Document MxvZ9v22LKVMvgD8YbZk84y7

SUBJECT FROM TO UNITED STATES ENVIRONMENTAL PROTECTION AGENCY X' Docket Submission jO*-l ^C^.LAjarf Sr C- PC (' A W ^ ^ Central Docket Section (A-130) E1W1RONMENTAL PROTECTION AGENCY _! p.M Z O 19B ' CENTRAL DOCKET SECTION Please place the enclosed material in docket ^ Ob . i-he f'ptSi'-VX A r>-.-1 i'n U(&ci ^ bfocDz/ /TLs ians/oi/nsc//ar v^fsil_ *1/^ Ji. C c*jr to Vi f i C-' <- r V, cHC P'rMit , t~ n , Jr . rt_ / T,,u tfrs ^rrsr r /L^W,c- Jrffci_ .ai/f/arU* E PA F.--1 1320- (R*. 3-761 TEH 0350039 N36689 Ame r ic an Jo u r n al or Ep id e mio l o g y Copyright < J935 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 121, No. 2 Printed in ILSA THE RELATIONSHIP BETWEEN BLOOD LEAD LEVELS AND BLOOD PRESSURE AND ITS CARDIOVASCULAR RISK IMPLICATIONS JAMES L. PIRKLE,' JOEL SCHWARTZ,1 J. RICHARD LANDIS' a n d WILLIAM R. HARLAN* pirkle, J. L. (CDC, Atlanta, GA 30333), J. Schwartz, J. R. Landis and W. R. Harien. The relationship between blood lead levels and blood pressure and its RECEIVED ,, ENVIRONMENTAL PROTECTION ag en c y cardiovascularrisk implications. Am J Epidemiol 1985;121:246-58. The relationship between blood pressure and blood lead levels in the second National Health and Nutrition Examination Survey (1976-1980) has been exam ined for white males aged 40-59 years. After adjustment for age, body mass JAM^S 198* Index, nutritional factors, and blood biochemistries in a multiple linear regression model, the relationship of systolic and diastolic blood pressures to blood lead CENTRAL DOCKET SECTION levels was statistically significant (p < 0.01). There was no evidence of a threshold blood lead level for this relationship. Although these data alone do not prove a causal relationship between low blood lead levels and blood pressure, the findings are consistent with current epidemiologic and animal studies, indicating that a causal relationship is probable. To examine the potential health risks, the multiple logfstic'risk factor coefficients from the Pooling Project and Framingham studies were used to predict the impact of the 37% decrease in mean blood lead levels which occurred in adult white males from 1976 to 1980. As a result of this blood lead decrease, the calculations predicted a 4.7% decrease in the incidence \ of fatal and nonfatal myocardial infarction over 10 years, a 6.7% decrease In the incidence of fatal and nonfatal strokes over 10 years, and a 5.5% decrease in ,\ ` , the incidence of death from all causes over 11.5 years. In addition, as a result ot ^ this blood lead decrease, the predicted number of white mates in this age group with hypertension (diastolic blood pressure >90 mmHg) decreased by 17.5%. V blood pressure; cerebrovascular disorders; lead; myocardial infarction i High blood pressure is a major public large data base of information on persons health problem in the United States. Epi demiologic studies have examined sociode mographic, nutritional, and physiologic correlates of high blood pressure to under stand better the causes and associated risk factors for hypertension. National health surveys have been useful in providing a selected to be representative of the civilian, noninstitutionalized US population. The National Health and Nutrition Examina tion Survey I (NHANES I) was conducted from 1971 to 1975 and provided extensive information on blood pressure and a large number of related variables. Several anal yses of the NHANES I blood pressure data Received for publication August 30, 1984. Abbreviation: NHANES, National Health and Nu trition Examination Survey. 'Center for Environmental Health, Centers for Disease Control, Atlanta, GA 30333. (Reprint requests to Dr. James L. Pirkle.) 'Environmental Protection Agency, Washington, DC. ' University of Michigan School of Public Health, Ann Arbor, Ml. * University of Michigan School of Medicine, Ann Arbor, Ml. base have been published (1-4). The second National Health and Nutri tion Examination Survey (NHANES II) was conducted from 1976 to 1980 and mea sured blood pressure as well as a broad array of pertinent sociodemographic, nutri tional, and physiologic factors. Whole blood lead measurements were included in NHANES II, but not in NHANES I. Initial 246 TEH 0350040 N36689.01 DUP050298669 ..No:; UNA. OOD S LAN4 S n d d a ve ig ne im ad his ice the i in not oup >ersons ivilian, n. The dueled tensive a large il anal- Nutri- 1 meabroad nutriNVhole nitial LEAD AND FLOOD F r.ESM KE 247 multiple regression analyses of the blood which traveled to 64 different sites across pressure data from NHANES II (5) found the United States. The medical evaluation that lead was significantly associated with included medical history, physical exami systolic and diastolic blood pressures in nation, anthropometric measurements, di males 12-74 years of age. Lead remained etary interview (24-hour recall and food statistically significant even after adjusting frequency), laboratory tests, electrocardi for age, race, body mass index, and the ograms, and radiographs. A manual of op other variables which were significant at erations, special interview and examination the 5 per cent level. In addition, lead and protocols, and specifically trained inter age were partially confounded in such a viewers and examiners were utilized to way that the strength of the relationship maintain structured and standardized con between lead and blood pressure may have duct of the survey at all sites. Dietary in been underestimated. takes were determined by 24-hour dietary The present analysis of the NHANES II recall and questionnaires about frequency blood pressure data had three objectives, of of food consumption. The nutrient intake which the first was to examine the_relation- from the 24-hour recall was quantified for ship of blood lead levels to blood pressure each individual using a current nutrient irrFTestxlttgr>ubs,bup7 white males aged data bank. 40-5Syears, Over this age range, age effects Blood samples were analyzed by the Clin on blood pressure are small, permitting bet ical Chemistry Division, Center for Envi ter separation of the effects of lead and age ronmental Health, Centers for Disease on blood pressure. In addition, confining the analyses to white males obviates the confounding of sex and race effects. The second objective was to estimate conservativelv the strengtb^ndlfulMen^&nce ofthe association between lead and blood pres sure. The third objective was to"use the multiple logistic risk factor coefficients from the Pooling Project and Framingham studies to estimate the potential public health implications of the relationship be