u

U.

The National Health and Nutrition Examination Surveys(NHANES)Jar-es L Pirkie. MD P*D 3ebra j Brody. MPH; Eiame W Gunter; Rachel A. Kramer. ScD;Dan el C Paschai. PhD. Katrenne M Flegal. PhD, MPH; Thomas D Matte, MD, MPH

Objective.—To describe trends in blood lead levels for the US population andselected population subgroups dunog the time period between 1976 and 1991.

Design.—Two nationally representative cross-sectional surveys and one cross-sectjonal survey representing Mex>can Americans in the southwestern UnitedStates.

Setting/Participants.—Participants in two national surveys that included bloodlead measurements: the second Natxxial Health and Nutrition Examination Survey,1976 to 1980 (n=9832), and phase 1 of the third National Hearth and Nutrition Ex-amination Survey, 1988 to 1991 (n=12 119). Also, Mexican Americans participat-ing m the Hispanic Hearth and Nutrition Examination Survey, 1982 to 1984(n=5682).

Results.—The mean bkxxj lead level of persons aged 1 to 74 years dropped 78%.from 0.62 to 0.14 umol/L (12.8 to 2.8 pg/dL). Mean Wood lead levels of children aged1 to 5 years declined 77% (0.66 to 0.15 umol/L [13.7 to 3.2 u dU) for non-Hispanicwhite children and 72% (0.97 to 0-27 umoi/L [20.2 to 5.6 yg«L]) for non-Hispanicblack children. The prevalence of Wood lead levels 0.48 umoVL (10 ug/dL) orgreaterfor children aged 1 to 5 years declined from 85.0% to 5.5% for non-Hispanic whitechildren and from 97.7% to 20.6% for non-Hispanic black children. Similar declineswere found in population subgroups defined by age, sex, race/ethnicity, incomelevel, and urban status. Mexican Americans also showed similar declines in bloodlead levels of a slightty smaJter magnitude over a shorter time.

Conclusions.—The resorts demonstrate a substantial decline in bkxxj lead lev-els of the entire US population and within selected subgroups of the population. Themajor cause of the observed decline in blood lead levels is most likely the removaJof 99.8% of lead from gasoline and the removal of lead from soldered cans.Although these data indicate major progress in reducing lead exposure, they alsoshow that the same sociodemographic factors continue to be associated with higherblood lead levels, including younger age, male sex, non-Hispanic black race/ethnicity, and tow income level. Future efforts to remove other lead sources (eg,paint, dust, and soil) are needed but will be more difficult than removing lead fromgasoline and soldered cans.

(JAMA. 19W;2722S4-291>

Fror- \f.e Division ot Environment --*eaiin uaoora- Prevention. Hyattsvtiie. Ma (Drs Kramer ano RegaJ and:ory Sc.ences iDrs P-ovie and Pascr* ano Ms Gutter)and Leaa Poisoning Prevention Brancr iDr Mar»). Na-!ior.ai Center 'or Environmental HeaHTv Centers 'orDisease Control and Prevention Atiarta. Ga. ana Drvi-s»on oi Heaitri ExammaiKXi Statistics National C*nter'or Heaitn Statistics Centers tor D'seas* Contra ana

t requests TO Divs*on o* Environmental HearthLaboratory Sciences. National Center (or Environmen-tal Heaitn Centers For Disease Control ana Prevention.MS-P20. Atlanta. GA 30333 (Of PtrWe)

LEAD has been dispersed in the envi-ronment in substantial quantities over along period of time. Compelling evidencefrom the scientific community on a widerange of adverse health outcomes hasplaced lead in the forefront of environ-mental health concerns. In the 1970s,federal regulatory and legislative effortswere undertaken to reduce lead haz-ards, including actions to limit the useof lead in paint and gasoline.1 The sec-ond National Health and NutritionExamination Survey (NHANES II, 1976to 1980) established baseline lead mea-surements for the US population anddemonstrated the pervasiveness of lead

S«e also pp 277 and 315.

