Document ymDBmdZV002kew6BrKoJJNEvE

PJMM4U& MEV. mmatnitti E. i. d u Po n t o e Ne mo u r s & Co mp a n y HtcowaiiATto Pe t r o l e u m La b o r a t o r y WILMINGTON. DELAWARE l*8Sa Te l e p h o n e Ar e a Co m COMSMMB 540-2874 December 5, 1983 Professor Ralph A. Bradley 325 Hickory Hill Drive Watkinsville, GA 30677 Dear Professor Bradley: Accompanying this letter are portions of a draft report on analysis of the NHANES II blood lead study, for your review prior to our meeting December 15 in Atlanta arranged by Ben Fort. The statistical analysis is primarily the work of Dr. Charles G. Pfeifer of the Du Pont Engineering Department. Dr. Pfeifer will attend the meeting on December 15, as will Dr. Ronald D. Snee, also of the Engineering Department. This draft material is intended to provide sufficient detail for your critique of the approach we have adopted, and its execution. Should you note any critical omission, please call me or Chuck Pfeifer (302-366-3540). I am also enclosing a copy of a report (PLMR 58-83) which presents part of our analysis in less detail than in the draft material. Perhaps it will be useful to give you an overview of our line of argument. I look forward to meeting you on the 15th. Very truly yours, N33844 JMP/er Enc. gg: Snee C. G. Pfeifer D. Lynam - Ethyl B. Fort - Ethyl BETTER THINGS FOR BETTER LIVING . . . THROUGH CHEMISTRY 0532679 N33844.01 TEH 0532680 DUP050033950 TEH 0532681 DUP050033951 A j TEH 0532682 DUP050033952 TEH 0532683 DUP050033953 ---- 7vTWt^ fob H TEH 0532685 DUP050033955 DUP050033956 f DUP050033957 TEH 0532688 DUP050033958 . ........ .............. Ia a J o^. \ ji y TEH 0532689 DUP050033959 TEH 0532690 DUP050033960 ANALYSI S OF NHANES I I BLOOD LEAD STUDY PLMR-58-83 J. ML PIERRARD PETROLEUM LABORATORY E. I. DU PONT DE NEMOURS & CO. SEPTEMBER 21, 1983 TEH 0532691 DUP050033961 N33844.03 The U.S. Government * s second National Health and Nu trition Examination Survey - known as NHANES II - was de signed to assess nutritional status of the population based on questionaires and examinations administered in 64 sample areas over a 4 year period. As an added part of the study, venous blood lead measurements were made on 9936 persons (1) The study was designed for cross-sectional analysis that is, to obtain a value for each measured variable repre sentative at the study midpoint - and not for chronological analysis. However, an apparent blood lead decrease of 37% over the 4 year study period prompted attempts to analyze the time trend (2). TEH 0532692 DUP050033962 NHANES H Assessment of National Health Status February 1976 to February 1980 64 Sampling Sites 6 Months to 74 Years of Age 27,801 Persons Selected -- 20,322 Examined 9,936 Venous Blood Lead Levels Determined Not Designed for Chronological Trend Analysis Apparent 37 Percent Reduction in Blood Lead DUP050033963 Attention was drawn to this representation of the par allel declines of six month average blood lead and total U.S. gasoline lead use (3). This figure might well have led some to infer that the blood lead decline was solely due to the decline in total gasoline lead use. TEH 0532694 DUP050033964 Gasoline Lead and NHANES n Blood Lead Levels (Feb, 1976 -- Feb. 1980) > cti t3o CD 2 oo a CD Q. CD S' TEH 0532695 DUP050033965 Once the basic data became available, further exami nation showed that there is substantial confounding because of npnuniformity of subject characteristics over the 4 year course of the NHANES II study. This figure shows one example of such confounding (4) These considerations resulted in a number of more de tailed analyses, including our own. To contrast the fea tures of the various analyses, it is helpful to consider the identifiable factors which might affect apparent blood lead. TCU Arrtn/'AA DUP050033966 Average Blood Lead Levels (//g/dl) Urban Dwellers Sampled and NHANES IX Blood Lead Levels {Feb. 1976 -- Feb. 1980) TEH 0532697 DUP050033967 It is generally agreed that exposure to lead may occur via a number of routes. In several analyses of the fJHAN.ES II data conducted or sponsored by government groups, it was assumed that only exposure to lead from gasoline decreased during the study period (5,6,7). Their assumption that food lead was un changed was