tween blood lead and blood pressure. Risk factor coefficients were available specifi cally for white males aged 40-59 years. Control (Atlanta, Georgia). Details of the methodology and quality control for the various analytes are available (7). Blood lead concentrations were determined by atomic absorption spectrophotometry by means of a modified Delves-cup micro method (8, 9). Cholesterol measurements were made by the George Washington LipidResearch Clinic Laboratory and referenced to the Abeil-Kendall method (10). Blood pressure measurements were made with an appropriate size cuff, following the recommendations of the American Heart Association. The first- and fifth-phase Ma t e r ia l s a n d me t h o d s Survey description Korotkoff sounds were taken as systolic - and diastolic pressures, respectively. Three blood pressures were recorded for each ex NHANES II was conducted from Feb ruary 1976 to February 1980 on a sample selected to be representative of the civilian, noninstitutionalized US population aged six months to 74 years. A total of 20,322 persons were examined. Details of the com aminee; a'seated blood pressure early in the examination, a recumbent blood pressure at the end of the examination, and a second seated blood pressure at the end of the examination. The second seated blood pressure was used In tKeseanalyses, but similar plex survey design, nonresponse adjust results were obtained by analysis of the first ments, the examination procedures, and seated blood pressure, the recumbent blood laboratory measurements have been pub pressure, and the mean of the first and lished (6, 31). Examinations were per second seated blood pressures. Regression formed in mobile examination centers analysis indicated that the relationship be- TEH 0350041 DU P0502 98670 248 PIHKLE ET AL tween blood pressure and blood lead was of the coefficient. Thus, standard weighted not significantly affected by whether or not' multiple linear regression usually finds persons were on hypertensive medication. more variables to be statistically significant Therefore, persons on hypertensive medi than after design effects are incorporated cation were included in the analyses. using SURREGR. The Pooling Project and Framingham The general procedure for variable selec studies were longitudinal epidemiologic in tion was as follows. First, weighted stepwise vestigations of cardiovascular morbidity multiple linear regression was used to de and mortality. In these studies, multiple termine which variables were significantly logistic regression analyses were conducted related (p < 0.05) to blood pressure (SAS; to evaluate risk factors for myocardial in Proc STEPWISE; Stepwise and MAXR farction (11), stroke (P. Sortie, National options). As noted above, weighted multiple Heart, Lung, and Blood Institute, personal linear regression selects the variables which communication, 1984), and death from all would be significant after the design effects causes (12). In these logistic analyses, fatal are incorporated with SURREGR but can and nonfatal myocardial infarctions were select additional variables that would not defined by standard electrocardiogram maintain statistical significance after in changes and, when available, serum enzyme corporating the design effects. Therefore, changes. Sudden death was attributed to fatal infarction when a man who was in apparent good health died within three hours of onset of symptoms with no history of violence or accident causing the fatal outcome. Fatal or nonfatal stroke included hemorrhagic and thrombotic strokes as well as transient ischemic attacks and was de fined according to published criteria (13). the set of variables chosen by weighted stepwise multiple linear regression was used as a starting variable set for SUR- I REGR. Since SURREGR cannot be imple mented in a stepwise fashion, subsequent variable selection was performed by back ward elimination using additional runs of SURREGR. This backward elimination step excluded only one variable in the sys Statistical procedures tolic blood pressure analysis and only one variable in the diastolic blood pressure- Multiple linear regression models were used to determine which variables were sig nificantly related to blood pressure. Since the NHANES II data, were obtained from a complex, multistage probability sample, final hypothesis testing in the multiple regression analysis needed to Incorporate the sampling weights and the complex sur vey design effects. Special computer pro grams are available (SURREGR (14, 15), REPERR (16)) which allow the design ef fects to be incorporated into the variance estimates. SURREGR, which runs in the Statistical Analysis System (SAS) environ ment, was used in this analysis. Compared with SURREGR, standard weighted mul tiple linear regression yields the same esti mates for regression coefficients, but in our analysis, so the design effects had little f influence on final variable selection. Un weighted stepwise multiple linear regres sion was also used to be sure there were no important differences in the significance of lead between unweighted and weighted analyses. The natural log of blood lead was more normally distributed and usually more sta tistically significant than untransformed blood lead. The results reported here are for the natural log of blood lead, but regressions using lead on the untransformed scale gave very similar results. Age was also mod eled as a single variable (sine of age) instead 1 of age and age-squared, with no change in results. The multiple logistic regressions were experience generally (although not always) also performed using programs from SAS. gives a lower estimate of the standard error Proc LOG1ST was used for the stepwise unweighted multzhs and Proc NLIN (d z used for the weirsa calculations. Calc u e: els for effects were s. on segmented regress Re t . Multiple reg-e The first goal of ih termine if blood lea: cantly related to bio males aged 40-59 w mented correlates of: sex, race, and one of ' height to weight. Bod. height5) was used in' iting our attention l white males, there w for race, sex, and. u Although age was oh leant in the stepwise ege-squared were for: regression model to b lead was independen range. The first set of reg tolic and diastolic bio males 40-59 years 1 consisting of age, a| index, and blood le. were done to detenni levels were significt systolic and diastoli: controlling for age, k index, which are wi lates of blood pressi cally significant (p < and diastolic blood regressions (unwei| weighted with desigi These models we porate additional va attention directed a' nrficance of the lead ence of nutritional chemistries. Our obj uate the possible ass or biochemical mes TEH 0350042 DUP050298671 "hus. standard weighted gression usually finds ; statistically significant ffects are incorporated sdure for variable selecFirst, weighted stepwise ession was used to de fies were significantly o blood pressure (SAS; Stepwise and MAXR ?ove, weighted multiple cts the variables which after the design effects `-h SURREGR but can .