Summary

Military and civilian workers at the U.S. Department of Defense (DoD) can be exposed to airborne lead at firing ranges and in other occupational settings. In 1978, the Occupational Safety and Health Administration (OSHA) set the current permissible exposure limit for airborne lead at 50 µg/m³. OSHA considers a blood lead level (BLL) of 40 µg/dL to be the upper acceptable limit to protect workers from adverse health effects. During the decades since OSHA set that limit, substantial scientific evidence pointed to a diversity of health effects associated with BLLs at less than 40 μg/dL. Based on that evidence and recommendations provided in 2013 by the National Research Council, DoD sought to identify BLLs lower than 40 µg/dL for worker removal and return to work. In addition, DoD sought to derive an airborne lead concentration corresponding to the BLL targeted by DoD management. That concentration is referred to as an occupational exposure limit (OEL), which is intended to represent the maximum contaminant concentration in the workplace that is intended to limit exposure concentrations and protect worker health.

DoD arranged for the use of a biokinetic model (also referred to as a physiologically-based pharmacokinetic [PBPK] model) to support the development of an OEL. Biokinetic modeling provides a mathematical technique for estimating absorption, distribution, metabolism, and excretion of chemicals, including particles and metals, in humans. Such models can be used to relate the amount of lead external exposure to the amount of lead found in the blood and other tissues at different points in time. DoD used a modified version of the O’Flaherty biokinetic model (referred to as the DoD-O’Flaherty model) to derive airborne concentrations of lead that correspond to BLLs for consideration by DoD management in establishing an updated OEL to replace the permissible exposure limit set by OSHA.

DoD requested that the National Academies establish an expert committee to evaluate whether the DoD-O’Flaherty model used to derive airborne lead concentrations from BLLs was appropriate. The committee was asked to consider whether an appropriate model was chosen, whether DoD’s modifications to the model were appropriately justified, and whether the assumptions in and inputs to the model were reasonable. The committee was asked not to recommend specific OEL values.

As part of carrying out its task, the committee was asked to provide an overall summary conclusion on DoD’s selected approach and the application of the approach for derivation of lead OEL values. It also was asked to address the following specific topics:

• Were the DoD-O’Flaherty model selection, parameterization, and validation appropriate, given the intended purpose—to develop OELs for DoD civilian and military workers?
• Were the inhalation rates used within the DoD-O’Flaherty model appropriate to represent DoD workers (military and civilian) who are occupationally exposed to lead?
• Were background levels of lead in air appropriately accounted for within the DoD-O’Flaherty model and representative of DoD workers who are occupationally exposed to lead?
• Is particle size variation appropriately accounted for within the DoD-O’Flaherty model and representative of lead absorption within the DoD workers (military and civilian) who are occupationally exposed to lead?

In its evaluation of model appropriateness, the committee considered questions posed in its Statement of Task and additional questions it selected from those commonly considered in reviews of PBPK model

aspects. Elements of a biokinetic model that had the greatest impact on the predicted relationship between exposure concentrations of airborne lead and BLLs of adults received the greatest attention. The questions addressed by the committee were organized into four broad categories:

1. Was an appropriate model chosen?
2. Were structural modifications appropriately justified?
3. Were model assumptions and inputs reasonable?
4. Was the model application appropriate?

OVERALL CONCLUSION

The committee commends DoD for undertaking a very substantial, deliberative process to establish a lead exposure monitoring program intended to be more protective of its workers who are exposed to lead. The committee recognizes DoD’s leadership in applying an innovative approach for establishing an OEL for lead using modern biokinetic modeling to develop quantitative relationships between occupational exposure and BLLs.

Overall, the committee found that the DoD-O’Flaherty modeling approach and application to support the development of an OEL for lead is appropriate. Specifically, an appropriate model was chosen, modifications to the model were appropriately justified, and the model assumptions and inputs were reasonable. The model was confirmed and shown to be sufficiently consistent with experimental data. The committee’s conclusions resulting from its main considerations are summarized below. In addition, recommended ways in which DoD can improve the DoD-O’Flaherty model, its application, and documentation are also provided.

