates of reinfection using prosthetic grafts and a limited time for venous harvest in an emergent setting. Recent studies have highlighted an increase in the use of Omniflow II biosynthetic vascular grafts (LeMaitre Vascular, Burlington, MA) for aortic reconstruction. The primary aim of the present study was to review the key outcomes for the Omniflow II graft in terms of reinfection and complications. METHODS: The National Healthcare Service healthcare databases advanced search function was used to search nine databases for the search term "Omniflow." The present study complied with the PRISMA (preferred reporting items for systematic review and meta-analysis) statement. Eligible studies related to aortic graft infection or in situ aortic reconstruction were selected in accordance with prespecified eligibility criteria and included for review. Data on the surgical technique, comorbidities, graft reinfection, mortality, and complications were combined. The data were analyzed using Stata/MP, version 17 (StataCorp, College Station, TX), and the probabilities were pooled using a DerSimonian and Laird random effects model with Freeman-Tukey arcsine transformation. RESULTS: Six studies with 60 patients (44 men; age range, 29-89 years) were included. Of the 60 patients, 25 had undergone surgical reconstruction because of early graft infection (<4 months after the index procedure), 24 for late graft infection, and 3 because of mycotic aneurysms. Eight high-risk patients had undergone surgical reconstruction for prevention of an initial graft infection. Staphylococcus aureus, Escherichia coli, and S. epidermis were the most common organisms. Early mortality was 8.83% (95% confidence interval [CI], 1.12%-20.53%), and late mortality was 18.49% (95% CI, 5.51%-35.34%). Follow-up varied from 9 months to 2 years. No graft rupture or graft degeneration had occurred during follow-up. However, 6.2% (95% CI, 0.39%-15.81%) had experienced early graft occlusion, and 3.83% (95% CI, 0.00%-16.34%) had developed early graft stenosis. Two cases of postoperative reinfection were reported. The freedom from reinfection was 97.71% (95% CI, 87.94%-100.00%). CONCLUSIONS: Use of the Omniflow II graft for aortic reconstruction is a feasible alternative with acceptable mortality and low reinfection rates. However, there is a risk of limb occlusion. Although these studies were of low quality, the Omniflow II graft shows promise in this difficult patient cohort, especially when bifurcated reconstruction is required.

Superior mesenteric artery syndrome secondary to thyrotoxicosis.

Superior mesenteric artery (SMA) syndrome is defined as a narrowed space and decreased angle between the SMA and aorta leading to partial or complete obstruction of the third portion of the duodenum. SMA syndrome patients may have comorbid conditions associated with extreme weight loss, hypermetabolism or malnutrition. We present the case of a 55-year-old male with SMA syndrome due to acute weight loss secondary to thyrotoxicosis. The patient was nutritionally optimized and euthyroid prior to undergoing a robotic-assisted duodenojejunostomy. In this patient, the thyrotoxicosis was controlled medically, and he remained euthyroid postoperatively. His duodenal obstruction was relieved by the operation and he continued to gain weight appropriately.

Platypnoea-orthodeoxia syndrome and hemidiaphragm paralysis.

A woman in her 70s was admitted to hospital with worsening shortness of breath and no prior respiratory history of note. This patient's shortness of breath was posture-dependent; symptoms were markedly worse and oxygen saturations were lower on sitting upright than in recumbency. Her shortness of breath had started several weeks prior to admission and had slowly worsened. Chest X-ray revealed a raised right hemidiaphragm. Further investigation revealed a patent foramen ovale, which was managed with percutaneous closure. This is one of several cases that demonstrate right-to-left shunting through a septal defect secondary to right hemidiaphragmatic paralysis. However, previous reports have not provided a clear guide for management of these cases. We suggest where patients are admitted with new onset breathlessness and platypnoea-orthodeoxia, a septal defect should be suspected. In this report, we have suggested a flowchart for the investigation and management of platypnoea-orthodeoxia syndrome.

Integration of Epigenetic Mechanisms into Non-Genotoxic Carcinogenicity Hazard Assessment: Focus on DNA Methylation and Histone Modifications.