-iables that would not significance after inign effects. Therefore, > chosen by weighted ineax regression was 'ariable eet for SURSGR cannot be implee fashion, subsequent s performed by back ing additional runs of backward elimination te variable in the sysanalysis and only one stone blood pressure ign effects had little triable selection. Unultiple linear regresbe sure there were no in the significance of ghted and weighted blood lead was more tnd usually more stathan untransformed 'ts reported here are lood lead, but regresuntransformed scale ts. Age was also mod3 (sine of age) instead i, with no change in ic regressions were urograms from SAS. ed for the stepwise LEAD AKD BLOOD PRESSURE 249 unweighted multiple logistic regression, Ta b l e 1 and Proc NLIN (nonlinear regression) was used for the weighted logistic regression calculations. Calculations of threshold lev Variables included in the stepidse regression analyses, white males aged 40-59 years, NHANESII, 1976-1980 els for effects were made using Proc NLIN Age* on segmented regression models (17). Age-squared* Body mass index Dietary iron! Dietary vitamin A! Dietary vitamin Ct Re s u l t s Dietary sodium! Salt shaker sodium Dietary thiamine! Dietary riboflavin! Multiple regression analyses The first goal of the analysis was to de Dietary sodium X salt Dietary niacin! shaker sodium Serum cholesterol! Dietary potassiumt Serum vitamin C! termine if blood lead levels were signifi Dietary aodium- Serum iron! cantly related to blood pressure in white males aged 40-59 years. The well docu mented correlates of blood pressure are age, sex, race, and one of the indices of relative potassium ratio Dietary calciumf. Dietary phosphorus! Dietary proteint Dietary fatt Serum transferrin saturation Serum zinc! Serum copper! Serum albumin! Hemoglobin! height to weight. Body mass index (weight/ height2) was usedlntEfslmaJyiisl By lim iting our attention to 40- to 59-year-old white males, there was no need to control for race, sex, and, to a large degree, age. Dietary carbohydrate! Red blood cell count Dietary cholesterol! Ethanol consumption/week!' Dietary saturated fatty Cigarettes smoked/day ^ acids! Total dietary gTamst Dietary oleic acid! Total dietary calories! Dietary linoleic acidt Although age was only occasionally signif icant in the stepwise analysis, both age and age-squared were forced into each multiple This variable was forced into each regression to remove any possible age effects on blood pressure. ! The natural log and squared transformations of regression model to be certain any effect of these variables were also included in the stepwise I lead was independent of age over this age regression. | range. I The first set of regressions analyzed sys- pressure, but rather to estimate conserva j tolic and diastolic blood pressures for white tively the strength and independence of the I - / males 40-59 years of age with a model relationship between blood pressure and j ill consisting of agel age-squared body; mass blood lead. Therefore, to provide an unu y index, and blood lead. Tiiese regressions sually rigorous test of the independent sig were done to determine whether blood lead nificance of blood lead, almost all the nu ! levels were significantly associated with tritional and biochemical variables in systolic and diastolic blood pressures after NHANES II were included in the stepwise : controlling for age, sex, race, and body mass regression. In addition, to account for pos I index, which are well documented corre sible curvilinear relationships, squared and lates of blood pressure. Lead was statisti natural logarithmic transformations of al cally significant (p < 0.01) for both systolic most all these variables were also included and diastolic., blood-pressures~&P"tdlAe (see table 1). j regressions (unweighted, weighted, and Including these additional 87 variables weighted with design effects). increases the probability of a variable being *' These, models^were expanded-to incor found statistically significant at the 5 per porate additl'orial''variaBIes"with particular cent level due to chance alone. This would < attention directed at the stability and sig- complicate the interpretation of nutritional / ntficance of the leadTcSEfficient tn {he pres and biochemical factors, but including ence of nutritional factors and blood bio these additional variables only makes it chemistries. Our objective was not to eval more difficult for lead to maintain its sig uate the possible associations of nutritional nificance. Thus, this nonselective inclusion or biochemical measurements with blood of variables is not recommended for inves- 7T TEH 0350043 DUP0502 98672 250 PIRKLE ET AL- tigating the independent contributions of starting model for SURREGR, which ad nutritional factors or blood analytes to ditionally incorporates the survey design blood pressure, and no conclusions are effects. For both systolic and diastolic pres drawn in this analysis concerning the pos sures, one variable from the weighted step sible importance of any of these factors. wise regression failed to maintain signifi However, if including so many terms and cance at the 5 per cent level after the design /their nonlinear transformations in the effects were incorporated. The final regres Stepwise regression does not appreciably sion results for systolic and diastolic blood alter the strength of the relationship be pressures, after accounting for the weight tween blood pressure and blood lead, this ing and design effects, are given in table 2. would provide considerable support for an After including the nutritional variables, independent relationship. the blood analytes, and their curvilinear From the 87 nutritional and biochemical transformations, lead remained signifi variables, the weighted stepwise regression cantly associated (p"<r0:01) with' bolFsyi- selected five additional variables for dia tolic and- diastolic blood pressures. The stolic pressure and six additional variables magnitude of this relationship, adjusted for for systolic pressure using a 5 per cent the other significant variables, is depicted significance level. These were used as