WAS AN APPROPRIATE MODEL CHOSEN?

DoD’s evaluation of lead biokinetic models focused on the Leggett+ and O’Flaherty models, which are used to estimate BLLs resulting from exposure to lead in environmental media.¹ Both models met criteria for having appropriate compartments or processes for describing lead biokinetics, addressing the essential exposure routes, handling background lead exposure and occupational lead exposure, and calculating the corresponding lead dose-metric. The committee found that both the O’Flaherty and Leggett+ models described available BLLs with similar accuracy. Minor differences were cited by DoD as potential reasons for selecting one model over the other. However, the committee did not recognize that assessment as a basis for determining that either model would be inappropriate for use by DoD in developing a lead OEL.

Both the O’Flaherty model and Leggett+ model have been repeatedly utilized for more than a decade to calculate BLLs, with some modification by individuals using the model. Many of those applications have included separate reviews of the model’s appropriateness.

DoD selected and modified the O’Flaherty model to support the development of a lead OEL. The O’Flaherty model has practical aspects that fit the purpose of DoD’s modeling approach for supporting development of an OEL. For example, the model could be modified to facilitate probabilistic simulations of DoD worker populations. In addition, the model benefits from its treatment of birth date as a factor in historical exposures (e.g., dynamic background lead exposure).

The approach DoD used to select the model was reasonable and included consideration of the right models. The selection of the O’Flaherty model for use in developing an OEL for lead was appropriate and effectively justified.

___________________

¹ The Leggett+ model is a version of the Leggett model that was modified by the California Environmental Protection Agency’s Office of Environmental Health Hazard Assessment.

WERE STRUCTURAL MODIFICATIONS TO THE MODEL JUSTIFIED?

DoD modified the O’Flaherty model code published in 2000 to run sensitivity analyses, Monte Carlo analyses, and other numerical simulations necessary to support the development of an OEL. The committee considered whether those changes might affect the representation of physiological compartments or processes that either alone or together impact lead biokinetics. The committee also considered changes made for supporting model operations, such as Monte Carlo analysis.

Model changes made by DoD in preparing the DoD-O’Flaherty model were justified appropriately. Changes made after publication of the O’Flaherty model in 2000, including those made by DoD, did not alter the model’s representations of physiology or biochemical processes that would, in turn, affect representations of lead biokinetics relative to the model version described in 2000.

WERE THE MODEL ASSUMPTIONS AND INPUTS REASONABLE?

Model Consistency, Calibration, and Confirmation

As part of its Statement of Task, the committee was asked to consider whether the DoD-O’Flaherty model was appropriately validated. While the term validation is routinely used, DoD’s efforts involved activities best described as model calibration and model confirmation. Calibration is the manipulation of parameter values for independent model variables to obtain a match between the observed and predicted dependent variables. Confirmation is the process of examining the consistency between model simulation results and observational/experimental data. The greater the number and diversity of confirmations that indicate consistency, the greater the likelihood the model accurately reflects the system under study.

To assess model calibration and confirmation, the committee focused on assessing whether the processes and results of selecting parameter values for calibration and confirmation were appropriate, whether appropriate data were used for model confirmation, and whether the consistency of the confirmation outcomes was adequate.

The calibration and confirmation of the DoD-O’Flaherty model were sufficient to conclude that, in general, the inputs and assumptions in the model were reasonable. The consistency of simulated BLLs between the DoD-O’Flaherty model and Leggett+ model provided additional evidence of the reasonableness of the model inputs and assumptions in the DoD-O’Flaherty model.

Though a comprehensive check of the correspondence between the code implementation and the external documentation of the DoD-O’Flaherty was not conducted by the committee, it performed several checks when questions arose about the code, particularly regarding the Monte Carlo analyses, and the committee found that the documentation faithfully reflected the implementation in the code. However a comprehensive error check of the model code is an important aspect of developing a biokinetic model for regulatory application.

