Continuous Model Averaging for Benchmark Dose Analysis: Averaging Over Distributional Forms.

When estimating a benchmark dose (BMD) from chemical toxicity experiments, model averaging is recommended by the National Institute for Occupational Safety and Health, World Health Organization and European Food Safety Authority. Though numerous studies exist for Model Average BMD estimation using dichotomous responses, fewer studies investigate it for BMD estimation using continuous response. In this setting, model averaging a BMD poses additional problems as the assumed distribution is essential to many BMD definitions, and distributional uncertainty is underestimated when one error distribution is chosen a priori. As model averaging combines full models, there is no reason one cannot include multiple error distributions. Consequently, we define a continuous model averaging approach over distributional models and show that it is superior to single distribution model averaging. To show the superiority of the approach, we apply the method to simulated and experimental response data.

BMDExpress 2: enhanced transcriptomic dose-response analysis workflow.

SUMMARY: A new version (version 2) of the genomic dose-response analysis software, BMDExpress, has been created. The software addresses the increasing use of transcriptomic dose-response data in toxicology, drug design, risk assessment and translational research. In this new version, we have implemented additional statistical filtering options (e.g. Williams' trend test), curve fitting models, Linux and Macintosh compatibility and support for additional transcriptomic platforms with up-to-date gene annotations. Furthermore, we have implemented extensive data visualizations, on-the-fly data filtering, and a batch-wise analysis workflow. We have also significantly re-engineered the code base to reflect contemporary software engineering practices and streamline future development. The first version of BMDExpress was developed in 2007 to meet an unmet demand for easy-to-use transcriptomic dose-response analysis software. Since its original release, however, transcriptomic platforms, technologies, pathway annotations and quantitative methods for data analysis have undergone a large change necessitating a significant re-development of BMDExpress. To that end, as of 2016, the National Toxicology Program assumed stewardship of BMDExpress. The result is a modernized and updated BMDExpress 2 that addresses the needs of the growing toxicogenomics user community. AVAILABILITY AND IMPLEMENTATION: BMDExpress 2 is available at https://github.com/auerbachs/BMDExpress-2/releases. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Quantitative Risk Assessment: Developing a Bayesian Approach to Dichotomous Dose-Response Uncertainty.

Model averaging for dichotomous dose-response estimation is preferred to estimate the benchmark dose (BMD) from a single model, but challenges remain regarding implementing these methods for general analyses before model averaging is feasible to use in many risk assessment applications, and there is little work on Bayesian methods that include informative prior information for both the models and the parameters of the constituent models. This article introduces a novel approach that addresses many of the challenges seen while providing a fully Bayesian framework. Furthermore, in contrast to methods that use Monte Carlo Markov Chain, we approximate the posterior density using maximum a posteriori estimation. The approximation allows for an accurate and reproducible estimate while maintaining the speed of maximum likelihood, which is crucial in many applications such as processing massive high throughput data sets. We assess this method by applying it to empirical laboratory dose-response data and measuring the coverage of confidence limits for the BMD. We compare the coverage of this method to that of other approaches using the same set of models. Through the simulation study, the method is shown to be markedly superior to the traditional approach of selecting a single preferred model (e.g., from the U.S. EPA BMD software) for the analysis of dichotomous data and is comparable or superior to the other approaches.

Quantitative meta-analytic approaches for the systematic synthesis of data and hazard identification: A case study of decreased pain sensitivity due to trimethylbenzene exposure.

Traditionally, human health risk assessments have relied on qualitative approaches for hazard identification, which involves weight of evidence determinations that integrate evidence across multiple studies. Recently, the National Research Council has recommended the development of quantitative approaches for evidence integration, including the application of meta-analyses, to help summarize and evaluate the results of a systematic review. In the meta-analytic approach, a pooled effect size is calculated after consideration of multiple potential confounding factors in order to determine whether the entire database under consideration indicates a chemical is a hazard. The following case-study applies qualitative and quantitative approaches to determine whether trimethylbenzene (TMB) isomers represent a neurotoxic hazard, specifically focusing on pain sensitivity. Following a thorough literature search, the only pain sensitivity studies available for TMBs initially seem discordant in their results: effects on pain sensitivity are seen immediately after termination of exposure, appear to resolve 24h after exposure, and then reappear 50 days later following foot-shock. Qualitative consideration of toxicological and toxicokinetic characteristics of the TMB isomers suggests that the observed differences between studies are likely due to testing time and the application of external stressors. Meta-analyses and -regressions support this conclusion: when all studies are included and possible confounders (isomer, testing time, laboratory, etc.) are accounted for, the pooled effect sizes are statistically significant, thus supporting that TMBs are a possible neurotoxic hazard to human health. Ultimately, this case study demonstrates how qualitative and quantitative methods can be combined to provide a robust hazard identification analysis by incorporating more of the available information.