the graphically in figures 1 and 2. Furthermore, Ta b l e 2 Diastolic and systolic blood pressure multiple regression results, including both sampling weights and design effects, for white males aged 40-59 years, NHAKES II. 1976-J980 Variable Unstandsrdixed rejrTeuton coefficient l statistic Signifi* csnce level segmented regression analyses indicated that there was not a threshold blood lead level in the data below which lead was not significantly related to blood pressure. As mentioned previously, the regression analysis was not directed at examining in dependent effects of nutritional and bio chemical factors. However, it should be Diastolic blood pressure (n * 554; r1 0.23) Age* i Age-squared* -Body mass'index''* TSiood lead (In U) scale) -- Dietary potassium'' -- Hemoglobin's \ Albumin / \ Dietary vitamin Ct; 0.2768 --0.0014 1.131 3.954 -0.0018 1.548 3.587 1.838 0.17 0.864 0.10 0.932 8.55 <0.001 2.83 0.008 4.92 <0.001 3.90 <0.001 2.50 0.018 4.65 <0.001 noted that each of the variables in the final systolic and diastolic models, except albu min and riboflavin, has previously been associafed"wTtS' Wood pressure (1-4,18,19). Albumin may reflect ionized calcium which has been related to blood pressure (201- In addition to tbe variables in table 1, several other variables were evaluated for their potential effect on the significance of blood lead; 1) whether or not on hyperten Systolic blood pressure (n ** 543, r1- 0.22) sive medication, 2) amount of recreational /" Age* fi | Age-squared* 1 Body mass index ` Blood lead (In 1 Beale) <!i Albumin Dietary vitamin Ct -- Dietary riboflavint Dietary oleic acidt -- Serum vitamin Ct 1.311 -0.0068 1.736 8.436 7.088 2.411 -5.509 3.992 -3.472 0.57 0.572 0.30 0.771 9.42 <0.001 3.24 0.003 2.50 0.018 3.84 <0.001 3.07 0.004 2.49 0.018 2.47 0.019 This variable was forced into the regression to remove any possible age effects on blood pressure. + The natural log (In) transformation was selected in the stepwise regression. and/or nonrecreational exercise, 3) resi dence in an urban versus rural area, 4) income greater versus less than 10,000, 5) educational level, 6) season of the year, 7) region of the country, 8) triceps skinfold thickness, and 9) family history of hyper tension. None of these factors appreciably altered the relationship between blood pressure and blood lead. In multiple regression analyses, another consideration is the possibility of signifi cant interaction terms. To evaluate this X I *> FlCURB 1. Adjus NHANESII. Both squared, body mas observations have 1 blood lead of 22 or regression line refl- 1 i at . m i Fig u r e 2. Ac KHANES 11. B< squared, body r observations ha blood lead of 21 regression line i 350o.44 DUP050298673 REGR, which adthe survey design and diastolic presthe weighted step> maintain signifivel after the design L The final regrestnd diastolic blood ng for the weightre given in table 2. .ritional variables, 1 their curvilinear remained signifi01) with both 6ys*d pressures. The nship, adjusted for iables, is depicted id 2. Furthermore, nalyses indicated -eshold blood lead hicb lead was not 'ood pressure, sly, the regression i at examining mtritional and biover, it should be riables in the final dels, except albus previously been ssure (1-4,18,19). red calcium which ! pressure (20). riables in table 1, vere evaluated for the significance of not on byperten:nt of recreational exercise, 3) resi ts rural area, 4) s than $10,000, 5) on of the year, 7) ' ) triceps skinfold history of hyperictors appreciably i between blood analyses, another ;ibility of signifiTo evaluate this LEAD AND BLOOD PRESSURE 251 ADJUSTED SLOOP HAD LEVEL* IMICRDORAMa/OECILiTERI FiGUKE 1. Adjusted diastolic blend pressure and adjusted blood lead levels for white males aged 40-59 years, NHANES II. Both blood pressure and blood lead have been adjusted by regression for the effects of age, agesquared, body mass index, and all other variables significant at the 5 per cent level {Bee table 2). The 564 observations have been reduced to 25 points for illustration. Each point is the mean blood pressure and mean blood lead of 22 or 23 consecutive observations, ordered by increasing blood lead levels. However, the plotted regression line reflects the slope coefficient obtained from the multiple regression analysis of all 564 points. w u r* xz x M it m ADJUITTTO BLOOD LUO LIVILS IMICBOOBAUS/DCCIUTSBI Fig v p.e 2. Adjusted s>-stolic blood pressure and adjusted blood lead levels for white males aged 40-59 years, NHANES II. Both blood pressure and blood lead hove been adjusted by regression for the effects of age, agesquared, body roas6 index, and all other variables significant at the 5 per cent level (see table 2). The 543 observations have been reduced to 25 points for illustration. Each point is the mean blood pressure and mean blood lead of 21 or 22 consecutive observations, ordered by increasing blood lead levels. However, the plotted regression line reflects the slope coefficient obtained from the multiple regression analysis of all 543 points. TEH 0350045 DUP050298674 > V'' K f/ Vi v> "V& :/ 252 PJRKLE FT AJL possibility, an additional weighted stepwise regression analysis was done for systolic and diastolic blood pressures. The variables consisted, of the linear interaction terms between the final variables ,1a the jppdel (shown in table 2) and the linear form of all the other variables originally selected for the initial stepw ise regression (table 1). This amounted to running a stepwise regression with 162 interaction terms added to the final regression models for systolic and diastolic pressures. Using this large set of variables gave a high probability that some variables would enter at the 5 per cent level by chance. However, the purpose was not to determine ifthese variables were independently significant but rather to test further the significance and independence of the relationship between blood pressure and blood lead. As expected, several inter action variables entered the systolic and diastolic regressions, but in each regression, the lead coefficient varied less than 10 per cent and remained significant (p < 0.015). Two other analyses were done to test this relationship further. First, the original weighted stepwise regression was extended to include variables through the 10 per cent significance level to see if marginally sig nificant variables influenced the signifi cance of lead. For both systolic and dia stolic pressures, lead remained significant, and there was little change in the magni tude of the coefficient. The second analysis was the most de manding test made of