exposure across race, urban and ruralresidence, and income levels.* Datafrom NHANES II showed a decline inblood lead levels from the beginning tothe end of the survey period that wasclosely correlated to declines in theuse of leaded gasoline during theseyears.*

Since 1980, intensive federal, state,and local actions directed at primaryprevention have been taken to furtherreduce lead exposure from gasoline,paint, solder, and other sources. Sec-ondary prevention activities, such asscreening for early detection and leadeducation programs, have also beenimplemented. New data from phase 1 ofthe third National Health and NutritionExamination Survey (NHANES IIIphase 1, 1988 to 1991) permit examina-tion of changes in blood lead levels since1980 in the US population and evalua-tion of the impact of these regulatory

284 JAMA July 27 1994—Vo« 272. No

actions. The Hispanic Health and Nu-trition Examination Survey i HHANES.19*2 to 19841 provides data on MexicanAmericans at *an intermediate timepoint. These analyses of trends in bloodlead levels serve both to evaluate theeffectiveness of prevention programsand to develop new strategies to furtherreduce lead exposure in the L'nitedStates.

METHODSDesign and Data Collection

The National Health and NutritionExamination Surveys (NHANES) aredesigned to measure and monitor thehealth and nutritional status of the USpopulation. The general design of theNHANES is a stratified multistageprobability cluster sample of householdswhose target population is civilian non-institutionalized persons residing in theL'nited States. Blood lead levels weredetermined in NHANES II (1976 to1980), HHANES (1982 to 1984), andNHANES III (1988 to 1994). The esti-mates from NHANES II and .NHANESIII are based on a national sample,whereas HHANES sampled three His-panic subgroups.*4

National trends of blood lead levelspresented in this article were based pri-marily on comparisons of data fromNHANES II and NHANES III phaseI (1988 to 1991). Trends for MexicanAmericans were based on a comparisonbetween the estimates from HHANESand NHANES III phase 1. TheHHANES also provides an intermedi-ate point in time between NHANESII and NHANES III. The HHANESsample of Mexican Americans includedonly those residing in the southwesternUnited States whereas the NHANESIII phase 1 sample represented Mexi-can Americans residing in the entireUnited States.

Venous blood lead measurementswere obtained for persons aged 6 monthsto 74 years in NHANES II; personsaged 4 to 74 years in HHANES; andpersons aged 1 year and older inNHANES III phase 1. Analysis waslimited to persons aged 1 to 74 years fornational trends and aged 4 to 74 yearsfor trends in the Mexican-Americanpopulation. The final samples used foranalyses included 9832 and 12119 forthe national trends from NHANES IIand NHANES III phase 1, respectively,and 5682 and 4067 Mexican Americansfrom HHANES and NHANES III phase1, respectively. Data from all of the sur-veys were collected using a householdinterview followed by a detailed medi-cal examination in a mobile examinationcenter.

25^

** 2°"g" 15 -

1 10-0>o.

0 -<

NHANES II

1* 1 II i il ', o ir~v— -

_^rfl ! i 1 1 1 rTrrrrr-^ .. . ,-

#

«"

£c.

25-,

20-

15-

10-

5 •

T

oim',

•'

'•

'

'

NHANES lit Phase i

\mif9~*JP

f ]^__L |RF ff Fbi-i r-

0.10 0.19 029 039 048 058 068 077 087 097 1 06 1 16 1 25 i 35 >T 45(2) (4) (6) <8l (10) <T2) (14) 06) MS) i20) (22) (241 i2fii 26) '>30i

Blood Lead Level.

Fig 1 —Stood lead levels (or persons aged 1 to 74 /ears: United States, second NatKxxai Heartn and Nu-intion Eiammatjon Survey (1976 to 1980, top) and pnase 1 of me third NaoonaJ Hea/ffi and Nutntwn Ex-amination Survey (1988 to 1991. bottom).

NHANES II

0.10 0.19 0.29 039 048 058 068 0.77 087(2) <4) (6) (8) (10) (12) (14) (18) (18)

Blood Lead Level, moVL

M6(24)

Fifl 2.—Stood lead tov*s lor children aoed t to 5 years; United States, second NabonaJ Hearth and NutrmonExamination Survey (1976 to i960, top) and phase 1 of tt» trwd MaaonaJ Health and Nutrition Exa/rwabonSurvey (1988 10 1991. bottom).

