Document B5KRgdGrEeN9maYDopMyZmgzX
Letters
EPA Review of Lead Study
scores. Some of these errors were not of
major magnitude but were indicative of
I would like to comment on the recent the care with which the report was as
article by Eliot Marshall (News and sembled and edited. A few errors were
Comment, 25 Nov., p. 906) dealing with serious, however, and questioned the
the dispute between myself and the En integrity of my data, myself, and my co
vironmental Protection Agency (EPA) workers. I was startled, upon finding
review panel headed by Lester Grant. these errors, to realize that three which
A subject as implant as the effects of appeared in the draft document had been
lead on the health ^f,children deserves pointed out earlier by me in written
careful scrutiny, and I have welcomed communications to Grant, on multiple
and received close examination of my occasions.
data by responsible scientists. Although The report stated that the rate of mis
I felt that the panel set up by Grant takes my colleagues and I committed in
circumvented the peer review process transferring data from the code sheets to
that EPA had established for the Lead the computer was 7 per 1000. The actual
Criteria Document, I welcomed them rate was lower by an order of magnitude:
and gave them unlimited access to my '7 per 10,000. This error appeared in the
data. Grant and I agreed that I Would be report despite the fact that I had brought
given a timely draft copy of their report it to Grant's attention in writing. It was
before its promulgation in order to cor also stated in the report that I did not
rect any errors offact. When the visiting furnish the distribution of lead levels
team arrived, on 30 March 1983, 1 pre across excluded subject groups, which
sented them with all of the bound print suggested that a source of bias could
outs of the study and an index to the have entered the study at this point.
data; I also provided them with a meet Actually, at Grant's request, 1 had sent
ing room. I met with them for two con him complete listings of all the dentine
secutive days to answer their questions. lead levels cross-tabulated by exclusion
At their request, certain portions of the ary class. These tabulations, which the
printouts were photocopied so that they report said I had not furnished, were
could work with the data sets individual cited in its bibliography!
ly. At the end of May, when the: report One member of tfie panel is quoted by
had not been sent to me, I called Grant's Marshall as saying that "[t]he basis for
office to ask for a copy of the draft. I was excluding certain children is still not
told it was being worked on and that 1 totally clear." Yet the reasons were de
would receive a copy as soon as it was lineated in our original paper (/), and the
ready. I continued to request the draft visitors were given copies of the tables
report at monthly intervals until Septem listing the excluded subject and the crite
ber without effect. On 30 September, at ria applied. Another panel member stat
the same time that a copy was sent to the ed that, because the computer program
EPA press in Cincinnati for publication, "kicked out" cases for which some co-
a copy was sent to me. On first reading, I variate data was missing, the actual sub
found at least 12 errors offact; almost all ject number for some of our analyses
were biased against the conclusions was less than 158, This is true. It is not
drawn by my study. By express mail, I true, as he stated, that because the N
informed Grant that, if the erroneous was reduced, the differences may not
report was not withdrawn, I would insist have been statistically significant. The
that a complete errata sheet be sent to program we employed (Statistical Pack
every recipient, with an acknowledg age for the Social Sciences) calculated
ment that my original agreement with the P values on the basis of the number
him had not been fulfilled. The report of cases actually analyzed, not the num
was withdrawn from the printers.
ber of cases in the data file.
The report incorrectly assigned den
The most serious criticism of my work
tine lead levels to my groups , incorrectly was the charge that I may have known
stated the type of dental tissue analyzed, the lead classification and outcome
incorrectly labeled the type of IQ test we scores of my subjects at the time I ex
employed, and incorrectly cited IQ cluded subjects from the study for rea
116
sons such as head injury. Had this been true, an important source of bias could have crept into the subject selection. My associates and I took particular pains to prevent this from happening. The report was incorrect, despite the fact that our blinding procedures had been brought to the attention of Grant and the panel on four occasions, twice in oral presenta tions and twice in written correspon dence. While this statement and the oth ers were withdrawn in the final docu ment, the conclusions drawn from them were allowed to stand.
Finally, Marshall indicates that, ac cording to Grant, I became uncooper ative when "problems turned up" and asked all of the reviewers to return ail the data before leaving his office. This is simply not true. Grant asked if I would allow them to take the photocopies I 'had provided them back to North Carolina. I asked him Whether their confidentiality would be shielded if I gave them to him. He said they would not. I said that, since these data represented work still in prog ress, they could not take them from the hospital but they could spend as much time as they Wished with the data in Pittsburgh.
Eight days after the panel visited me in Pittsburgh, I received a letter from Grant, in which he wrote:
Thank you for taking time out of a busy schedule to meet last week with myself and members of the expert committee that I have asked to review ydur studies on association between neiirpbehavioral effects and low lev el lead exposures in children. Your assistance in providing an introductory overview of your studies, in making computer printouts ofdata and statistical analyses results available for inspection by the committee, and in answer ing questions posed in regard to procedures employed in collecting the data and analyzing the results of your studies was quite helpful and appreciated by us, as was the assistance provided by your secretary during our visit.
It is not clear to me why there is a disparity between these two statements by Grant.
I have never claimed that my work is flawless. Epidemiology in the real world does not permit such hubris. It is,, how ever, more than a little frustrating to see refractory misinterpretations of fact and inference applied to it in the face of repeated efforts at correction,
He r b e r t L. Ne e d l e ma n
Bekayioral Science Division, Children's Hospital ofPittsburgh, 125 DeSoto Street, Pittsburgh, Pennsylvania 15213
References
1. H. L. Needleman et at.. New Engl. J. Med. 300, 68? (1979);_______ A. Leviton, 1?. Bellinger, ibid. 306, 367 (1982).
SCIENCE, VOL. 223
TEH 0532674