HOME
    SHOSHONE NEWS-PRESS, Feb. 9, 2002, p. 5.
    Common Sense

    Federal agency fails to paint a convincing picture on paint

    By Ron Roizen, Ph.D.
    SNRC Science Committee

    Lead-based paint is a well known source of childhood lead exposure.  There are many old houses in the Cd'A Basin, perhaps particularly in Wallace, where the entire city is listed on the National Historical Register.  Hence, the notion that old paint may be a significant source of elevation in CDA basin childhood blood-lead levels (BLLs) merits close scientific examination.


    TABLE 1:  Blood lead & paint hazard
    relationship, raw data
    10+ BLL 
    - YES
    (Col. A)    		 10+ BLL
    - NO
    (Col. B)    		 TOTALS
    paint hazard - YES
    (Row 1)    		 6 17 23
    paint hazard
    - NO
    (Row 2)    		 14 179 193

    TOTALS    		 20 196 216

    As it happens, the Human Health Risk Assessment report for the basin considered paint but downplayed its importance.  In the end, a paint variable or parameter was not even included in the IEUBK models used to predict childhood BLLs here.

    How was this downplaying justified in the HHRA?  The answer to that question is an object lesson in how ostensibly objective conclusions may be tilted in a desired direction.

    According to HHRA's text, 524 "observations" of childhood BLLs were available from blood-lead surveys conducted from 1996 to 1999, inclusive.   Among these same 524 "observations," 216 "observations" had accompanying house-paint assessments -- that is, evaluations of the lead-paint hazard in the interior of a child's home.

    Though the HHRA did not present a table displaying the bivariate relationship between paint and BLL, it is an easy matter to construct what such a table would comprise.  Table 1 presents the raw data for the EPA's 216 "observations" for whom both both BLLs and paint-condition assessments were available.

    Table 1 shows that a total of 23 "observations" had hazardous paint environments and a total of 20 "observations" had BLLs of 10 micrograms per deciliter or higher (ug/dL).  Some six "observations" had both hazardous paint and 10+ BLLs.

    The question is:  Do these data show a strong or a weak relationship between paint and BLLs?

    The data can be presented in different ways depending upon whether the analyst wished to downplay or play-up the strength of the paint-BLL relationship.

    For example, an analyst wishing to downplay the relationship could report that:

    • Only 30 percent of all "observations" with 10-plus BLLs also resided in dwellings with paint hazards;
    • That the great majority of "observations" with elevated BLLs -- 70 percent, to be exact -- lived in dwellings with no paint hazard; and, in short,
    • That most of the elevated BLLs did not derive from paint.
    Both the 30 percent and 70 percent statistics derive from vertically percentaging Col. A of Table 1 (see Table 1-V, Col. A, below) -- 6 divided by 20 equals 30 percent and 14 divided by 20 equals 70 percent.

    On the other hand, a quite different picture of the paint-BLL relationship is suggested when Table 1 is percentaged horizontally rather than vertically (see Table 1-H, below).  Percentages [sic] in the horizontal direction, it now becomes possible to assert, for example:

    • That whereas more than a quarter (26 percent, to be exact) of "observations" with a paint hazard had elevated BLLs, only 7 percent of "observations" without a paint hazard had elevated BLLs;
    • That the "observations" living in a dwelling with a paint hazard had a 3.7-times greater chance or risk of elevated BLLs than "observations" without a paint hazard (27 percent divided by 7 percent equals 3.7); or even

      • That "observations" in a dwelling with a paint hazard had a 370% greater chance of elevated BLL than "observations" without a paint hazard.
    Obviously, much hangs on whether Table 1 is percentaged vertically or horizontally, and much hangs on how those percentages are in turn presented in the report's text.  The HHRA's authors in general opted for the paint-downplaying message of vertical percentaging.

    But our story isn't quite finished yet.

    I have been using the awkward word "observations" to refer to these data because some children were measured more than once in the 1996-1999 series of blood-lead surveys.