The response rates for blood lead col-lection in the three surveys ranged from61% to 69%. Previous nonresponse biasanalyses conducted for NHAN'ES II,HHANES, and NHANES III phase 1indicated that there was no apparentbias due to nonresponse.7 *Laboratory Method*

All venous blood specimens were col-lected in the mobile examination cen-ters, frozen, and shipped on dry ice tothe NHANES laboratory, Division ofEnvironmental Health Laboratory Sci-

ences. National Centers for Environ-mental Health, Centers for Disease Con-trol and Prevention, Atlanta, Ga, foranalysis. The methods for determininglead in blood, including descriptions ofquality control and assurance proce-dures, have been described for each sur-vey.*-11 Comparability has been estab-lished for the method used in NHANESHand HHANES (modified Delves cup)and that used in NHANES III phase 1(graphite furnace atomic absorptionspectrophotoraetry), as described byMiller et al.u In each of the three sur-

JAMA Jury 27 1994—Vol 272. No 4 in 5ood Lea-2 Leve-s—^"V e e; a- 285

8 8 8 S 8 8 8 S 8 S

14x»33

00019

PREDICTED BLOOD Pb (ug/dL)

8 8 * 8 8 3 8 8 8I2J

I*•»»M

I . I I I I I I 1

HOURS OF PLAY ON FLOOR VS. DISTANCE I-ROM NL SITE

2 4 6 8 10 12

DISTANCE (rings of about 1/8 mile)

i.oo

0.00 -

A PredictedN - 477

v ObservedN - 490

teCVALIDl.ASCl

1 2 3 4

LOG(e) BLOOD LEAD CONCENTRATION (ug/dL)

PERCENT WITH AIR CONDITIONING VS. DISTANCE FROM NL S

2 4 6 8 1 0 1 2

DISTANCE (rings of about 1/8 mile)

HOURS OF OUTDOOR PLAY VS. DISTANCE FROM NL SITE

<3 -

co 2oc

0 2 4 6 8 10 12

DISTANCE (rings of about 1 /8 mile)

y , j - . .. . ...... ,. -

• - - n , . . n . . . : . . .

NUMBER OF CHILDREN IN HOUSE VS. DISTANCE FROM NL

2.0

5 10

DISTANCE (rings about 1/8 mile)

MEAN BUILDING CONDITION VS. DISTANCE FROM NL SITE

0 6 10DISTANCE (rings about 1/8 mRe)

IS

MEAN YEARS OF EDUCATION VS. DISTANCE -ROM NL SITE

14 -

13 - T

11

12 - 1

0 5 10DISTANCE (rings about 1/8 mite)

fe-

INCOME. DOLLARS <£m

§c

o

> enOrn

CD

3<D _*2. O

&

01

* |O CO ^. l ' '5 o o o >S a a o ~^:> o o o ^> O 0 0 _

"71 1

1 ——— -^ ———— 1

"\.

'JL'i\ *i — 1« ^ *^^^

i i

^_oom<CO

oCD— I.

ZOmTl

O

zr/"***UJ

Hm

<•

TOTAL DUST LOADING VS. DISTANCE FROM NL SITE

2.0

0 5 10DISTANCE (rlnQS about 1 /8 mite)

10

LOG OF DUST LEAD LOADING VS. DISTANCE FROM NL SITE

0.5

2 -0.5Q

6-1.5

-2.5

-3.50 5 10

DISTANCE (rings about 1 /8

re. //

1)

LOG EXTERIOR LEAD PAINT x CONDITION

enH>gmx-xn.300CDorO&«x\00

O0mxHm3)ODrm>oT)

CO

gCD

om

oZrCO

m

fi

03-IZOm

CO

LOG WATER LEAD CONCENTRATION

f~OQO~n

IDr5DOOO

COOCO

om3)O

COHm

m

Iento

X00

O

LOG DUST LEAD CONCENTRATIONCD CD

OCDO~nac00Hr~m>oOOp00

goo

OmTl33O

00Hm

r-ooz— im

LOG INTERIOR LEAD PAINT x CONDITION 33O

G) N> -» O -* -JJO | ——————————————————————————————————

QH

6m2 01<pToCD

£003® ?i^^ O

i i i.