DoD should conduct and document an error check of the DoD-O’Flaherty model to assure there are no mathematical errors or errors in the code and equations, and that the model reasonably reproduces the analytic results published in the 2019 model support document.

Particle Size Variation and Absorption Factor for Inhaled Lead

Particle size is an important variable in lead biokinetic modeling because it is an important determinant of the percentage of inhaled lead that is transferred to the blood, which is represented by the inhalation transfer coefficient (ITC).

Studies of airborne lead particle-size distributions, both in firing ranges and in other workplaces show particle diameters ranging from ultrafine size (< 0.1 µm) up to about 80 µm. However, definitive studies of the ITC for lead for a range of particle size distributions and activity levels have yet to be conducted. DoD assumed a point estimate of 30% for the ITC.

The approach used by DoD to assign an ITC value was reasonable, given the absence of definitive studies of the ITC and the wide range of airborne particle sizes expected in DoD occupational settings. However, DoD should consider evaluating the evidence of a wider band of ITCs, including the use of a local sensitivity analysis that is focused on examining the sensitivity of the model output to a higher deposition rate. Evidence supporting a role for tracheo-bronchial absorption of lead would be one factor that could influence the ITC. Strong evidence of a wider range of ITCs would justify inclusion of this factor in the Monte Carlo simulations used to establish the OEL.

An important part of relying on measurements of airborne lead concentrations to estimate BLLs is to use measurement methods that reliably sample the inhalable particle size fraction of airborne lead. The 37-mm plastic cassette is the typical sampling method used in the United States and many other countries for measuring airborne lead concentrations.

The typical 37-mm cassette-sampling device can result in airborne lead measurements that underreport total inhalable lead. DoD should verify that the sampling method used to implement the OEL utilizes a sampling device that measures total inhalable lead and does not suffer from the limitations of the typical 37-mm cassette sample.

Background Airborne Lead Concentrations

DoD updated the previous estimates of background air concentrations of lead used in the O’Flaherty model to reflect recent measurements that would better represent the airborne lead concentrations occurring during the lifetime of the DoD cohort. The updated background airborne lead concentrations used in the DoD-O’Flaherty model were obtained from the most recent U.S. Environmental Protection Agency (EPA) Integrated Science Assessment for Lead.

The use of airborne lead concentrations from that science assessment is appropriate, with the qualification that, according to EPA, the concentrations are heavily influenced by source monitors in the network. Source-oriented monitoring sites (e.g., next to airports used by aircraft that use leaded aviation fuel) are required near sources of lead emissions that contribute, or are expected to contribute, to ambient air lead concentrations that exceed National Ambient Air Quality Standards.

Therefore, measurements from source monitors may not reflect airborne lead concentrations experienced by DoD workers living and/or working at a distance from those sources. Conversely, they may better represent exposures for those that live in proximity to such sources. A more spatially and temporally informed approach was not available to DoD.

DoD made an additional adjustment, using a population modifier (EXPOSMOD), to background lead exposures to assure that total variability in BLLs was consistent with the BLL population variability reported in the scientific literature. The application of EXPOSMOD jointly to the oral (dietary) and inhalation components of exposure was appropriate. However, because the BLL distribution of the general population has changed over time, the correspondence between the model predictions and measured BLLs (both central tendency and geometric standard deviation [GSD]) are also variable.² Therefore, a single value for EXPOSMOD may not accurately represent all years considered in DoD’s modeling approach.

Because dietary intake of lead tends to be the largest source of background lead exposure, estimates of the magnitude of dietary component can have a substantial effect on model estimates of non-occupational lead exposures. Previous versions of EPA’s Air Quality Criteria for Lead may provide evidence of lower dietary lead concentrations prior to 1980 compared to those currently used in the model.