Contribution of particle-size-fractionated airborne lead to blood lead during the National Health and Nutrition Examination Survey, 1999-2008.

The objective of this work is to examine associations between blood lead (PbB) and air lead (PbA) in particulate matter measured at different size cuts by use of PbB concentrations from the National Health and Nutrition Examination Survey and PbA concentrations from the U.S. Environmental Protection Agency for 1999-2008. Three size fractions of particle-bound PbA (TSP, PM10, and PM2.5) data with different averaging times (current and past 90-day average) were utilized. A multilevel linear mixed effect model was used to characterize the PbB-PbA relationship. At 0.15 µg/m(3), a unit decrease in PbA in PM10 was significantly associated with a decrease in PbB of 0.3-2.2 µg/dL across age groups and averaging times. For PbA in PM2.5 and TSP, slopes were generally positive but not significant. PbB levels were more sensitive to the change in PbA concentrations for children (1-5 and 6-11 years) and older adults (≥ 60 years) than teenagers (12-19 years) and adults (20-59 years). For the years following the phase-out of Pb in gasoline and a resulting upward shift in the PbA particle size distribution, PbA in PM10 was a statistically significant predictor of PbB. The results also suggest that age could affect the PbB-PbA association, with children having higher sensitivity than adults.

Associations between prenatal exposure to air pollution, small for gestational age, and term low birthweight in a state-wide birth cohort.

A range of health effects, including adverse pregnancy outcomes, have been associated with exposure to ambient concentrations of particulate matter (PM) and ozone (O3). The objective of this study was to determine whether maternal exposure to fine particulate matter (PM2.5) and O3 during pregnancy is associated with the risk of term low birthweight and small for gestational age infants in both single and co-pollutant models. Term low birthweight and small for gestational age were determined using all birth certificates from North Carolina from 2003 to 2005. Ambient air concentrations of PM2.5 and O3 were predicted using a hierarchical Bayesian model of air pollution that combined modeled air pollution estimates from the EPA׳s Community Multi-Scale Air Quality (CMAQ) model with air monitor data measured by the EPA׳s Air Quality System. Binomial regression, adjusted for multiple potential confounders, was performed. In adjusted single-pollutant models for the third trimester, O3 concentration was positively associated with small for gestational age and term low birthweight births [risk ratios for an interquartile range increase in O3: 1.16 (95% CI 1.11, 1.22) for small for gestational age and 2.03 (95% CI 1.80, 2.30) for term low birthweight]; however, inverse or null associations were observed for PM2.5 [risk ratios for an interquartile range increase in PM2.5: 0.97 (95% CI 0.95, 0.99) for small for gestational age and 1.01 (95% CI 0.97, 1.06) for term low birthweight]. Findings were similar in co-pollutant models and linear models of birthweight. These results suggest that O3 concentrations in both urban and rural areas may be associated with an increased risk of term low birthweight and small for gestational age births.

Early prenatal exposure to air pollution and its associations with birth defects in a state-wide birth cohort from North Carolina.

BACKGROUND: Few studies have examined the potential relationship between air pollution and birth defects. The objective of this study was to investigate whether maternal exposure to particulate matter (PM2.5 ) and ozone (O3 ) during pregnancy is associated with birth defects among women living throughout North Carolina. METHODS: Information on maternal and infant characteristics was obtained from North Carolina birth certificates and health service data (2003-2005) and linked with information on birth defects from the North Carolina Birth Defects Monitoring Program. The 24-hr PM2.5 and O3 concentrations were estimated using a hierarchical Bayesian model of air pollution generated by combining modeled air pollution predictions from the U.S. Environmental Protection Agency's Community Multi-Scale Air Quality model with air monitor data from the Environmental Protection Agency's Air Quality System. Maternal residence was geocoded and assigned pollutant concentrations averaged over weeks 3 to 8 of gestation. Binomial regression was performed and adjusted for potential confounders. RESULTS: No association was observed between either PM2.5 or O3 concentrations and most birth defects. Positive effect estimates were observed between air pollution and microtia/anotia and lower limb deficiency defects, but the 95% confidence intervals were wide and included the null. CONCLUSION: Overall, this study suggested a possible relationship between air pollution concentration during early pregnancy and certain birth defects (e.g., microtia/anotia, lower limb deficiency defects), although this study did not have the power to detect such an association. The risk for most birth defects does not appear to be affected by ambient air pollution.