\^s*1*^

t . jj rr „(^ ,%^j ""• — ^— — i* H^^^

i ! «^

^i - '

I 1 i

r~mDuZ-i

OTOGO— 1

Om~n33O

Zr~GOHm

MEAN LOG BLOOD LEAD VS. DISTANCE FROM NL SITE

22

0 2 4 6 8 1 0 1 2

DISTANCE (rings of about ' /8 mile)

LOG OF SOIL LEAD CONC. VS. DISTANCE FROM NL SITE

8

< 7cc '

UJ

Q6

_Jco 5

0 5 10

DISTANCE • (rings about 1/8 mile)

s

PERCbN wriH BLOOD LEAD 20+ VS. DSTANCE FRON/ NL S

2 4 6 8 10 12

DISTANCE (rings of about 1/8 mile)

MEAN BLOOD LEAD VS. DISTANCE FROM NL SITE

0 2 4 6 8 10 12

DISTANCE (rings of aocut 1/8 mile)

re.

: • r • v • r T •* -;: • * " : ' " : 2 "* L;

PERCENT WITH BLOCD LEAD 1C+ VS. DISTANCE FROM NL 3

2 4 6 8 10

DISTANCE (rings 9bout 1/8 mile)

12

PERCENT WITH BLOOD L£AD 1t>+ VS. DISTANCE FROM NL 3

20

z 15 -

XOu-O

LUOccUJCL 5 -

0 2 4 6 8 1C

DISTANCE (rings about 1/8 mile)

12

and predicted blood lead. The over-all goodness of fit was comparable or better to thatfor the calibration community, the lead smelter community of Mid?ate, Utah. (Figures18, 19, 20).

3. Sensitivity analyses were based on a range of values for the contribution oflead-contaminated soil to household dust. The default assumption, that theconcentration of soil-derived lead in house dust is 0.70 of the sofl lead concentration,was judged to he appropriate, and also provided a very good fit to the chOd blood leaddata from the Madison County study. Alternative values in the sensitivity analyses werebased on statistical analyses from study data: 0.29 (distances up to 1/4 mfie), 0.46 (alldata), and 0.55 (distances to 3/8 mile). The curvilhiearity in the blood lead T»,environmental lead relationship was characterized by a pa&sh-e-to-total gut leadabsorption fraction of 0.20, as found from In-vitro rtudlM. The higher dust/soflcoefficients of 0.70 and 0.55 are more appropriate for risk assessment, more realistic forproperties of the site, and provide a good fit to the data.

4. Remediation goals for sofl abatement were calculated from the IEUBK Modelso as to generate not more than 5 percent of children of ages 6 to 84 months with bloodlead 10 ug/dl or greater. The calculated soil lead concentrations depended on theassumptions one made about soil to dust transport, but otherwise assumed only defaultparameters. The soil remediation levels ranged from 340 ppm (soil-to-dust coefficient »0.70) to 480 ppm (solMo-dust coefficient of 0.29). This suggests a range of soil leadcleanup values of 400 to 500 ppm. The results are shown in Table 1.

REFERENCES

1. Dlinoii Department of Public Health, 1994. Ml4»ofl County Lead Exposure Study.Citv. niinoii. Draft for public comment. Springfield, Iffinois, Feb. 1994.

2. Marcus A.H., Hogan K.f White P., Van Leeuwen P. 1994. Comnyttaofl MadisonCounty Lead Exposure Study. Granite City, Illinois. la-house memo, U.S. EnvironmentalProtection Agency, May 1994; corrected draft, Sept. 18, 1994. Research Triangle Park,NC.

3. Agency for Toxic Substances and Disease Registry, 1994. Cnmmmu on Mnlfojfcand Cadmium Tfoposure study with Biological Markers Incorporated. Review draft Atlanta,Georgia.