In general, background concentrations of airborne lead are appropriately accounted for in the DoD-O’Flaherty model. However, DoD should consider the evidence for a lower or declining BLL GSD and further consider if different values for EXPOSMOD over time may improve the model performance and accuracy of predictions for current and future OELs.

___________________

² The modeled GSD is expected to directly influence the derived OELs.

In addition, DoD should consider reviewing the 1977 and 1986 EPA Air Quality Criteria for Lead to determine if using a lower dietary lead concentration for the pre-1980 background exposures would be more appropriate than those currently used in the DoD-O’Flaherty model.

Inhalation Rates

A key challenge for modeling DoD occupational lead exposure scenarios is to estimate long-term average daily lead intake via inhalation by using inhalation rates that adequately represent an expected range of activity patterns across the TriServices (U.S. Army, Navy, and Air Force). The committee considered two primary factors in evaluating the appropriateness of inhalation rates: (a) whether daily activity patterns were adequately represented, and (b) the strength of the underlying inhalation rate data for deriving distributions of inhalation rates. With respect to representing inhalation rates for daily activity patterns, DoD developed exposure scenarios that encompass activities of both typical workers and those who more likely engage in higher inhalation-rate activities. Under that approach, separate parameter values can be estimated for men and women. DoD intends for the DoD-O’Flaherty model to yield reasonable estimates of the relationship between exposure and BLLs to support the selection of an OEL intended to protect nearly all full-time military and civilian workers, including firing range personnel.

DoD’s approach is reasonable for estimating inhalation rates of a general worker population and the use of gender specific inhalations rates is appropriate. The inclusion of the 95th percentile is reasonable to account for the higher activity patterns of some workers in the population.

The committee considered the strengths and limitations of the underlying inhalation rate data used to derive the inhalation rate distributions for derivation of the lead OEL. EPA’s Exposure Factors Handbook was the primary source of data on inhalation rates. The handbook reports summary statistics (e.g., arithmetic mean, standard deviation, 95th percentile) grouped by age and gender. Overall, the data sources used to support inhalation rates for the model appear to be fit for purpose. The key studies are relatively current (published 2006 to 2009) and span survey years during the past 15 to 20 years. A major source of uncertainty of these data sources stems from the question of representativeness of the study populations (i.e., general worker populations) to the combination of military and civilian workers. It is conceivable that inhalation rates of military personnel are higher than average when they are engaged in strenuous activities. The extent to which the upper end of the distribution of inhalation rates proposed for derivation of the lead OEL adequately represents such high-end activity patterns of firing range personnel is unclear. This uncertainty may be offset to some degree by the inherent bias associated with the study protocols. Specifically, variability in inhalation rates measured during short periods is likely to be greater than variability in long-term average inhalation rates, which is the focus of DoD’s modeling exercise. That may mean that the high-end estimate of the probability distribution (truncated at ± 2 standard deviations) from a study used to establish inhalation rates for the DoD analysis likely exaggerates long-term average daily inhalation rates for some military and civilian staff.

The data sources and general approach for developing the probability distributions of inhalation rates are reasonable. However, DoD should consider conducting additional Monte Carlo simulations at the candidate OELs using a distribution of inhalation rates (and cardiac outputs) representative of personnel with higher activity levels, such as those that might occur on a firing range. A comparison of the resulting BLL distributions to those used to derive the OELs should be used to determine the fraction or percentile of DoD workers in a higher activity group that would have BLLs below each target level. The analysis would illustrate the sensitivity of the model to inhalation rates in alternative exposure scenarios and the influence of uncertainty in the inhalation rate on outcomes. It would also help risk managers understand the level of protection afforded individuals with inhalation rates higher than those used to derive the candidate OELs.