Epigenetics involves a series of mechanisms that entail histone and DNA covalent modifications and non-coding RNAs, and that collectively contribute to programing cell functions and differentiation. Epigenetic anomalies and DNA mutations are co-drivers of cellular dysfunctions, including carcinogenesis. Alterations of the epigenetic system occur in cancers whether the initial carcinogenic events are from genotoxic (GTxC) or non-genotoxic (NGTxC) carcinogens. NGTxC are not inherently DNA reactive, they do not have a unifying mode of action and as yet there are no regulatory test guidelines addressing mechanisms of NGTxC. To fil this gap, the Test Guideline Programme of the Organisation for Economic Cooperation and Development is developing a framework for an integrated approach for the testing and assessment (IATA) of NGTxC and is considering assays that address key events of cancer hallmarks. Here, with the intent of better understanding the applicability of epigenetic assays in chemical carcinogenicity assessment, we focus on DNA methylation and histone modifications and review: (1) epigenetic mechanisms contributing to carcinogenesis, (2) epigenetic mechanisms altered following exposure to arsenic, nickel, or phenobarbital in order to identify common carcinogen-specific mechanisms, (3) characteristics of a series of epigenetic assay types, and (4) epigenetic assay validation needs in the context of chemical hazard assessment. As a key component of numerous NGTxC mechanisms of action, epigenetic assays included in IATA assay combinations can contribute to improved chemical carcinogen identification for the better protection of public health.

Expression Levels of Three Key Genes CCNB1, CDC20, and CENPF in HCC Are Associated With Antitumor Immunity.

INTRODUCTION: Hepatocellular carcinoma (HCC) is the most common primary liver cancer with a low 5-year survival rate. The heterogeneity of HCC makes monotherapy unlikely. The development of diagnostic programs and new treatments targeting common genetic events in the carcinogenic process are providing further insights into the management of HCC. The aim of this study was firstly to validate key genes that are involved in promoting HCC development and as biomarkers for early diagnosis and, secondly, to define their links with antitumor immunity including inhibitory checkpoints. METHODS: Multiple databases including Gene Expression Omnibus (GEO), Gene Expression Profiling Interactive Analysis (GEPIA), Kaplan-Meier Plotter, UALCAN, and Oncomine were used for target gene screening and establishment of a co-expression network. Clinical data and RNAseq of 367 HCC patients were downloaded from the Cancer Genome Atlas (TCGA) database. The diagnostic and prognostic value of screened genes were tested by receiver operating characteristic (ROC) curve and correlation analysis. The links with the key genes in HCC and antitumor immunity were defined using both blood and liver tissue collected prospectively from HCC patients in our center. RESULTS: Upregulation of CCNB1, CDC20, and CENPF was commonly observed in HCC and are involved in the p53 signal pathway. The hepatic mRNA expression levels of these three genes were strongly associated with patients' prognosis and expressed high value of area under the ROC curve (AUC). Further analysis revealed that these three genes were positively correlated with the gene expression levels of IFN-γ, TNF-α, and IL-17 in peripheral blood. In addition, the expression of CENPF showed positive correlation with the percentage of CD8pos T cells and negative correlation with the percentage of CD4pos T cells in the peripheral blood. In the HCC microenvironment, the transcript levels of these three genes and inhibitory checkpoint molecules including PD-1, CTLA-4, and TIM-3 were positively correlated. CONCLUSION: The upregulation of CCNB1, CDC20, and CENPF genes was a common event in hepatocarcinogenesis. Expression levels of CCNB1, CDC20, and CENPF showed potential for early diagnosis and prediction of prognosis in HCC patients. There is a close association between three genes and Th1/Th17 cytokines as well as the count of CD4pos and CD8pos T cells. The positive correlation between the three genes and inhibitory checkpoint genes, PD-1, CTLA-4, and TIM-3, indicates that these genes are linked with weakened antitumor immunity in HCC. Our findings may provide further insights into developing novel therapies for HCC.

A comprehensive view on mechanistic approaches for cancer risk assessment of non-genotoxic agrochemicals.

Currently the only methods for non-genotoxic carcinogenic hazard assessment accepted by most regulatory authorities are lifetime carcinogenicity studies. However, these involve the use of large numbers of animals and the relevance of their predictive power and results has been scientifically challenged. With increased availability of innovative test methods and enhanced understanding of carcinogenic processes, it is believed that tumour formation can now be better predicted using mechanistic information. A workshop organised by the European Partnership on Alternative Approaches to Animal Testing brought together experts to discuss an alternative, mechanism-based approach for cancer risk assessment of agrochemicals. Data from a toolbox of test methods for detecting modes of action (MOAs) underlying non-genotoxic carcinogenicity are combined with information from subchronic toxicity studies in a weight-of-evidence approach to identify carcinogenic potential of a test substance. The workshop included interactive sessions to discuss the approach using case studies. These showed that fine-tuning is needed, to build confidence in the proposed approach, to ensure scientific correctness, and to address different regulatory needs. This novel approach was considered realistic, and its regulatory acceptance and implementation can be facilitated in the coming years through continued dialogue between all stakeholders and building confidence in alternative approaches.

