Document NG6o96z7gN4XGyMJnwwNVom2D

s DEPARTMENT OF HEALTH A HUMAN SERVICES T Public Health Service Centers for Disease Control March 18, 1983 David Weil Ph.D. EPA Annex Building; Alexander Drive Research Triangle,, North Caroline 27711 Dear Dr. Weil, Enclosed are three documents for the NHANES II Time Trend Review Panel: 1) Duplicate measurements differing by more than 7 ug/dl (in response to questions from Dr. Smith) 2) Appendix M: Tabulations by Demographic Variables (at the request of Dr. Smith) 3) Appendix N: Regression Analysis of Urban and Rural Population Subgroups I hope this information will be of assistance to you and the panel* James L. Pirkle M.D. Ph.D. Clinical Chemistry Division Center for Environmental Health Centers for Disease Control TEH 0533056 Implicate measurements differing by more than 7 ug/dl Vhen duplicate lead measurements of a whole blood sample differed by note than 7 ug/dl, the analysis of the sample was repeated (again in duplicate) and the results of the repeat analysis were used. Experience with the modified Delves cup technique for measuring lead in whole blood indicated that if duplicate measurements differed by more than 7 ug/dl, it vas likely that an unusual, non-random analytical error had occurred. Example* of such errors are: microclots in the whole blood sample, slight misalignment of the Delves cup, uneven distribution of the blood inside the cup and physical deterioration (pitting, hairline cracking) of the cup. These errors can cause a large variation between duplicate measurements. For example, if a microclot (a small aggregate of red blood cells) was dispersed in the duplicate analysis of a whole blood sample, it would contain so many more red blood cells than a typical sample that the duplicate measurement could artifaetually be as much as 2 to 3 times higher (e.g. 15 ug/dl to 45 ug/dl). The other sources of error listed could produce changes as large as 50-10OX. The cutoff of 7 ug/dl was selected to detect any lead measurements which might potentially be affected by such errors. Duplicate measurements not meeting the 7 ug/dl cutoff were not included in the final lead concentration calculation in order to insure that the reported lead measurement was not influenced by any of these atypical analytical errors. For the sake of detail, an analysis was performed to determine what would have happened had these values not been excluded from the final lead concentration calculation. The analysis is meant only to be illustrative since measurements are included which would have been rejected on analytical grounds as described above. Out of the approximately 10,000 whole blood lead samples, 458 were repeated because duplicate measurements differed by more than 7 ug/dl. Thirty-six of these 458 had duplicate measurements which differed by more than 25 ug/dl and clearly were cases of a microclot, pitted cup, alignment problem etc. The remaining 422 samples had a difference TEH 0533057 DUP050034327 between duplicate determinations greater than 7 ug/dl either from one of the previously mentioned aoureea (i.e. microclot, cup alignment, etc.) or from random sources of error typically associated with analytical procedures (e.g. pipetting, instrument noise, etc.). A new lead value was calculated by averaging the duplicates which differed by more than 7 ug/dl with the duplicates of the repeat analysis. Thus four values were averaged together* t4 o determine this new lead value. The difference between this new lead value and the reported value is plotted versus time in Figure 1 with a linear regression fit and a 951 confidence interval about the mean. Table 1 gives the linear regression results. The plot and the regression analysis indicates that in samples that had duplicates greater than 7 ug/dl (about 4.51 of the 9937 NHANES II lead measurements), reporting the repeat measurement (again done in duplicate) did not contribute to a downward trend in the NHANES II blood lead data over time. Table 1 Regression of difference in calculated new lead value and reported lead value over time (see text for details) ~ DEP VARIABLE: MFF SOM OF SOURCE DF SQUARES MODEL I 50.480216 ERROR 420 , 3921.905 C TOTAL 421 3972.386 ROOT MSE 3.055793 DEP MEAN 1.080569 C.V. 282.7949 VARIABLE DF PARAMETER ESTIMATE INTERCEP 1 TIME 1 0.605344 0.025015 MEAN SQUARE 50.480216 9.337870 R-SQUARE ADJ R-SQ F VALUE 5.406 0.0127 0.0104 PROB>F 0.0205 STANDARD T FOR HO: ERROR PARAMETER**) 0.252791 0.010759 2.395 2.325 PR0B > JTJ 0.0171 0.0205 0533058 DUP050034328 **-iDX.rM* _ , 3emo> D~y. c skS O ws3 wm gc;is OCD ID tf) Ul 9T1 zi--n: <C luzo k T; x u j l u P> iCuCsIUc c ., tciDoccsmeccucej <rj wets: rt! fJ e usj: j 03 u 525 fcq OS 03 P a 99 *9 I r* 9 *ai *4 ! + n+ t } J ii + ; t;it + *+ 4 ts! 11 ++ + i'i *+ili * *f *+ .,If+ + + 5 +**!r #tHHJtI<U{i:4T*&4+x*tVT4* I ++v +V* i ht +% ** #+ *1 it I* ++ ++ *% 4* ++ 4+ * if * ++i+t. *4 & J*j*.+ - l in oin in sr - Utof - oto UJ - in : (v> o tv - in *o - in 1 n|nTn|rm <|rn 'Tfr rnryr 1111 -rp (-V* --CO in (VI CD to (O 0r nr[i mrTyn'i i i'| i"i i > 'p' CO * (IO OI* wM I 0*-*U.U.LJCCUJZOUJ -o in DUP050034329 n*?33059