Pb NAAQS Human Health Risk Assessment – Overview of Design and Implementation November 12 th, 2008 Dr. Zachary Pekara and Dr. Jee-Young Kimb a - Office of Air Quality Planning and Standards (OAQPS), USEPA b – National Center for Environmental Assessment (NCEA), USEPA
Overview of presentation ¢ ¢ ¢ ¢ Background – the role of risk assessment in the National Ambient Air Quality Standards (NAAQS) Key attributes of Pb from a risk assessment standpoint Case study approach Air-quality scenarios Sensitive populations, sentinel health endpoint and blood Pb metric Types of exposure and risk metrics modeled Conceptual framework for the Pb NAAQS risk assessment More detailed overview of indoor dust modeling step Blood Pb results Concentration-response function(s) for IQ loss Key IQ loss (risk) results Areas for refinement of risk assessment approach ADDITIONAL SLIDES 2
Background on NAAQS Process: Statutory Considerations and Role of Administrator ¢ ¢ ¢ 3 NAAQS includes a primary standard (human health focus) and secondary standard (welfare and ecosystem) Primary standard (for public health protection) – judged by the Administrator to protect public health with an adequate margin of safety l Includes consideration for sensitive subpopulations Administrator considers risk and evidence-based information (provided by staff) along with peer-review and public comments in making decision regarding appropriate NAAQS
Background on NAAQS Process: Risk Assessment and Evidence. Based Analysis ¢ Risk assessment – application of more complex step-wise analysis of exposure and resulting risk for residential populations associated with selected case studies l Mechanistic and empirical modeling elements: • Exposure modeling framework • Health impact (risk) modeling framework l ¢ Estimate distribution of exposure and risk for populations within specific study areas (e. g. , area surrounding smelter facility) Evidence-based analysis – use data obtained directly from the literature (empirical) to estimate risk estimates using simple analysis framework l For Pb, have air-to-blood ratio to estimate exposure and simple CR function slope to translate that into IQ loss • IQ loss = Pb-air * AB ratio * IQ loss slope l 4 Generate simple estimate of risk (no characterization of risk distribution across population)
Background on NAAQS Process: Indicator, Level, Averaging Time and Form ¢ ¢ ¢ Indicator: chemical species or mixture that is to be measured (Pb NAAQS is TSP) Level: amount of Pb that can be in ambient air Averaging time: period over which air measurements are averaged to arrive at a level to compare to the level Form: air quality statistics (e. g. , max, or second max) that is to be compared with the level (works with averaging time) EXAMPLE: Current NAAQS: 0. 15 µg/m 3 max rolling 3 month average • Level: 0. 15 ug/m 3 • Averaging time: rolling 3 month average • Form: maximum ¢ 5 Risk Assessment informs: level and to a certain extent averaging time
Key Attributes of Pb-Related Risk with Implication for the NAAQS Review – Multi-pathway and persistent nature of Pb Simplified representation Pb in ambient air penetrates indoors Pb paint deposition to indoor dust 6 deposition outdoor soil Food (crops) Drinking water Auto Pb ingestion of indoor dust inhalation Re-entrainment ingestion of outdoor soil dietary and drinking water ingestion
Key Attributes of Pb-Related Risk with Implication for the NAAQS Review – Air-related and background pathways Non-air related (background) Air-related (policy-relevant) Pb in ambient air penetrates indoors deposition to indoor dust ingestion of indoor dust inhalation 7 deposition outdoor soil ingestion of outdoor soil Food (crops) Drinking water dietary and drinking water ingestion
Key Attributes of Pb-Related Risk with Implication for the NAAQS Review – Non-linearity of Exposure and Risk Modeling ¢ Non-linearity in Pb exposure modeling and IQ concentrationresponse requires consideration of total Pb exposure (not just air-related) in order to representatively “place” a modeled child on the CR function curve 6 pts IQ loss 1 pt 1. 0 8 Blood Pb level (ug/d. L) 10
Design Aspects: Case study approach General urban case study Location-specific urban case study Primary Pb smelter case study Comparatively small area 2 km radius study area 5 -20 km Pb smelter facility Small neighborhood with ambient air levels at standard One single exposure zone (uniform ambient air Pb level and demographics) 9 Larger urban area with varying ambient air Pb levels and demographics 2 km radius residential area surrounding large Pb smelter with varying ambient air Pb levels and demographics Each US Census block is a separate exposure zone (varying ambient air Pb levels and demographics across study area)
