Office of the administrator science advisory board

Dr. Philip Goodrum

Section 4.5.1 was fairly well organized, and it was particularly useful to begin with reiterating the analysis and findings of Brunekreef et al (1984) based a meta-analysis of multiple studies. It is not obvious upon first reading how the slope term was calculated from the parameters of the log-log regression. Several short examples should be added:

1. 
   Brunekreef et al (1984) - the equations presented in the title of Figure 4-19 should be carried over to the main text and example calculations should be provided. There are errors in lines 9 and 11 on page 4-79 – the slopes presented for both data groups are based on an increase in air Pb concentration from 0.50 to 1.5 µg/m³, rather than 0.15 to 1.5 µg/m³. Text and tables throughout the section should be closely reviewed for similar typographical errors.
   

2. 
   Table 4-11: I’d suggest re-orienting the table to landscape view, and expanding the “Model Description” from 1 column into 3, with the following headers: Model (e.g., log-log), Parameters (provide BOTH slope and intercept terms), Description (the remaining information)
   

3. 
   Table 4-11: Slope calculations – references to footnotes suggest that either the slope was calculated by integration, or more likely some interval of air lead concentrations was used. It is not clear how the slopes were calculated, nor can the values presented be readily identified from the cited literature. In the text or table footnotes, provide a few additional example calculations to demonstrate how the slopes were determined.
   

It is unclear if the increase in blood lead concentration that is referred to in discussion of the slopes is actually predictive of a change in the geometric mean (GM) or arithmetic mean. Given the reliance on log-transformed variables in the regressions and based on a cursory review of the primary literature, it appears more supportable to interpret these deltas as changes in the GM. A short discussion of the difference between arithmetic mean and GM for lognormal distributions should be added and, if there are differences in interpretation – either provide both the AM and GM, or describe how the conversion to GM was conducted.

• 
  At the relatively low blood lead concentrations under consideration, it is likely that lead in diet contributes a substantially higher fraction of the average daily lead intake than lead in air (even after accounting for fate and transport processes that suggest air Pb contributes some fraction to soil and dust Pb). Present the FDA dietary exposure estimates and show how changes in lead in diet correspond to changes in population parameters of the blood lead distribution. Also present the assumptions in the IEUBK and ALM models regarding relative contributions of diet, water, air, soil, and dust as well as the air-to-dust partitioning assumptions.
  
• 
  The graphical summary of the regression models used to define the blood-air slopes should be accompanied by a more in-depth overview of the information provided: 1) overall variance explained by the models used to derive the slope (e.g., what is the importance of a low coefficient of determination?); 2) range of blood lead concentrations in the empirical study – particulary if it is representative of individuals with relatively high blood lead concentrations; 3) consistency of slope from each new study with comparable studies presented previously (i.e., is the new study within the same range?); 4) relevance of the new slope estimates the key decision points regarding the standard. As noted in the graphics below, these decision points include: a) the change in the GM and 95^th percentile blood lead concentrations that are considered significant at the population level; and b) the plausible range of slope factors that could trigger an exceedance of the thresholds established in (a).
  
• 
  While historically, much attention is given to predicted (and measured) blood lead concentrations in the fetus and developing child, older age groups also warrant concern because the association between average air Pb concentrations and health outcomes may be stronger for older age groups, and available studies / empirical data on older age groups can be more directly included in the assessment. Currently, the assessment is restricted by the tools that are publically available. Specifically, the IEUBK and ALM models restrict the assessment to data on children and adults of child-bearing age. A concerted effort to finalize the All Ages Model would break down this barrier and allow for a more comprehensive evaluation of lifetime exposures to be conducted.