ToxicR: A computational platform in R for computational toxicology and dose-response analyses.

The need to analyze the complex relationships observed in high-throughput toxicogenomic and other omic platforms has resulted in an explosion of methodological advances in computational toxicology. However, advancements in the literature often outpace the development of software researchers can implement in their pipelines, and existing software is frequently based on pre-specified workflows built from well-vetted assumptions that may not be optimal for novel research questions. Accordingly, there is a need for a stable platform and open-source codebase attached to a programming language that allows users to program new algorithms. To fill this gap, the Biostatistics and Computational Biology Branch of the National Institute of Environmental Health Sciences, in cooperation with the National Toxicology Program (NTP) and US Environmental Protection Agency (EPA), developed ToxicR, an open-source R programming package. The ToxicR platform implements many of the standard analyses used by the NTP and EPA, including dose-response analyses for continuous and dichotomous data that employ Bayesian, maximum likelihood, and model averaging methods, as well as many standard tests the NTP uses in rodent toxicology and carcinogenicity studies, such as the poly-K and Jonckheere trend tests. ToxicR is built on the same codebase as current versions of the EPA's Benchmark Dose software and NTP's BMDExpress software but has increased flexibility because it directly accesses this software. To demonstrate ToxicR, we developed a custom workflow to illustrate its capabilities for analyzing toxicogenomic data. The unique features of ToxicR will allow researchers in other fields to add modules, increasing its functionality in the future.

Introduction to benchmark dose methods and U.S. EPA's benchmark dose software (BMDS) version 2.1.1.

Traditionally, the No-Observed-Adverse-Effect-Level (NOAEL) approach has been used to determine the point of departure (POD) from animal toxicology data for use in human health risk assessments. However, this approach is subject to substantial limitations that have been well defined, such as strict dependence on the dose selection, dose spacing, and sample size of the study from which the critical effect has been identified. Also, the NOAEL approach fails to take into consideration the shape of the dose-response curve and other related information. The benchmark dose (BMD) method, originally proposed as an alternative to the NOAEL methodology in the 1980s, addresses many of the limitations of the NOAEL method. It is less dependent on dose selection and spacing, and it takes into account the shape of the dose-response curve. In addition, the estimation of a BMD 95% lower bound confidence limit (BMDL) results in a POD that appropriately accounts for study quality (i.e., sample size). With the recent advent of user-friendly BMD software programs, including the U.S. Environmental Protection Agency's (U.S. EPA) Benchmark Dose Software (BMDS), BMD has become the method of choice for many health organizations world-wide. This paper discusses the BMD methods and corresponding software (i.e., BMDS version 2.1.1) that have been developed by the U.S. EPA, and includes a comparison with recently released European Food Safety Authority (EFSA) BMD guidance.

Bayesian hierarchical dose-response meta-analysis of epidemiological studies: Modeling and target population prediction methods.

When assessing the human risks due to exposure to environmental chemicals, traditional dose-response analyses are not straightforward when there are numerous high-quality epidemiological studies of priority cancer and non-cancer health outcomes. Given this wealth of information, selecting a single "best" study on which to base dose-response analyses is difficult and would potentially ignore much of the available data. Therefore, systematic approaches are necessary for the analysis of these rich databases. Examples are meta-analysis (and further, meta-regression), which are well established methods that consider and incorporate information from multiple studies into the estimation of risks due to exposure to environmental contaminants. In this paper, we propose a hierarchical, Bayesian meta-analysis approach for the dose-response analysis of multiple epidemiological studies. This paper is the second of two papers detailing this approach; the first covered "pre-analysis" steps necessary to prepare the data for dose-response modeling. This paper focuses on the hierarchical Bayesian approach to dose-response modeling and extrapolation of risk to populations of interest using the association between bladder cancer and oral inorganic arsenic (iAs) exposure as an illustrative case study. In particular, this paper addresses the modeling of both case-control and cohort studies with a flexible, logistic model in a hierarchical Bayesian framework that estimates study-specific slopes, as well as a pooled slope across all studies. This approach is akin to a random effects model in which no assumption is made a priori that there is a single, common slope for all included studies. Further, this paper also details extrapolation of the estimates of logistic slope to extra risk in a target population using a lifetable analysis and basic assumptions about background iAs exposure levels. In this case, the target population was the general United States population and information on all-cause mortality and incidence and mortality from bladder cancer was used to perform the lifetable analysis. The methods herein were developed for general use in investigating the association between any pollutant and observed health-effects in epidemiological studies. In order to demonstrate these methods, inorganic arsenic was chosen as a case study given the large epidemiological database that exists for this contaminant.