Air quality scenarios evaluated ¢ ¢ ¢ 10 Current conditions scenario PREVIOUS – 1978 NAAQS scenario (urban case studies hypothetically assumed to have ambient air Pb levels just meeting current NAAQS) l Assume proportional rollup for location specific urban case studies based on TSP monitor data Alternate (lower) standard levels l 0. 5, 0. 2, 0. 05, and 0. 02 ug/m 3 l Varying averaging times (max monthly and max quarterly)
Sensitive populations, sentinel health endpoint and blood Pb metric selected for risk modeling ¢ ¢ 11 Neurological effects in children (0 -7 yrs of age): developing nervous system in children most sensitive and effects shown to occur at lower blood Pb levels l Evidence for neurological effects is well supported by epi and tox studies l Available epi studies support derivation of CR functions for IQ loss Epi studies investigating neurological effects have focused on number of blood Pb metrics (concurrent, lifetime average, peak, and early childhood). l All 4 metrics have been correlated with IQ, but the concurrent and lifetime average have been shown to have the strongest association (in the Lanphear 2005 pooled analysis) l Concurrent (strongest association of the 4) emphasized in presenting final results
Risk (Pb-related IQ loss): l Population-weighted distributions of total IQ loss l Population incidence estimates • Number of children with total Pb related IQ loss greater than 1 IQ point, 5 IQ points, 7 IQ points, etc. 12 50 th % 95 th % Blood Pb levels (ug. d. L) % of pop ¢ Exposure: l Population-weighted distributions of blood Pb levels 50 th % 95 th % Points of IQ loss % of pop ¢ % of pop Types of Exposure and Risk Metrics: population-weighted distributions and population incidence 1, 350 kids with > 4 IQ points lost Points of IQ loss
Conceptual framework for risk assessment - 1 Location-specific urban case study Single central tendency blood Pb level for entire study area Single population distribution of blood Pb levels for entire study area STEP 1: Multipathway blood Pb modeling % of pop Blood Pb levels (ug. d. L) STEP 3: Application of IQ loss functions 13 % of pop STEP 2: Application of geometric standard deviation (GSD) Single population distribution of IQ loss for entire study area Points of IQ loss
Conceptual framework for risk assessment - 2 Ambient air Pb levels MODEL blood Pb levels (IEUBK) – central -tendency levels for EACH exposure zone MODEL indoor dust Pb levels Background Pb levels (diet and drinking water) Exposure Analysis (central-tendency level) % of pop Soil Pb levels • multi-pathway intake modeling • biokinetic BLL modeling Blood Pb levels (ug. d. L) Demographic data for exposure zones Exposure Analysis (population distribution) MODEL Populationdistribution of blood Pb levels for ENTIRE study area % of pop Inter-individual variability in residential blood Pb levels (GSD) Blood Pb levels (ug. d. L) 14 Risk Characterization (IQ loss) Estimate policy-relevant IQ loss for population percentiles of interest % of pop CR functions relating blood Pb levels and IQ loss MODEL Populationdistribution of IQ points lost for entire study area Points of IQ loss
Modeling Approach: Characterizing indoor dust Pb levels - 1 General urban and location-specific urban case studies • Hybrid model: mechanistic-empirical model • SUB-MODEL 1: Mechanistic compartmental model to predict indoor Pb loadings given ambient air Pb levels (recent-air contribution). Considers: air exchange rates, deposition velocity, cleaning rates and efficiency. Dynamic mass-balance model which is solved for steady-state. • • Background (non-air) component of indoor dust Pb loading estimated by subtracting airrelated estimate from total residential Pb loading estimate. Total estimate of indoor dust Pb levels obtained from HUD dataset (median of US residential range). SUB-MODELS 2 and 3: Empirical-based log-log regression equations used to (critical non-linearity): a) b) Convert wipe equivalent loadings (from mechanistic model) to vacuum loadings and then Convert vacuum loadings to concentrations 15 Primary Pb smelter case study • Log-log regression model based on sitespecific data from the remediation zone. Data include: • Indoor dust Pb concentrations from 17 houses in remediation zone (units of analysis). Temporally-averaged values were used for each house. • Annual-average Pb concentrations from US Census block centroids located within 200 m of each house • Road dust measurements within 300 m of each house • Post-remediation yard soil Pb levels for each house • Model selected relates natural log of ambient air Pb to natural log of indoor dust Pb (this model had better predictive power compared with models which included soil or road dust variables).