justified on the basis of calculated lead content for a constant market basket of foods (8). However, 64$ of the U.S. households surveyed in 1979 by the Economics and Statistics service of the U.S. Department of Agriculture reported diet changes during the period 1977 to 1979 (9). Furthermore, alumi num cans were displacing soldered cans during this period (10). Therefore, we made provision in our analysis for possible changes due to these or other unquantifiable factors. Returning to gasoline lead, it should be noted that the other analyses referred to used total national gasoline lead use as the index of exposure. let, the 64 sites visited dur ing NHANES II had a wide range of lead exposures as shown by air lead data. To account for this variability, we calculated the site-specific gasoline lead density for each location vis ited. Each site consisted of one or more counties. The gaso line lead consumed in the site county at the time of the visit was calculated from state gasoline consumption and lead dosage, and county and state populations, using the fact that gasoline consumption is highly correlated with population in most of the U.S. (11). The site-specific lead density then was calculated by dividing county lead use by county area. This approach allows for exposure to gasoline lead by ingestion as well as inhation. The study design factors listed in the chart were treated as classification variables. Caravan itinerary covers region of the country and season. Personal variables are sex, age, and race, and residence variables are size of city or town and in or outside of the core city of a Standard Metropolitan Sta tistical Area. Quality control data on the laboratory analysis of blood leads for the last 3 years of NHANES II do not indicate any significant problem. However, there were no check samples car ried through the sequence of field collection, storage, ship ment and final analysis at the home laboratory, so we allowed for a possible trend contribution from this source. TFH 0532698 DUP050033968 Identifiable Factors Potentially Influencing Blood Lead Trend Exposure Air Lead from Gasoline Food and Beverages Water Paint Dust Study Design Caravan Itinerary Personal Variables Residence Characteristics Study Execution Laboratory Analysis Sample Handling TEH 0532699 DUP050033969 We investigated the time trend contribution attributable to the demographic factors alone, for demographics and gasoline lead, and for demographics, gasoline lead and time, which was used as a surrogate for unquantifiable time-related blood lead effects* Details are given in our paper (12) which is appended to this report. Of particular interest is this comparison of national gaso line lead use and site-specific lead density. National gasoline use is not statistically significant as a blood lead predictor in the presence of the time surrogate for other time-related blood lead effects. On the other hand, site-specific lead den sity and its logarithm retain their significance. TEH 0532700 DUP050033970 Significance of National Gasoline Lead and Site Specific Lead Density Gasoline Lead Exposure Variable National Site Density Log (Site Density) Significance Level* Lead Exposure Variable Time 0.8850 <0.0001 0.0022 < 0-0001 0.0019 <0.0001 All models included the full set of 30 demographic terms. TEH 0532701 DUP050033971 The histogram shows the frequency distribution of the correlations with time of the site-specific lead densities at each of the NHANES II sites over the 4 years of the study. The arrow at A shows that total national gasoline lead use is highly correlated with time. But, as indicated by the arrow at B, the site-specific lead densities at the sites when they were actually visited is poorly correlated with time. Thus, national gasoline lead use not only fails to re produce the time history of lead exposure as the caravans moved from site to site, but it also obscures the influence of other potential time-related effects because it is highly correlated with time. TEH 0532702 DUP050033972 Correlation of Gasoline Lead Density and Time Correlation Coefficient TEH 0532703 DUP050033973 Using site-specific lead density as the measure of ex posure to gasoline lead, stepwise regression analysis