Systematic dose-response of environmental epidemiologic studies: Dose and response pre-analysis.

Meta-analysis approaches can be used to assess the human risks due to exposure to environmental chemicals when there are numerous high-quality epidemiologic studies of priority outcomes in a database. However, methodological issues related to how different studies report effect measures and incorporate exposure into their analyses arise that complicate the pooled analysis of multiple studies. As such, there are "pre-analysis" steps that are often necessary to prepare summary data reported in epidemiologic studies for dose-response analysis. This paper uses epidemiologic studies of arsenic-induced health effects as a case example and addresses the issues surrounding the estimation of mean doses from censored dose- or exposure-intervals reported in the literature (e.g., estimation of mean doses from high exposures that are only reported as an open-ended interval), calculation of a common dose metric for use in a dose-response meta-analysis (one that takes into consideration inter-individual variability), and calculation of response "effective counts" that inherently account for confounders. The methods herein may be generalizable to 1) the analysis of other environmental contaminants with a suitable database of epidemiologic studies, and 2) any meta-analytic approach used to pool information across studies. A second companion paper detailing the use of "pre-analyzed" data in a hierarchical Bayesian dose-response model and techniques for extrapolating risks to target populations follows.

Use of study-specific MOE-like estimates to prioritize health effects from chemical exposure for analysis in human health assessments.

There are unique challenges in estimating dose-response with chemicals that are associated with multiple health outcomes and numerous studies. Some studies are more suitable than others for quantitative dose-response analyses. For such chemicals, an efficient method of screening studies and endpoints to identify suitable studies and potentially important health effects for dose-response modeling is valuable. Using inorganic arsenic as a test case, we developed a tiered approach that involves estimating study-specific margin of exposure (MOE)-like unitless ratios for two hypothetical scenarios. These study-specific unitless ratios are derived by dividing the exposure estimated to result in a 20% increase in relative risk over the background exposure (RRE20) by the background exposure, as estimated in two different ways. In our case study illustration, separate study-specific ratios are derived using estimates of United States population background exposure (RRB-US) and the mean study population reference group background exposure (RRB-SP). Systematic review methods were used to identify and evaluate epidemiologic studies, which were categorized based on study design (case-control, cohort, cross-sectional), various study quality criteria specific to dose-response analysis (number of dose groups, exposure ascertainment, exposure uncertainty), and availability of necessary dose-response data. Both case-control and cohort studies were included in the RRB analysis. The RRE20 estimates were derived by modeling effective counts of cases and controls estimated from study-reported adjusted odds ratios and relative risks. Using a broad (but not necessarily comprehensive) set of epidemiologic studies of multiple health outcomes selected for the purposes of illustrating the RRB approach, this test case analysis would suggest that diseases of the circulatory system, bladder cancer, and lung cancer may be arsenic health outcomes that warrant further analysis. This is suggested by the number of datasets from adequate dose-response studies demonstrating an effect with RRBs close to 1 (i.e., RRE20 values close to estimated background arsenic exposure levels).

Application of systematic evidence mapping to assess the impact of new research when updating health reference values: A case example using acrolein.