Chemical carcinogen safety testing: OECD expert group international consensus on the development of an integrated approach for the testing and assessment of chemical non-genotoxic carcinogens.

While regulatory requirements for carcinogenicity testing of chemicals vary according to product sector and regulatory jurisdiction, the standard approach starts with a battery of genotoxicity tests (which include mutagenicity assays). If any of the in vivo genotoxicity tests are positive, a lifetime rodent cancer bioassay may be requested, but under most chemical regulations (except plant protection, biocides, pharmaceuticals), this is rare. The decision to conduct further testing based on genotoxicity test outcomes creates a regulatory gap for the identification of non-genotoxic carcinogens (NGTxC). With the objective of addressing this gap, in 2016, the Organization of Economic Cooperation and Development (OECD) established an expert group to develop an integrated approach to the testing and assessment (IATA) of NGTxC. Through that work, a definition of NGTxC in a regulatory context was agreed. Using the adverse outcome pathway (AOP) concept, various cancer models were developed, and overarching mechanisms and modes of action were identified. After further refining and structuring with respect to the common hallmarks of cancer and knowing that NGTxC act through a large variety of specific mechanisms, with cell proliferation commonly being a unifying element, it became evident that a panel of tests covering multiple biological traits will be needed to populate the IATA. Consequently, in addition to literature and database investigation, the OECD opened a call for relevant assays in 2018 to receive suggestions. Here, we report on the definition of NGTxC, on the development of the overarching NGTxC IATA, and on the development of ranking parameters to evaluate the assays. Ultimately the intent is to select the best scoring assays for integration in an NGTxC IATA to better identify carcinogens and reduce public health hazards.

Nomenclature Used by Family Caregivers to Describe and Characterize Neuropsychiatric Symptoms.

BACKGROUND AND OBJECTIVES: Neuropsychiatric symptoms (NPS) are a core feature of Alzheimer's disease and related dementias that are characterized by a fluctuating course. NPS are challenging to manage and contribute to high rates of burden among family caregivers. Successful information exchange between clinicians and family caregivers is critical for facilitating effective management of NPS. However, this communication is often challenging due to inconsistent terminology and classification of symptoms and limited understanding of how family caregivers recognize and describe symptoms. The objective of this study was to examine the language family caregivers' use to describe and contextualize NPS. RESEARCH DESIGN AND METHODS: Descriptive qualitative study of 20 family caregivers in a mostly urban county in the Midwestern United States using semistructured interviews. Caregiver descriptions of NPS were analyzed using directed content and text analysis to examine terminology, followed by a thematic analysis approach to examine contextualization of NPS. RESULTS: Caregivers employed shared terminologies to describe NPS that differed substantially from clinical terminology used to classify symptoms. Caregivers frequently engaged sense-making as a strategy to explain NPS. This sense-making served to contextualize patterns in behavior and was characterized by explanatory, situational, and strategy-oriented frameworks for understanding behavior in terms of its purpose and meaning. Caregivers' descriptions of NPS reflected broad overlap between individual NPS (i.e., agitation and care resistance) that would generally be considered clinically distinct symptoms. DISCUSSION AND IMPLICATIONS: Nomenclature surrounding NPS may vary considerably between family caregivers and clinicians, and should be evaluated in partnership with people with dementia and their caregivers to ensure supportive interventions and resources are responsive to caregivers' interpretation of symptoms and sense-making.

Analysis of exposure margins in developmental toxicity studies for detection of human teratogens.

The draft Step 2 ICH S5(R3) guideline includes an exposure-based endpoint as an option for selecting the high-dose in reproductive and developmental toxicity studies. To help determine an appropriate exposure margin for embryofetal developmental toxicity testing, a retrospective analysis was undertaken to determine what threshold would have been sufficient to detect hazards to embryofetal development in rats and rabbits for 18 known and 4 presumed human teratogens. The analysis showed that using a high dose that provided at least a 6-fold exposure margin in the developmental toxicity studies would have been sufficient to detect the teratogenic hazard with relevance for humans for all these therapeutics. With regards to human risk assessment practices for developmental toxicity, the analysis showed that, after excluding lenalidomide and pomalidomide data in rats, all available AUC margins at the NOAEL for the induction of malformations or embryofetal lethality were <4-fold of the exposure at the MRHD for all 22 therapeutics. These data support the proposed general approach of increased level of concern for human risk when exposure margins of the NOAEL to the MRHD are <10-fold, reduced concern when the exposure margins are 10- to 25-fold, and minimal concern when the exposure margin is > 25-fold.

Skin sensitization testing needs and data uses by US regulatory and research agencies.