showed that Over half the apparent blood lead decrease can be accounted for by changing demographics. About one-tenth of the apparent blood lead de crease can be accounted for by gasoline lead exposure. A significant time trend still remains after accounting for demographics and gasoline lead exposure. TEH 0532704 DUP050033974 Contributions to Blood Lead Decrease 1976-80 DUP050033975 To complement the stepwise regression analysis reported in (12), we analyzed the NHANES II blood lead data by the method of split-plot covariance analysis (13)* This method takes its name from agricultural applications, where a field or "plot" is "split" and planted with two varieties of plant. To validly compare performance of the two varieties grown un der the same conditions "within-plot" corrections - such as for different numbers of the plant varieties - must be made. To compare performance of the varieties in different fields, or plots - say under different climatic conditions - "betweenplot" corrections are required. The situation is analogous for the NHANES II blood lead data set. Here, corrections are required for within-site differences in the characteristics of the subjects. The within-site analysis adjusts the raw observed blood lead value of each individual at a site to the value he would have if his personal and residence characteristics were the average values for the site. The between-site analysis amounts to an adjustment of the site-adjusted blood leads to the value each would have if every site had the same value of each of the personal, residence and itinerary variables. After these adjustments all observations are on a common basis relative to demographics, and the remain ing time trend is attributable to lead exposure factors and/or other trend influences. By omitting or including site-specific lead density in the adjustment process, we were able to estimate the contribution of gasoline lead to the blood lead decrease. TEH 0532706 DUP050033976 Steps in Blood Lead Trend Analysis Adjust for Differing Characteristics of individual Subjects Further Adjust for Differing Site Characteristics -- With and Without Gasoline Lead Quantify Unexplained Time Trend Compare Unexplained Time Trend With and Without Gasoline Lead Exposure TEH 0532707 DUP050033977 The split-plot analytical approach was applied to 216 models for blood lead as listed. Ah important result is that use of the: NHANES II blood lead weighting factors, or the basis of residual weighting, or the transformation of the dependent variable blood lead did not affect the overall results, which were consistent throughout. TEH 0532703 DUP050033978 Analyses for Residual Time Trend Factor BPb and In BPb Pf R, I and 26 combinations of current and lagged site specific gasoline lead density and their logarithms With and without BPb weighting factors Residuals weighted according to 1/S2 and N Options 2 27 2 2 Number of models = 2x27x2x2 = 216 Consistent results regardless of BPb weights, residual weighting,, and BPb or In BPb TEH 0532709 DUP050033979 Because the results were consistent, the overall picture can be grasped by examining a few cases. For untransformed blood lead as the dependent Variable, before adjustment to remove any effects, the 4-year decrease is 5.8 ug/dl. After within-site and between-site adjustment for itinerary, personal, and residence (DEM) variations, this is reduced to about 1.6 /ig/dl, with r |=0.53. Including current site-specific lead density (G Pb) or its lagged values (G Pbgo - 30 days lag, G Pbijc - 45 days, lag) in the model accounts for up to 0.5 yug/nP of tne 4-year decrease. Note that the addition of lagged lead density terms after the first one has little effect on either the blood lead decrease attributable to gasoline lead or on R?. Also note in the last two lines that making the adjustment for gasoline lead before any adjustment for demographics reduces the 4-year decrease by less than 1 yug/dl, but with R? of only 0.20 to 0.21. TEH 0532710 DUP050033980 Typical Model Results* Effects Removed None DEM DEM. GPb DEM, GPb. GPb30 DEM. GPb. GPb3a GPb4S DEM. tog GPb DEM, log GPb, log GPb^ DEM. log GPb. log GPb30. log GPb45 GPb log GPb Blood Lead, ug/dl Remaining Pour Year Decrease $.80 Attributable to Gasoline Lead ; j r !