BACKGROUND: The environmental health community needs transparent, methodologically rigorous, and rapid approaches for updating human health risk assessments. These assessments often contain reference values for cancer and/or noncancer effects. Increasingly, the use of systematic review methods are preferred when developing these assessments. Systematic evidence maps are a type of analysis that has the potential to be very helpful in the update process, especially when combined with machine-learning software advances designed to expedite the process of conducting a review. OBJECTIVES: To evaluate the applicability of evidence mapping to determine whether new evidence is likely to result in a change to an existing health reference value, using inhalation exposure to the air pollutant acrolein as a case example. METHODS: New literature published since the 2008 California Environmental Protection Agency's Office of Environmental Health Hazard Assessment (OEHHA) Reference Exposure Level (REL) for acrolein was assessed. Systematic review methods were used to search the literature and screening included the use of machine-learning software. The Populations, Exposures, Comparators and Outcomes (PECO) criteria were kept broad to identify studies that characterized acute and chronic exposure and could be informative for hazard characterization. Studies that met the PECO criteria after full-text review were briefly summarized before their suitability for chronic point of departure (POD) derivation and calculation of a reference value was considered. Studies considered potentially suitable underwent a targeted evaluation to determine their suitability for use in dose-response analysis. RESULTS: Over 15,000 studies were identified from scientific databases. Both machine-learning and manual screening processes were used to identify 60 studies considered PECO-relevant after full-text review. Most of these PECO-relevant studies were short-term exposure animal studies (acute or less than 1 month of exposure) and considered less suitable for deriving a chronic reference value when compared to the subchronic study in rats used in the 2008 OEHHA assessment. Thirteen epidemiological studies were identified but had limitations in the exposure assessment that made them less suitable for dose-response compared to the subchronic rat study. Among the 13 studies, there were four controlled trial studies that have the potential to be informative for future acute reference value derivation. Thus, the 2008 subchronic rat study used by OEHHA appears to still be the most appropriate study for chronic reference value derivation. In addition, advances in dosimetric modeling for gases, including new evidence pertinent to acrolein, could be considered when updating existing acrolein toxicity values. CONCLUSIONS: Evidence mapping is a very useful tool to assess the need for updating an assessment based on understanding the potential impact of new studies on revising an existing health reference value. In this case example, the focus was to identify studies suitable for chronic exposure dose-response analysis, while also identifying studies that may be important to consider for acute exposure scenarios, hazard identification, or for future research. This allows the evidence map to be a useful resource for a range of decision-making contexts. Specialized systematic review software increased the efficiency of the process in terms of human resources and time to conduct the analysis.

Can data science inform environmental justice and community risk screening for type 2 diabetes?

BACKGROUND: Having the ability to scan the entire country for potential "hotspots" with increased risk of developing chronic diseases due to various environmental, demographic, and genetic susceptibility factors may inform risk management decisions and enable better environmental public health policies. OBJECTIVES: Develop an approach for community-level risk screening focused on identifying potential genetic susceptibility hotpots. METHODS: Our approach combines analyses of phenotype-genotype data, genetic prevalence of single nucleotide polymorphisms, and census/geographic information to estimate census tract-level population attributable risks among various ethnicities and total population for the state of California. RESULTS: We estimate that the rs13266634 single nucleotide polymorphism, a type 2 diabetes susceptibility genotype, has a genetic prevalence of 56.3%, 47.4% and 37.0% in Mexican Mestizo, Caucasian, and Asian populations. Looking at the top quintile for total population attributable risk, 16 California counties have greater than 25% of their population living in hotspots of genetic susceptibility for developing type 2 diabetes due to this single genotypic susceptibility factor. CONCLUSIONS: This study identified counties in California where large portions of the population may bear additional type 2 diabetes risk due to increased genetic prevalence of a susceptibility genotype. This type of screening can easily be extended to include information on environmental contaminants of interest and other related diseases, and potentially enables the rapid identification of potential environmental justice communities. Other potential uses of this approach include problem formulation in support of risk assessments, land use planning, and prioritization of site cleanup and remediation actions.

An exploratory analysis of the relationship between ambient ozone and particulate matter concentrations during early pregnancy and selected birth defects in Texas.

We performed an exploratory analysis of ozone (O3) and fine particulate matter (PM2.5) concentrations during early pregnancy and multiple types of birth defects. Data on births were obtained from the Texas Birth Defects Registry (TBDR) and the National Birth Defects Prevention Study (NBDPS) in Texas. Air pollution concentrations were previously determined by combining modeled air pollution concentrations with air monitoring data. The analysis generated hypotheses for future, confirmatory studies; although many of the observed associations were null. The hypotheses are provided by an observed association between O3 and craniosynostosis and inverse associations between PM2.5 and septal and obstructive heart defects in the TBDR. Associations with PM2.5 for septal heart defects and ventricular outflow tract obstructions were null using the NBDPS. Both the TBDR and the NBPDS had inverse associations between O3 and septal heart defects. Further research to confirm the observed associations is warranted.

Cross-species coherence in effects and modes of action in support of causality determinations in the U.S. Environmental Protection Agency's Integrated Science Assessment for Lead.