Modeling Approach: Characterizing indoor dust Pb levels - 2 16
Modeling Approach: Estimating blood Pb levels (IEUBK modeling) Media Pb concentrations (air, soil, indoor dust, diet, drinking water) (single value across all 7 years) Ingestion and inhalation rates (7 values – differentiated by child age) IEUBK blood Pb model Lifetime average BLL estimate (average of 6 th month to 7 th year) Concurrent BLL estimate (7 th year estimate) Combined with Geometric Standard Deviation (GSD) characterizing inter-individual blood Pb level variability in population 17
Modeling Approach: Blood Pb results (and performance evaluation) Comparison – Modeled Concurrent BLLs for Case Studies Compared to NHANES-IV Data (modeled results are for current conditions) 18
Modeling Approach: Specification of CR Functions for IQ Loss – 1 ¢ Lanphear et al. (2005) – An international pooled analysis from seven prospective cohorts l Development of regression model involved multistep process • First examined fit of linear model then considered quadratic and cubic terms to examine non-linearity • Restrictive cubic spline function indicated that log-linear model provided a good fit to the data l Ten potential confounders considered • Final model adjusted for site, HOME score, birth weight, maternal IQ, and maternal education • Addition of child’s sex, tobacco and alcohol exposure during pregnancy, maternal age at delivery, marital status, and birth order did not alter effect estimate l Four measures of BLL examined • Concurrent, peak, early childhood, and lifetime average all highly correlated, but concurrent BLL exhibited strongest relationship with IQ l Stability of model evaluated • Results of random-effects model were similar to fixed-effects model • Identical log-linear models that were fit with each model omitting data from one of the sites indicated that the pooled analysis did not depend on data from any single cohort 19
Modeling Approach: Specification of CR Functions for IQ Loss – 2 Relationship between Blood Pb and Children’s IQ in Lanphear et al. (2005) Log-linear model (95% CI shaded) for concurrent blood lead concentration adjusted for HOME score, maternal education, maternal IQ, and birth weight. The mean IQ (95% CI) for the intervals <5, 5 -10, 10 -15, 15 -20, and >20 µg/d. L are shown. (Lanphear et al. , 2005) 20 Log-linear model for concurrent blood lead concentration along with linear models for concurrent blood lead levels among children with peak blood lead levels above and below 10 µg/d. L. (Lanphear et al. , 2005)
Modeling Approach: Specification of CR Functions for IQ Loss - 3 Plot of four CR functions specified for the risk assessment (based on Lanphear et al. , (2005) pooled analysis results) Stratified at 7. 5 peak BLL Stratified at 10 peak BLL 21
Modeling Approach: Risk Estimation – Prediction of IQ Loss four CR functions relating blood Pb levels to IQ loss Results of risk modeling LLL function % of pop Results of exposure modeling Blood Pb levels (ug. d. L) Points of IQ loss General urban case study (current conditions, LLL CR function) TOTAL IQ loss Blood Pb Level (concurrent: ug/d. L) 50 th% -4. 5 1. 9 95 th% -7. 7 6. 5 Pathway contribution (based on fraction of total UPTAKE) Population percentile Diet Drinking Water 18% 10% Background 22 Inhalation Indoor dust (air) Indoor dust (other) Outdoor soil/dust 0. 5% 28% 6% 28% Recent Air Past Air
Modeling Approach: Risk Estimation – Risk Results Median population percentile risk (IQ loss) results (LLL CR function) General Urban Case Study LOW BOUND HIGH BOUND Air-related (policy-relevant) risk 23