United States regulatory and research agencies may rely upon skin sensitization test data to assess the sensitization hazards associated with dermal exposure to chemicals and products. These data are evaluated to ensure that such substances will not cause unreasonable adverse effects to human health when used appropriately. The US Consumer Product Safety Commission, the US Environmental Protection Agency, the US Food and Drug Administration, the Occupational Safety and Health Administration, the National Institute for Occupational Safety and Health, and the US Department of Defense are member agencies of the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM). ICCVAM seeks to identify opportunities for the use of non-animal replacements to satisfy these testing needs and requirements. This review identifies the standards, test guidelines, or guidance documents that are applicable to satisfy each of these agency's needs; the current use of animal testing and flexibility for using alternative methodologies; information needed from alternative tests to fulfill the needs for skin sensitization data; and whether data from non-animal alternative approaches are accepted by these US federal agencies.

Status of acute systemic toxicity testing requirements and data uses by U.S. regulatory agencies.

Acute systemic toxicity data are used by a number of U.S. federal agencies, most commonly for hazard classification and labeling and/or risk assessment for acute chemical exposures. To identify opportunities for the implementation of non-animal approaches to produce these data, the regulatory needs and uses for acute systemic toxicity information must first be clarified. Thus, we reviewed acute systemic toxicity testing requirements for six U.S. agencies (Consumer Product Safety Commission, Department of Defense, Department of Transportation, Environmental Protection Agency, Food and Drug Administration, Occupational Safety and Health Administration) and noted whether there is flexibility in satisfying data needs with methods that replace or reduce animal use. Understanding the current regulatory use and acceptance of non-animal data is a necessary starting point for future method development, optimization, and validation efforts. The current review will inform the development of a national strategy and roadmap for implementing non-animal approaches to assess potential hazards associated with acute exposures to industrial chemicals and medical products. The Acute Toxicity Workgroup of the Interagency Coordinating Committee on the Validation of Alternative Methods (ICCVAM), U.S. agencies, non-governmental organizations, and other stakeholders will work to execute this strategy.

Human relevance of rodent liver tumors: Key insights from a Toxicology Forum workshop on nongenotoxic modes of action.

The Toxicology Forum sponsored a workshop in October 2016, on the human relevance of rodent liver tumors occurring via nongenotoxic modes of action (MOAs). The workshop focused on two nuclear receptor-mediated MOAs (Constitutive Androstane Receptor (CAR) and Peroxisome Proliferator Activated Receptor-alpha (PPARα), and on cytotoxicity. The goal of the meeting was to review the state of the science to (1) identify areas of consensus and differences, data gaps and research needs; (2) identify reasons for inconsistencies in current regulatory positions; and (3) consider what data are needed to demonstrate a specific MOA, and when additional research is needed to rule out alternative possibilities. Implications for quantitative risk assessment approaches were discussed, as were implications of not considering MOA and dose in hazard characterization and labeling schemes. Most, but not all, participants considered the CAR and PPARα MOAs as not relevant to humans based on quantitative and qualitative differences. In contrast, cytotoxicity is clearly relevant to humans, but a threshold applies. Questions remain for all three MOAs concerning what data are necessary to determine the MOA and to what extent it is necessary to exclude other MOAs.

Origin of the TTC values for compounds that are genotoxic and/or carcinogenic and an approach for their re-evaluation.

The threshold of toxicological concern (TTC) approach is a resource-effective de minimis method for the safety assessment of chemicals, based on distributional analysis of the results of a large number of toxicological studies. It is being increasingly used to screen and prioritize substances with low exposure for which there is little or no toxicological information. The first step in the approach is the identification of substances that may be DNA-reactive mutagens, to which the lowest TTC value is applied. This TTC value was based on the analysis of the cancer potency database and involved a number of assumptions that no longer reflect the state-of-the-science and some of which were not as transparent as they could have been. Hence, review and updating of the database is proposed, using inclusion and exclusion criteria reflecting current knowledge. A strategy for the selection of appropriate substances for TTC determination, based on consideration of weight of evidence for genotoxicity and carcinogenicity is outlined. Identification of substances that are carcinogenic by a DNA-reactive mutagenic mode of action and those that clearly act by a non-genotoxic mode of action will enable the protectiveness to be determined of both the TTC for DNA-reactive mutagenicity and that applied by default to substances that may be carcinogenic but are unlikely to be DNA-reactive mutagens (i.e. for Cramer class I-III compounds). Critical to the application of the TTC approach to substances that are likely to be DNA-reactive mutagens is the reliability of the software tools used to identify such compounds. Current methods for this task are reviewed and recommendations made for their application.