    As it happens, paint-assessing data were collected only in the 1996 survey.  The 1997, 1998, and 1999 surveys collected only blood-lead data and no data on paint condition.  A total of 98 individual children contributed data in the 1996 survey.

    Yet there are a total of 216 "observations" employed in this paint analysis.  This implies that a majority of the "observations" in Table 1 were repeat observations taken in 1997, 1998, and/or 1999 -- in years when no associated paint data were collected.

    And that implication, in turn, contributes significant ambiguity to the analysis.

    Consider the following hypothetical scenario:

    Suppose that a child had both (a) a paint hazard and (b) an elevated BLL in 1996.  That "observation" would contribute one case to Row 1:Col. A of Table 1.

    Logically, the more "observations" that fall in the table diagonal defined by Row 1:Col. A and Row 2:Col. B, the stronger the paint:BLL relationship.  Conversely, cases falling on the diagonal defined by Row 1:Col. B and Row 2:Col. A suggest a weakened, nonexistent, or even negative paint:BLL relationship.

    Suppose, in turn, that the Panhandle Health District's follow-up of the 1996 survey resulted in advice to the parents of this child that the paint problem should be remedied, and they promptly carried out that remedy.

    Now the same child participates in the 1997, 1998, and/or 1999 survey waves -- and in all these subsequent waves no elevated BLLs are reported.  All these subsequent "observations" will fall in Row 1:Col. B -- that is, positive on paint hazard but negative on elevated BLL -- because no subsequent measure of paint condition was available in survey years 1997, 1998, and 1999.  Therefore, the 1996 paint measure will be attached to this child in subsequent surveys whether paint has or has not in fact been remediated.

    This [sic -- should be "These"] hypothetical subsequent "observations" in turn will tend to weaken the cross-sectional relationship between paint and BLL as pictured in Table 1 -- because 1997-1999 "observations" will fall in a cell that in effect says paint is not related to elevated BLL.

    The result is ironic because the post-1996 change of this child's blood level from elevated to non-elevated tends to support [the] importance of paint's contribution to elevated BLL.  But the cross-sectional mode employed in Table 1 and [the] absence of paint data for 1997-1999 in effect transform that post-1996 change into one or more "observations" tending to undercut the paint-BLL relationship.

    The above hypothetical scenario -- and other problematic scenarios -- might have been eliminated by restricting Table 1's analysis entirely to 1996 data -- in other words, including only data for a year in which both paint and BLL data were available.  This is a check the HHRA's authors did not provide in that document's text.  The absence of this sort of 1996-data-only check in the HHRA analysis is, in this writer's judgment, a glaring and telling omission.

    A footnote:  EPA retained an outside consultant, Paul Mushak, Ph.D., to make an independent assessment of comments to the draft HHRA.  At one point in Dr. Mushak's remarks he wrote:  "Everything that could be done to give lead paint a thorough assessment was done in the HHRA" (Mushak, Appendix W, CD p. 76).  Given the weaknesses in the bivariate analysis discussed above, Mushak's broad assertion suggests that both the HHRA's analysis and its review by hired consultants merit close scrutiny.

    TABLE 1-V:  Blood lead & paint hazard
    relationship, percentaged vertically
    10+ BLL 
    - YES
    (Col. A)    		 10+ BLL
    - NO
    (Col. B)    		 TOTALS
    paint hazard - YES
    (Row 1)    		 30% 9% 11%
    paint hazard
    - NO
    (Row 2)    		 70 91 89

    TOTALS    		 100%
    (20)    		 100%
    (196)    		 100%
    (216)
    TABLE 1-H:  Blood lead & paint hazard
    relationship, percentaged horizontally
    10+ BLL 
    - YES
    (Col. A)    		 10+ BLL
    - NO
    (Col. B)    		 TOTALS
    paint hazard - YES
    (Row 1)    		 26% 74 100%  (23)
    paint hazard
    - NO
    (Row 2)    		 7% 93 100% (193)

    TOTALS    		 9% 91 100% (216)

    "Common Sense" is a column by members of the Shoshone Natural Resouces Coalition's Science Committee.  Responses are invited from those who agree and those who disagree.