Areas for Potential Refinement of the Pb NAAQS Risk Assessment Approach ¢ Exposure modeling: l l l ¢ Further refine indoor dust modeling (provide coverage for foot tracking mechanism that links ambient air to indoor dust Pb) Develop probabilistic approach for modeling inter-individual variability in multi-pathway exposure to Pb (with emphasis on ambient-air related pathways) – alternate to GSD approach Refine ability to pathway-apportion exposure (and risk) particularly for higher population percentiles Enhance ability to relate shorter-term changes in Pb exposure to blood Pb levels (enhance shorter-term blood Pb modeling) Refine our ability to model the impact of ambient air-Pb changes on adult blood Pb levels Risk modeling: l l 24 Further refine our understanding of low-exposure (low-blood Pb) IQ loss with the goal of enhancing our CR functions Refine our ability to model other low-exposure related health endpoints
ADDITIONAL SLIDES 25
Policy-relevant apportionment of risk estimates (policy-relevant versus background) Background sources Policy-relevant sources Ambient air “Recent air” Indoor dust • newly emitted lead • resuspension of historically emitted and deposited lead Indoor dust “Past air” Outdoor soil • historically emitted and deposited lead • paint • Diet • Drinking water paint Total risk = recent air pathways + past air pathways + background pathways • The risk assessment simulates attainment of alternate NAAQS by reducing recent air exposures. • In fact, attaining alternate NAAQS could also involve reduction of past air 26 exposures (e. g. , historically emitted and deposited lead).
Conceptual framework for risk assessment - Extra Location-specific urban case study Census block #1 % of pop Census block #n Blood Pb levels (ug. d. L) % of pop Blood Pb levels (ug. d. L) Points of IQ loss Populationweighted aggregation 27 % of pop 100 children Points of IQ loss 900 children
Modeling Approach: Characterizing ambient air Pb levels, inhalation exposure air concentrations, and background (diet and drinking water) concentrations Media category General urban case study single ambient air Pb level assumed across entire study area (mean values from urban areas with > 1 million people). Location-specific urban case study US Census block groups within study areas assigned to nearest TSP monitor (point source and nonpoint source monitors handled differently). ¢ 6 to 11 exposure zones depending on location. Primary Pb smelter case study Ambient air Pb levels ¢ Inhalation exposure air concentrations National Air Toxics Assessment (NATA) – derived ratios of modeled Pb air exposure levels to ambient air Pb levels. The average ratio for the overall NATA analysis was used. NATA-derived ratios estimated for set of relevant US Census tracts Outdoor soil Pb levels Arithmetic mean from HUD data set intended to characterize residential soil Pb levels across houses constructed between 1940 and 1998. Site-specific post-remediation soil Pb measurement data (for subarea) Dietary Pb levels Based on (a) Pb food residue data from US FDA Total Diet Study (2001) and (b) food consumption data from NHANES III (CDC, 1997) Drinking water Pb 28 levels Geometric mean of values reported in studies of US and Canadian populations (residential water). ¢ ISC-PRIME dispersion modeling (NAAQS attainment scenario) used to estimate centroid levels for US census blocks and block groups. ¢ 22 US Census block groups and 115 blocks. ¢
Modeling Approach: Specification of CR Functions for IQ Loss - Extra l Log-linear function • • l n = 1, 333 Median concurrent BLL 9. 7 μg/d. L β = -2. 70 (95% CI: -3. 74, -1. 66) Estimate IQ point decrement: 3. 9 points for BLL 2. 4 to 10 μg/d. L; 1. 9 for BLL 10 to 20 μg/d. L Dual linear stratified at peak BLL 10 μg/d. L • n = 244 • GM concurrent BLL 4. 3 μg/d. L • β = -0. 80 (95% CI: -1. 74, 0. 14) for <10 μg/d. L β = -0. 13 (95% CI: -0. 23, -0. 03) for ≥ 10 μg/d. L l Dual linear stratified at peak BLL 7. 5 μg/d. L • n = 103 • GM concurrent BLL 3. 2 μg/d. L • β = -2. 94 (95% CI: -5. 16, -0. 71) for <7. 5 μg/d. L β = -0. 16 (95% CI: -0. 23, -0. 08) for ≥ 7. 5 μg/d. L 29
Скачать презентацию Pb NAAQS Human Health Risk Assessment Overview
b0d46b9f57e50e13dae54af0a94d3da1.ppt