Prediction of skin sensitization potency using machine learning approaches.

The replacement of animal use in testing for regulatory classification of skin sensitizers is a priority for US federal agencies that use data from such testing. Machine learning models that classify substances as sensitizers or non-sensitizers without using animal data have been developed and evaluated. Because some regulatory agencies require that sensitizers be further classified into potency categories, we developed statistical models to predict skin sensitization potency for murine local lymph node assay (LLNA) and human outcomes. Input variables for our models included six physicochemical properties and data from three non-animal test methods: direct peptide reactivity assay; human cell line activation test; and KeratinoSens™ assay. Models were built to predict three potency categories using four machine learning approaches and were validated using external test sets and leave-one-out cross-validation. A one-tiered strategy modeled all three categories of response together while a two-tiered strategy modeled sensitizer/non-sensitizer responses and then classified the sensitizers as strong or weak sensitizers. The two-tiered model using the support vector machine with all assay and physicochemical data inputs provided the best performance, yielding accuracy of 88% for prediction of LLNA outcomes (120 substances) and 81% for prediction of human test outcomes (87 substances). The best one-tiered model predicted LLNA outcomes with 78% accuracy and human outcomes with 75% accuracy. By comparison, the LLNA predicts human potency categories with 69% accuracy (60 of 87 substances correctly categorized). These results suggest that computational models using non-animal methods may provide valuable information for assessing skin sensitization potency. Copyright © 2017 John Wiley & Sons, Ltd.

Multivariate models for prediction of human skin sensitization hazard.

One of the Interagency Coordinating Committee on the Validation of Alternative Method's (ICCVAM) top priorities is the development and evaluation of non-animal approaches to identify potential skin sensitizers. The complexity of biological events necessary to produce skin sensitization suggests that no single alternative method will replace the currently accepted animal tests. ICCVAM is evaluating an integrated approach to testing and assessment based on the adverse outcome pathway for skin sensitization that uses machine learning approaches to predict human skin sensitization hazard. We combined data from three in chemico or in vitro assays - the direct peptide reactivity assay (DPRA), human cell line activation test (h-CLAT) and KeratinoSens™ assay - six physicochemical properties and an in silico read-across prediction of skin sensitization hazard into 12 variable groups. The variable groups were evaluated using two machine learning approaches, logistic regression and support vector machine, to predict human skin sensitization hazard. Models were trained on 72 substances and tested on an external set of 24 substances. The six models (three logistic regression and three support vector machine) with the highest accuracy (92%) used: (1) DPRA, h-CLAT and read-across; (2) DPRA, h-CLAT, read-across and KeratinoSens; or (3) DPRA, h-CLAT, read-across, KeratinoSens and log P. The models performed better at predicting human skin sensitization hazard than the murine local lymph node assay (accuracy 88%), any of the alternative methods alone (accuracy 63-79%) or test batteries combining data from the individual methods (accuracy 75%). These results suggest that computational methods are promising tools to identify effectively the potential human skin sensitizers without animal testing. Published 2016. This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Integrated decision strategies for skin sensitization hazard.

One of the top priorities of the Interagency Coordinating Committee for the Validation of Alternative Methods (ICCVAM) is the identification and evaluation of non-animal alternatives for skin sensitization testing. Although skin sensitization is a complex process, the key biological events of the process have been well characterized in an adverse outcome pathway (AOP) proposed by the Organisation for Economic Co-operation and Development (OECD). Accordingly, ICCVAM is working to develop integrated decision strategies based on the AOP using in vitro, in chemico and in silico information. Data were compiled for 120 substances tested in the murine local lymph node assay (LLNA), direct peptide reactivity assay (DPRA), human cell line activation test (h-CLAT) and KeratinoSens assay. Data for six physicochemical properties, which may affect skin penetration, were also collected, and skin sensitization read-across predictions were performed using OECD QSAR Toolbox. All data were combined into a variety of potential integrated decision strategies to predict LLNA outcomes using a training set of 94 substances and an external test set of 26 substances. Fifty-four models were built using multiple combinations of machine learning approaches and predictor variables. The seven models with the highest accuracy (89-96% for the test set and 96-99% for the training set) for predicting LLNA outcomes used a support vector machine (SVM) approach with different combinations of predictor variables. The performance statistics of the SVM models were higher than any of the non-animal tests alone and higher than simple test battery approaches using these methods. These data suggest that computational approaches are promising tools to effectively integrate data sources to identify potential skin sensitizers without animal testing. Published 2016. This article has been contributed to by US Government employees and their work is in the public domain in the USA.