the independence of the relationship between blood pressure and blood lead. Models for diastolic and Bystolic blood pressures were fit by weighted stepwise regression to the original model variables (table 1), excluding lead. This gave all the other variables and their curvilinear transformations a maximal op portunity to explain variation that could also be explained by lead. After obtaining this new final model without lead, a single regression was run adding the lead variable to the variables of this new final model. For both systolic and diastolic pressures, lead was still statistically significant (p < 0.016), and the magnitude of the lead coef ficient changed less than 10 per cent from those obtained in the original analysis (ta ble 2). The results of these analyses indi cated that the strength and independence of the relationship between blood pressure and blood lead were quite stable. Public health implications Two factors prompted analysis of the potential public health implications of the relationship of blood lead and blood pres sure. First, if causally related, the regres sion results indicated that changes in blood lead levels formerly thought to be of little importance in adults may produce practi cally significant changes in blood pressure. Second, support for causality can be found in animal studies. In male rats, low blood lead levels (approximately 40 pg/dl) have been found to cause systolic blood pressure increases of 15-20 mmHg (21). The follow ing public health implications are based on the .assumption of causality, which is ad dressed later. The general approach to estimating po tential public health implications consisted of two steps: to calculate the change in blood pressure that might result from a specified change in blood lead levels and to use the Pooling Project and Framingham multiple' logistic regression coefficients to calculate the change in incidence of cardi ovascular and cerebrovascular events that would result from the lead-mediated changes in blood pressure. It is assumed in these calculations that the effect of lead is onlv on blood pressure. The three outcomes examined were fatal, and ponfatal myocaTdl^ jnfarcti.on, Mai and nonfaOTitrokes^ and death from aH causes. Analyses specific for white inales aged 40-59 years are available for fatal and nonfatal myocardial infarctions from the Pooling Project final report and for fatal and nonfatal strokes from the Framingham Study. Analysis of death from all causes is available for white males aged 40-54 years from Framings years is not co* up period, eaci the multiple 1oj of risk factors systolic and dh age) for the out base also inclu risk factors for to 59-year age j Therefore, b: regression coefi each outcome f< 59 years in tbs calculated in i The mean risk males in this eg ing the mean oi survey samplin calculating this be representstn aged 40-59 year for calculations causes which < Multiplying the mated number age group yield, events. The baseline formed using tl diastolic or syst individual. The (or decrease) in' risk of an outcoi ing each individ specific blood ; crease) and reca predicted risk, risk was then c; sampling weight the effect of a 11 pressure on risk cardial infarctic stoiic pressure v and for each in risk for myocar lated. Then, e. weighted by its si a new weighted TEH 0350046 DU P050298675 gnificant (p < of the lead coef'.0 per cent from nal analysis (ta e analyses indi* id independence n blood pressure stable, '-ications analysis of the plications of the and blood pres ited, the regreschanges in blood ht to be of little produce practi1 blood pressure, .ity can be found a rats, low blood f 40 #ig/dl) have ic blood pressure (21). The followons are based on ity, which is ad- o estimating po rtions consisted a the change in \l result from a ead levels and to uid Framingham n coefficients to cidence of cardirular events that a lead-mediated . It is assumed in ; effect of lead is .mined were fatal infarction, fatal i death from all for white males 3ble for fatal and ret ions from the ort and for fatal the Framingham from all causes is aged 40-54 years LEAD AND BLOOD PRESSURE 253 from Framingham, but the age range 55-59 which represented white males aged 40-59 years is not covered. For a specified follow years. Multiplying this new mean risk rimes up period, each of these analyses provides the estimated number of US white males the multiple logistic regression coefficients in this age group yielded the new predicted of risk factors (i.e., smoking, cholesterol, number of events. systolic and diastolic blood pressures, and This procedure was used to calculate the age) for the outcome. The NHANESII data effect of blood pressure changes on inci base also includes measurements of these dence of serious outcomes. Over the period risk factors for each white male in the 40- 1976-1980, mean blood lead levels for white to 59-year age group. adult males in the United States decreased Therefore, based on the multiple logistic 37 per cent (from 16.7 to 10.5 pg/dl) (22). regression coefficients, the predicted risk of From a public health perspective, an im each outcome for each white male aged 40- portant calculation is the predicted effect 59 years in the NHANES II data can be' this change in blood lead had on blood calculated in. a straightforward manner. pressure and the incidence of serious out The mean risk of each outcome for white comes which are related to blood pressure. males in this age group is obtained by tak Based on the Pooling Project logistic coef- .vr ing the mean of the individual risks. If the ficients for white males aged 40-59 years, vd ^ survey sampling weights are included in the 37 per cent decrease in blood lead levels ' calculating this mean risk, the result should resulted in a predicted 4.7 per cent decrease I be representative of white males who were in the incidence of fatal and nonfatal myo | aged 40-59 years from 1976 to 1980 (except cardial infarctions oveT 10 years (table 3). i for calculations involving death from all Similarly, based on Framingham logistic causes which covers ages 40-54 years). coefficients, the blood lead decrease re Multiplying the mean risk times the esti sulted in a predicted 6.7 per cent decrease mated number of US white males in this in incidence of fatal and nonfatal strokes age group yields the predicted number of over 10 years. For death from all causes, events. the 37 per cent decrease in blood lead levels The baseline risk calculation was per resulted in a predicted 5.5 per cent decrease formed using the NHANES II measured over the subsequent 11.5 years. Calcula diastolic