_ 1.62~ 0.53 1.27 0.35 0.59 121 0.41 0.59 1,19 0.43 0.58 1.48 0.14 0.57 1.10 0.52 0.60 1.09 0.53 0.59 5.17 0.20 4.93 o.2i Dependent variable 0Pb. residuals weighted by t/ST no BPb weights TEH 0532711 DUP050033981 Prom the site-by-site analysis (12) and from the additional split-plot analysis of covariance on the individual observations just discussed, we conclude that national gasoline lead use is not a valid index because it does not reflect the exposures due to lead use at the NHANES II sites at the times they were sampled. Further, Confounding of total national gasoline lead use with time precludes estimation of other unquantifiable, time-related effects. We found that over half of the apparent blood lead decrease is traceable to changes in site subject demographics. From a large number of models we estimate, that 5 to 10% (or 0.3 to 0.6yug/dl) of the blood lead decrease is attri butable to gasoline lead exposure. This estimate is consistent with the existing literature. It agrees with,the estimate based on an air lead decrease of about 0,3 jug/ra3, an dC value in the range 1 to 2 (14,15). It is also consistent with the Frankfurt study (16). Finally, after adjustment for both demographics and gasoline lead exposure, a significant blood lead time trend remains. TEH 0532712 DUP050033982 Conclusions National gasoline lead use is not a valid index of exposure at NHANESII sites Over 50% of the apparent blood lead decrease is attributable to demographics 5 to 10% of the apparent blood lead decrease is attributable to gasoline lead exposure The blood lead decrease of 0.3 to 0.6//g/d! attributable to gasoline lead is consistent with the observed air lead decrease A significant blood lead time trend remains after accounting for demographics and gasoline lead exposure TEH 0532713 DUP050033983 REFERENCES 1. United States. National Center for Health Statistics. Plan and operation of the second National Health and Nutrition Examination Survey, 1976-1980. Prograins and Collection Procedures, Series 1, No. 15. 2. United States. Center for Disease Control. Morbidity and Mortality Weekly Report 30, 132 (1982). 3- Houk, V. Statement at Environmental Protection Agency Hearing on Regulation of Fuel and Fuel Additives, Lead Phasedown Regulation, Notice of Proposed Rulemaking, April 15, 1982. 4. Petroleum Chemicals Division, E. I, du Pont de Nemours & Co., Inc. Statement presented to Environmental Protection Agency, Science Advisory Board, Environmental Health Committee at an Open Meeting on Public Health Impacts Associated with Lead in Gasoline Phasedown Program, July 7, 1982. 5. Schwartz, J. U.S, EPA Memorandum "Health Effects of Gasoline Lead Emissions", May 11, 1982. 6. ICF Incorporated "The Relationship Between Gasoline Lead Usage and Blood Lead Levels in Americans: A Statistical Analysis of the NHANES II Data", Report to EPA, December, 1982. 7- Pirkle, J.L. U.S. CDC Memorandum "Comments on the Dupont , .(sic) Analysis of the NHANES II Blood Lead Data", December 1 1982 8. Jelinek, C.F, "Levels of Lead in the United States Food Supply", J. Assoc. Off. Anal. Chem. 65, 942-6 (1982). 9. "Growing Concern for Nutrition", Food Processing, p.40, March, 1981. 10. "Beverage Can Shipments", Modern Metals, p. 34, July, 1980, 11. Pierrard, J.M., et al, "Vehicle Emissions Controls and Ambient Air Quality", SAE Australia, Jubilee Year Conference, Melbourne (1977). 12. Pierrard, J.M. et al, "Assessment of Blood Lead Levels in the U.S.A. from NHANES II Data", Proceedings of the Inter national Conference on Heavy Metals in the Environment, p. 421-4, September, 1983. TEH 0532714 2 *. 13- Kempthorne, 0. Design and Analysis of Experimentst John Wiley and Sons, New York, 1952. 14. U.S. EPA Office of Air Quality Standards. National Air Quality and Emissions Trend Deport, 1981. 15. Snee, R.D. "Evaluation of Studies of the Relationship Between Blood Lead and Air Lead", Int. Arch. Occup. Environ. Health 4j8, 219-42 (1981). 16. Sinn, W. "On the Relationship Between Lead in Air and Blood Lead Content of Persons Living and Working in the Centre of a City (Frankfurt Blood Lead Study)", Int. Arch. Occup. Environ. Health 47, 93-118 (1980), 48; 1-23 (1981). .j. ' x- V v TEH 0532715 DUP050033985