The peer-reviewed literature on the health and ecological effects of lead (Pb) indicates common effects and underlying modes of action across multiple organisms for several endpoints. Based on such observations, the United States (U.S.) Environmental Protection Agency (EPA) applied a cross-species approach in the 2013 Integrated Science Assessment (ISA) for Lead for evaluating the causality of relationships between Pb exposure and specific endpoints that are shared by humans, laboratory animals, and ecological receptors (i.e., hematological effects, reproductive and developmental effects, and nervous system effects). Other effects of Pb (i.e., cardiovascular, renal, and inflammatory responses) are less commonly assessed in aquatic and terrestrial wildlife limiting the application of cross-species comparisons. Determinations of causality in ISAs are guided by a framework for classifying the weight of evidence across scientific disciplines and across related effects by considering aspects such as biological plausibility and coherence. As illustrated for effects of Pb where evidence across species exists, the integration of coherent effects and common underlying modes of action can serve as a means to substantiate conclusions regarding the causal nature of the health and ecological effects of environmental toxicants.

Effect measure modification of blood lead-air lead slope factors.

There is abundant literature finding that susceptibility factors, including race and ethnicity, age, and housing, directly influence blood lead levels. No study has explored how susceptibility factors influence the blood lead-air lead relationship nationally. The objective is to evaluate whether susceptibility factors act as effect measure modifiers on the blood lead-air lead relationship. Participant level blood lead data from the 1999 to 2008 National Health and Nutrition Examination Survey were merged with air lead data from the US Environmental Protection Agency. Linear mixed effects models were run with and without an air lead interaction term for age group, sex, housing age, or race/ethnicity to determine whether these factors are effect measure modifiers for all ages combined and for five age brackets. Age group and race/ethnicity were determined to be effect measure modifiers in the all-age model and for some age groups. Being a child (1-5, 6-11, and 12-19 years) or of Mexican-American ethnicity increased the effect estimate. Living in older housing (built before 1950) decreased the effect estimate for all models except for the 1-5-year group, where older housing was an effect measure modifier. These results are consistent with the peer-reviewed literature of time-activity patterns, ventilation, and toxicokinetics.

Associations of Ozone and PM2.5 Concentrations With Parkinson's Disease Among Participants in the Agricultural Health Study.

OBJECTIVE: This study describes associations of ozone and fine particulate matter with Parkinson's disease observed among farmers in North Carolina and Iowa. METHODS: We used logistic regression to determine the associations of these pollutants with self-reported, doctor-diagnosed Parkinson's disease. Daily predicted pollutant concentrations were used to derive surrogates of long-term exposure and link them to study participants' geocoded addresses. RESULTS: We observed positive associations of Parkinson's disease with ozone (odds ratio = 1.39; 95% CI: 0.98 to 1.98) and fine particulate matter (odds ratio = 1.34; 95% CI: 0.93 to 1.93) in North Carolina but not in Iowa. CONCLUSIONS: The plausibility of an effect of ambient concentrations of these pollutants on Parkinson's disease risk is supported by experimental data demonstrating damage to dopaminergic neurons at relevant concentrations. Additional studies are needed to address uncertainties related to confounding and to examine temporal aspects of the associations we observed.

Influence of urbanicity and county characteristics on the association between ozone and asthma emergency department visits in North Carolina.

BACKGROUND: Air pollution epidemiologic studies, often conducted in large metropolitan areas because of proximity to regulatory monitors, are limited in their ability to examine potential associations between air pollution exposures and health effects in rural locations. METHODS: Using a time-stratified case-crossover framework, we examined associations between asthma emergency department (ED) visits in North Carolina (2006-2008), collected by a surveillance system, and short-term ozone (O3) exposures using predicted concentrations from the Community Multiscale Air Quality (CMAQ) model. We estimated associations by county groupings based on four urbanicity classifications (representative of county size and urban proximity) and county health. RESULTS: O3 was associated with asthma ED visits in all-year and warm season (April-October) analyses [odds ratio (OR) = 1.019; 95% CI: 0.998, 1.040; OR = 1.020; 95% CI: 0.997, 1.044, respectively, for a 20-ppb increase in lag 0-2 days O3]. The association was strongest in Less Urbanized counties, with no evidence of a positive association in Rural counties. Associations were similar when adjusted for fine particulate matter in copollutant models. Associations were stronger for children (5-17 years of age) compared with other age groups, and for individuals living in counties identified with poorer health status compared with counties that had the highest health rankings, although estimated associations for these subgroups had larger uncertainty. CONCLUSIONS: Associations between short-term O3 exposures and asthma ED visits differed by overall county health and urbanicity, with stronger associations in Less Urbanized counties, and no positive association in Rural counties. Results also suggest that children are at increased risk of O3-related respiratory effects.