or systolic blood pressure for each tions not incorporating the sample weights individual. The effect of a specific increase yield very similar per cent changes in inci (or decrease) in blood jpressure pn predicted dence of serious outcomes; thus, the sample risk ofan outcome was measured by chang weighting has little effect on these esti ing each individual's blood pressure by the mates. specific blood pressure increase (or de For white males aged 40-59 years, the crease) and recalculating each, individual's effect of different mean blood lead levels predicted risk. The new mean predicted on the predicted incidence of fatal and non- risk was then calculated using the survey falal myocardial infarctions, fatal and non sampling weights. For example, to measure fatal strokes, and death from all causes is tie effect of a 1 mmHg increase in diastolic graphed in figures 3-5. From these graphs, pressure on risk of fatal or nonfatal myo the effect of changes in mean blood lead cardial infarction, each individual's dia levels can be estimated. stolic pressure was incremented 1 mmHg, Stepwise multiple logistic regressions and for each individual, a new predicted were performed to determine if lead was a risk for myocardial infarction was calcu significant predictor of diastolic blood pres lated. Then, each individual risk was sure greater than or equal to 90 mmHg. weighted by its survey sampling weight, and These 6tepwise logistic analyses found a a new weighted mean risk was calculated similar set of statistically significant vari- TEH 0350047 DUP0502 98676 254 FIRKLE ET AL. Ta b l e 3 Predicted change in the number of Qerious outcomes as a result of the 37% decrease in blood lead levels in adult white males aged 49-59 years. NHANESII, 1976-1980 Follow-up period (yean) Predicted no. of events during follow-up period Before 37% decrease in blood lead After 3796 decrease in blood lead Absolute difference is so. of vents % difference is DO. of events Fatal and nonfatal myocardial in farctions* Fatal and nonfata) strokest Death from all causes} 10 10 11.5 1,638,200 413,200 1,353,600 1,560,900 385,700 1,279,700 77,300 27,500 73,900 4.7 6.7 5.5 * Men with a history of heart attack were excluded; based on Pooling Project multiple logistic coefficients for diastolic blood pressure, smoking, cholesterol, and age. t Men with a history of previous stroke were excluded; based on Framingham multiple logistic coefficients for systolic blood pressure. Brooking, cholesterol, and age. t Data available only for ages 40-54 years; based on Framingham multiple logistic coefficients for diastolic blood pressure, smoking, and cholesterol. 1750 - A Z a 1700 - o E V<i. X 1650 ~4 *5 < -o is 1600 o| s 1550 1 X Q 1500 z t- // 6 1- - -t .- 10 14 18 22 26 MEAN BLOOD LEAD LEVEL {ps/dll 30 Fig u r e 3. Predicted number of fatal and nonfstal myocardial infarctions for a follow-up period of 10 years (using the Pooling Project multiple logistic regression coefficients) and mean population blood lead levels for white males aged 40-59 years, NHANES 11. Men with a history of a bean attack were excluded. ables and found that blood lead was signif- have shown that the body lead burden of ^ leant (p < 0.01). From these results, the 37 modem humans is about 500 times higher V per cent decrease in blood lead levels than that of preindustrial humans. Thus, ^ yielded a predicted 1.3 million fewer white although blood lead levels less than 30 pg/ ^ males aged 40-59 years with hypertension dl are designated as "low," they are actually (diastolic &90 mmHg), a decrease of 17.5 high on a historical scale. The effect of per cent. current lead exposure on human health is Dis c u s s io n therefore an important area of investiga tion. Lead is a toxic substance with no known After adjusting for age, body mass index, physiologic function. Isotopic studies (23) nutritional factors, and blood biochemis- FlGURE 4. Predi Framingham multii aged 40-59 years, N 1 Fig u r e 5- Pr Framingham mu aged 40-54 years TEH 0350048 DU P 050298677 od lead levels in adult tnce 3. Of nu ft difference is so. of event* w 4.7 00 6.7 00 5.5 le logistic coefficients le logistic coefficients efficients foi di&stolic LEAD a n d b l o o d p r e s s u r e 460 r 440 S EtZ n hi 420 IfuS5E>lCfjCi 5S9 400 380 s o. 380 255 6 10 14 18 22 26 30 MEAN BLOOD LEAD LEVEL (p/dl) FIGURE 4. Predicted number of fata! and nonfstal strokes for a follow-up period of 10 years (using the Framingham multiple logistic regression coefficients) and mean population Mood lead levels for white males Aged 40-59 years, NHANES II. Men with a history of a stroke were excluded. jp period of 10 years blood lead levels for dueled. lead burden of >00 times higher humans. Thus, less than 30 fig/ they are actually The effect of iuman health is ea of investiga- ody mass index, ood biochemis- 1500 1450 mx <bQ"bl. mUatl 1400 V 03- 1350 ids 2d 5 n< O Z c o O w 1300 hoixt bce. Uol 1250 1200J- V,, 10 14 __1__ 16 22 26 ~30 MEAN 8L0OD LEAD LEVEL fpg/dl) Fig u r e 5. Predicted number of deaths from all causes for a follow-up period of 11.5 years (using the Framingham multiple logistic regression coefficients) and mean population blood lead levels for white males aged 40-54 years (rather than 40-59 years). > TEH 0350049 DUP050298678 256 P1P.KLE ET AL. tries, whole blood lead was significantly sure and blood lead. The possibility re related to both systolic and diastolic blood mains that an unknown omitted variable pressures in white males aged 40-59 years. exerts a significant effect on both blood In a conservative estimate of the strength pressure and blood lead, but it is unlikely. and independence of this relationship, cur The strength and independence of the vilinear transformations of almost all the relationship between blood pressure and variables were available for inclusion in the blood lead do not alone prove a causal stepwise regressions to account for possible association. A causal association between important nonlinear relationships. More lead and blood pressure has been investi over, a large set of interaction terms was gated in animal studies covering a large included in the stepwise regression to ac range of lead exposure. Of particular rele count for their potential influence. Includ vance to this analysis are studies in rats ing both curvilinear transformations and exposed to low levels of lead, an exposure interaction terms had little effect on the insufficient to cause renal damage. Few statistical significance or magnitude of the studies axe available which examined the blood lead coefficient. effect of such low doses. Victery