Correlation Between Cardiac Output and Ventilation Rate

Ventilation rate and cardiac output are inherently correlated. The committee identified two potential issues related to the independence of cardiac output and ventilation rate in DoD’s Monte Carlo analysis. First, if the Monte Carlo simulations included conditions where the expected ratio of inhalation rate to cardiac output was significantly violated, non-plausible physiological conditions could have arisen. The second issue has to do with the relationship between the inhalation rate and the glomerular filtration rate (GFR), which control the most significant rates of lead intake and elimination, respectively. The GFR is highly correlated with cardiac output, which is, in turn, highly correlated with inhalation rate. Changing inhalation rate, without corresponding physiologically accurate changes in cardiac output and GFRs, could establish unrealistic scenarios in which a lead dose rate increases but lead elimination through a correlated process decreases, instead of increasing. A main question is whether either issue would change the final distributions of BLLs for a given airborne lead concentration used to produce the final BLL distributions. The resulting BLL population distributions would then be in error. The committee notes that the coefficient of variation for the inhalation rate (0.2) may be small enough that perhaps there is little impact on the final BLL distribution from the ventilation rate-cardiac output correlation.

Varying cardiac output and ventilation rates may separately create non-physiological conditions in which a lead dose rate and renal clearance of lead do not increase and decrease together. DoD should explore the impact of correlated increases in ventilation rates and cardiac output on BLLs to determine if these parameters should be varied together, rather than independently, in the modeling of BLLs.

Model Documentation

Model documentation was spread among several documents, two technical reports, and the model code itself (which comprises many source-code files). This diversity of sources, style, and level of detail makes scrutiny of the mathematical and computational model rather burdensome. Though examination of the body of documentation permitted an evaluation of the model, it would have been highly desirable to have a single document that detailed the model structure, equations, parameters, and assumptions.

Documentation of the DoD-O’Flaherty model needs to be improved. DoD should prepare a support document for the DoD-O’Flaherty model in a manner similar to EPA’s documentation of the Integrated Exposure Uptake Biokinetic Model. In addition, the support document for the DoDO’Flaherty model should include:

• An illustrative figure representing the compartmental structure, blood flows, and mass transfers.
• Information contained in DoD’s response to the committee’s information request of 2019.³
• Documentation of an error check of the DoD-O’Flaherty model code, and assurance that the model reasonably reproduces the analytic results published in the 2019 model support document.
• Strategies that would allow the DoD-O’Flaherty model to be usable in the future because the model relies on software that is no longer supported by the company that developed it.

___________________

³ On June 27, 2019, the committee submitted a written request to DoD for information on the DoD-O’Flaherty modeling approach. The information topics included the DoD-O’Flaherty model structure, changes DoD made to the 2000 version of the O’Flaherty model, the basis for DoD’s estimated average removal duration for DoD workers, who exhibited elevated BLLs; DoD job activities that have the potential to result in lead exposure; modeled exposure scenarios; and approaches for selecting model parameters for the Monte Carlo analyses.

WAS THE APPLICATION OF THE MODEL APPROPRIATE?

In evaluating the overall approach and application of the DoD-O’Flaherty model for derivation of candidate OELs for lead, the committee considered the appropriateness of the model, the model assumptions and inputs, and several other factors.

In general, the committee agreed that the approach of using a biokinetic model to establish monitoring equivalent air concentrations representative of upper-bound BLLs is sound and well justified. The modeled population reasonably represented the worker population that DoD seeks to monitor and protect.

The assumptions and inputs to the model were largely considered appropriate. The approach considered variability in important exposure, physiological, and biokinetic parameters, including each in a Monte Carlo simulation producing likely distributions of resulting BLLs from which an OEL could be established. However, the committee observed that the results of the Monte Carlo analyses were not presented in a manner that gave the reader an appreciation for the prediction intervals or envelope. The results of Monte Carlo analysis would be more useful to the reader if they included mean values of measures with prediction intervals based on model uncertainty and variability/error in the data used for parameterization.

In carrying out its task, the committee’s overall conclusion is that the DoD-O’Flaherty modeling approach and application to support the development of a lead OEL are appropriate. The committee’s recommendations provide ways in which DoD can improve the DoD-O’Flaherty model, its application, and documentation.