et al. (21) The possibility of omitting a relevant found that male rats with blood lead levels \ variable faceB every epidemiologic study or of about 40 ftg/dl developed significant sus- \ "7 survey. To provide the most conservative tained increases in systolic blood pressure i analysis, almost all the nutritional factors (15-20 mmHg) compared with controls. ' and blood biochemistries in the NHANES Using doses of lead much less than those II data base were included in the stepwise causing rat blood lead levels of 40 /ig/dl, analysis, whether or not they had previ Perry et al. (24) and Kopp et al. (25) also ously been associated with blood pressure. found increases in systolic pressure in rats. Despite the inclusion of all these variables, By contrast, Victery et al. noted that male there was little effect on the strength of the rats with blood lead levels of 71 ^g/dl relationship between blood pressure and showed no change in systolic pressure com blood lead, and this extensive screening of pared with controls, suggesting a biphasic variables provides strong support for an blood pressure response to blood lead levels. independent effect. It is_still pgssible that A biphasic response could account for the a variable related to blood pressure baa results of animal studies that did not find been_ omitted. To affect the results of this a blood pressure effect at much higher doses analysis, such an omitted variable would of lead exposure. It may also help to clarify have to be related to both blood pressure inconsistent epidemiologic findings of the and blood lead and not be well represented effect of lead on blood pressure in persons by any of the variables already in the step without nephropathy who have mild to wise regression. There is adequate unex moderate as contrasted with marked ele plained variation in blood pressure for an vations of blood lead. omitted variable to be significantly related Webb et al. (26) further investigated the to blood pressure and still have a negligible mechanism of the effect of lead on blood effect on the relationship between blood pressure in male rats. They found that the pressure and blood lead. In addition to the blood pressure changes in male rats with variables in table 1, variables indicating low blood lead levels were associated with residence, income status, exercise, use of an increased vascular responsiveness to hypertensive medication, educational level, alpha adrenergic agonists, which could ac season of the year, region of the country, count for the pressor effect of lead. Animal triceps skinfold thickness, and family his studies can be summarized as indicating a tory of hypertension did not appreciably causal relationship between relatively low alter the relationship between blood pres blood lead levels and increases in blood pressure. Th=*blood pressun rationalizaum ; imal and epics: A statisticaltween blood 'es. without rena. in two other = Also, bypercsn ment have has stores of lest lead (29). Thsassociation cr : lead and blocc Additional s: useful in ever rent findings: females. Epic have to carer for blood lead a coefficient c and the expec a blood lead mmHg. Afri-: study alone c ship betweei blood pressn with curren studies, indi ship is probe Assuming dieted the e decrease in males on th (table 3). T and nonfatt cent), fatal cent), and cent) const benefits. T of events w levels are these calct to affect oi lations in< blood lead tional deci events. Additio: TEH 0350050 l DUP050298679 'he possibility re- j omitted variable ;ct on both blood but it is unlikely, iependence of the ood pressure and le prove a causal sociation between has been rnvesti- covering a large Df particular relexe studies in rats lead, an exposure nal damage. Few ich examined the Victory et ah (21) a blood lead levels ed significant soslie blood pressure sd with controls, h less than those evels of 40 pg/dl, pp et al (25) also c pressure in rata. 1. noted that male vels of 71 fig/dl olic pressure cornresting a biphasic j blood lead levels, d account for the that did not find nuch higher doses Jso help to clarify ic finding of the assure in persons 10 have mild to with marked ele- r investigated the of lead on blood ?y found that the a male rats with e associated with sponsiveness to , which could ac t of lead. Animat d as indicating a en relatively low creases in blood LEAD AND BLOOD PRESSURE 257 pressure. There is 8 suggestion of a biphasic analyses were done to estimate the effect blood pressure response which provides a of the 37 per cent decrease in blood lead rationalization for the findings ofother an from 1976 to 1980 on the number of persons imal and epidemiologic studies. with diastolic pressures 2:90 mmHg. The A statistically significant association be Joint National Committee on Detection, tween blood lead and blood pressure in men Evaluation and Treatment of High Blood without renal impairment has been found Pressure uses 90 mmHg as its cutoff to in two other epidemiologic studies (27,28). define hypertension and recommends Also, hypertensive men with renal impair treatment of diastolic pressures of 90 ment have been found to have larger body mmHg or higher (30). The 37 per cent stores of lead as measured by chelatable decrease in mean blood lead levels resulted lead (29). There appears to be either a lesser^ in a predicted 1-3 million fewer white males association or no association between blood / aged 40-59 years with hypertension (dia lead and blood pressure in women (5, 27). j stolic pressures ^90 mmHg), a decrease of Additional epidemiologic studies will be0 17-5 per cent. useful in evaluating the consistency of cur Finally, the absolute magnitude of the rent findings and in exploring the effect in effect of lead on the incidence of these females. Epidemiologic investigations will cardiovascular and cerebrovascular events have to carefully consider statistical power, in adult males is almost certainly underes- ^ for blood lead measurements typically have timated since 1) persons were excluded who a coefficient of variation of 10-20 per cent, had a history of a myocardial infarction or and the expected blqod pressure change for stroke, and in these persons blood pressure a blood lead ct^rige may be only a few increases likely increase the chance of a mmHg. Although the results of the current" repeat event/2)/the calculation was limited 7 study alone do not prove a causal relation-1 to the age range 40-59 years (40-54 years ^ Bhip between low blood lead levels andj for death from all causes); and 3) races blood pressure, our findings are consistent other than white were not included. with current epidemiologic and\animal RErERENCES studies, indicating that a-causal relation ship is probable. Assuming a causa] relationship, we pre dicted the effect of the recent 37 per cent decrease in blood lead levels of adult white males on the incidence of serious outcomes (table 3). The predicted decreases in fatal and nonfatal myocardial infarction (4-7 per cent), fatal and nonfatal strokes (6-7 per rent), and death from all causes (5.5 per cent) constitute substantial public health benefits. The relative changes in number of events with changes in mean blood lead levels are striking (figures 3-5), and in these calculations, lead has been assumed to affect only blood pressure. These calcu lations indicate that lowering the mean blood lead further would result in addi tional decreases in the number of predicted events. Additional multiple logistic regression 1. Harlan WR, Hull AL, Schmouder RL, et al. Die tary intake and cardiovascular risk [acton. Pan I. Blood pressure correlate*. United State* 197175, Hyattsville, MD: National Center for Health Stalistica, 1982. (Vital and health statistics. Series 11: Data from the National Health Survey, no. 226) (DHHS publication no. (PH51S3-1646). 2. Harlan WR, Hull AL, Schmouder RL,et al. Blood . pressure and nutrition in adults. The National Health and Nutrition Examination Survey. Am J Epidemiol 1984:120:17-28. 3. McCarron DA Morris CD, Henry HJ, et al. Blood pressure and nutrient intake in the United States. 4. Science *984:224:1392-8. Stanton JL, Braitman LE, Riley AM, et al. De mographic, dietary, life style and anthropo metric correlates of blood pressure. Hypertension 1982;4(Suppl ni):135-42. -- ----------------------------------5. Harlan WR, Landis JR. Schmouder RL, et al. Relationship of blood lead and blood pressure in the adolescent and adult US population. JAMA (in press). 6. National Center for Health Statistics. Plan and operation of the second National Health and Nu trition Examination Survey, 1976-1980. Wash ington, DC: National Center for Health Statistics, 1981. (Vital and health statistics. Series 1, no. 15) (DHHS publication nD. (PHS)B1-1317). TEH 0350051 DUP0 502 98680 I j 1 < 9 1 `1 I 258 P1RKLE FT AL. 7. Gunter EW, Turner WE, Neese JW. et al. Labo ratory procedures used by the Clinical Chemistry Division. Centers for Disease Control, for the second National Health and Nutrition Examina tion Survey (NHANES II), Atlanta. GAi Centers for Disease Control, 1981. 8. Barthel WF, Smrek AL, Angel GP, et al. Modified Delves cup atomic absorption determination of lead in blood. J Assoc Off Anal Chem 1973; 66:1252-6. 9. Delves HT. A micro-sampling method for the rapid determination of lead in blood by atomicabsorption spectrophotometry. Analyst 1970;95: 431-8. 10. National Center for Health Statistics. National Health and Nutrition Examination Survey 19761980 (NHANES II). Serum cholesterol special data tape. Catalog no. 54.1C. Second release. Na tional Center for Health Statistics, Washington, DC, 1984. 11. The Pooling Project Research Group. Relation ship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of the pooling project. J Chronic Dis 1978;31:201306. 12. McGee D, Gordon T. The results of the Framing ham Study applied to four other US-based epide miologic studies of coronary heart disease. The Framingham Study. Section 31. DHEW publica tion no. (NIH)76-1083. National Institutes of Health. Washington, DC; US GPO, 1976. 13. Shurtleff D. Some characteristics related to the incidence of cardiovascular disease and death: Framingham Study, 18-year follow-up. Section 30. DHEW publication no. (NIH174-599. National Institutes of Health. Washington, DC: US GPO, 1974. 14. Shah BV. SURREGR: standard errors of regres sion coefficients from sample survey data. Re search Triangle Park, NC: Research Triangle In stitute. 1982. 15. Shah BV. Software for survey data analysis. Am Stat 1984;38:68-9. 16. Survey Research Center Support Group. REPERR: repeated replication sampling error analysis, OSIRIS IV user's manual. Institute for Social Research, 1979. 17. Hudson DJ. Fitting segmented curves whose join points have to be estimated. 3 Am Stat Assoc 1966:61:873-80. 18. McCarron DA. Stanton 3, Henry H, et al. Assess ment of nutritional correlates of blood pressure. Ann Intern Med 19S3;98(part 2):715-19. 19. Langford HG. Dietary potassium and hyperten sion: epidemiologic data. Ann intern Med 19S3;S8(part 2):770-2. 20. McCarron DA. Low serum concentrations of ionited calcium in patients with hypertension. N Engl 3 Med 1982;307:226-8. 21. Victefy W, Vender AJ, Sbulak JM, et al. Lead, hypertension and the renin-angiotensin system in rats. J Lab Clin Med 1982;99:354-62. 22. Annest JL, Pirkle JL, Makuc D, et al. Chronolog ical trend in blood lead levels between 1976 and 1980. N Engl J Med 1983;308:1373-7. 23. Settle DM, Patterson CC. Lead in albacore: guide to lead pollution in Americans. Science 1980;207:1167-76. 24. Perry HM, Erlanger M, Perry EF. Increase in the systolic pressure of rats chronically fed cadmium. Environ Health Perspect 1979;28:251-60. 25. Kopp SJ, Perry HM, Glonek T. Cardiac physio logic-metabolic changes after chronic low-level heavy metal feeding. Am J Physiol 1980;239:H2230. 26. Webb RC, Winquist RJ, Victory W, et aL In vivo and in vitro effects of lead on vascular reactivity in rata. Am 3 Physiol 19S1;241:H211-16. 27. BeeverBDG,ErskineE,RobertsonM,etal.B!oodlead and hypertension. Lancet 1976;2:1-3. 28. KromhoutD, Couland CL. Trace metals and CHD risk indicators in 152 elderly men (the Zutphen study), Eur Heart 3 19S4;5(abstr suppl 1):101. 29. 3atuman V, Landy E, Maesaka JK, et aL Contri bution of lead to hypertension with renal impair ment. N Engl J Med 1983;309:17-21. 30. Joint National Committee on Detection, Evalua tion. and Treatment of High Blood Pressure. The 1934 Report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure. Arch Intern Med 1984:144:104657. 31. National Center for Health Statistics. Blood lead levels for persons ages 6 months-74 years: United States, 1976-80. Washington, DC: National Cen ter for Health Statistics, 1984. (Vital and health statistics, Series 11, no. 233) (DHHS publication no. (PHS) 84-1683). .KMERICA* Copyright ft 19K * All rights reserves A> 0* SPEC ROKr Use. 323C1* Wo k * at a =a Anc empfci wes grab* 162 a sow cefea 6S53C 6co- tops OCKI & Outbrea itis in co: Received final fora J 'Health Health ant bassee. FL Mr. Lieh v Texas Scb atHousvr ' Divisi Office, Ce "Past* Port Rid `Natit eases, Nr `Univ Health S * Cun Valley, 1 Repr PDHF.C Taliaha DHf the Na' supper Tb! Ken i Healt assist TEH 0350052 DU P050298681