4. Wilkinson, L. 1992. SYSTAT! Jtis System for Statistics. Systat Inc., Evanston IL.

TABLE 1SENSITIVITY ANALYSES FOR SOIL LEAD CLEANUP LEVELS

SOIL CONTRIBUTIONTO HOUSE DUST

0.70 (default)0.550.460.29

SOIL LEAD CLEAN-UPCONCENTRATION (ppm)

340370420480

6

factor in childhood blood lead in Madison County, and is much less important than soillead and dust lead, (Figures 4, 5, 6, 7, 8).

-4. Tap water lead concentrations are highest, on average, in the area closest tothe NL smelter, but show little relation to distance from the she farther away. Thissuggests that tap water lead may be a contributing factor, hut is not th« primaryenvironmental factor in childhood blood lead In the Madison County study. (Figure 9).

5. Total dust loading shows almost no relationship to distance from the NL site,on average, so that increased dustlness of homes cannot explain the higher householddust lead loadings found near the NL site; it is the higher concentration of lead inhousehold dust that accounts for higher dust lead loadings near the NL she. Fartherstudies are needed to determine whether this la an artifact of the method of house dustsampling, or whether this is a generalizable conclusion. (Figures 10, 11).

6. Lead in household dust is the primary exposure pathway for young children*Lead in soil and lead in deteriorating lead-based paint are primary sources for lead inhouse dust, with substantial variability from one household to another*

7. Many sodo-demographk factors are related to individual childhood bloodlead concentrations, and also show a systematic relationship to distance from the NLtit*. Theco include increasing parental education, Increasing Income, and decreasingnumbers of pre-school children per household with Increasing distance from the smelter.Households In the study with the most children and the fewest resources to cope withlead poisoning are located closest to the NL site. (Figures 12, 13, 14, 15).

8. Individual chfld-spedfk behaviors may affect blood lead concentration, withsubstantial differences among children. These Include hours of outdoor play, frequencyof mouthing non-food objects, and hours of Indoor play on the floor. Thera are somesystematic relationships, such as a highly significant tendency for children in the studywho live closer to fee site ta have more hours of outdoor play, on average, than childrenwho live farther away (Figures 16, 17).on«c 1

9. In view of correlations that were found between distance from the NL sits,blood lead, environmental lead, household sodo-demographk characteristics, andtypical child behavior, then are some concerns that tht sample of children may not berepresentative of die community. TUs study used volunteer subject*. The lowestresponse rate was in the zone farthest from the NL rite, 39 percent. Response rateswere similar la tht other three cones closer to the NL site, respectively 51 percent In theclosest zone, 60 percent in the neit closest, and 53 percent in ths next closest. One mustassume that there were no systematic biases In recruitment related to important factorsthat affect child blood lead, such as socio-economic status or behavior,

10. Percentage of explained variability In the logarithm of blood lead is not a

useful criterion in model assessment, since it depends on the range and the distributionof predictor variables within the data set. In comparison with an other studies of childblood lead data that EPA has performed, including urban and rural sites, active andinactive lead smelter or lead mining sites, the child blood lead data from the MadisonCounty study has a higher percentage of explained variation (40 percent) than mostother studies* and environmental lead explains n comparable percentage of Taritnc* (18percent in our analyses) to other inactive smettitr sites. Both the magnitude andstrength of the relationships between blood lead and environmental lead art comparableto those we have seen at other sites. The linear regression relationships for blood leadvs. lead in sod, dust, and drinking water are statistically significant In an appropriatemodel specifications.

3. CALCULATION OF SQTTf T.1C/H1 r.TJEAN^P CONCENTRATIONS.

3.1. SUMMARY AND CONCLUSIONS! SITE-SPECinC PROPERTIES FOR RISK

1. The NL/Taracorp site appears to haw properties that are characteristic ofother recently inactive lead smelter sites. The areas closest to the site have sofl and dustlead concentrations that are appropriate to alrtorae particulates from smelter emissions.These partldtt are generally easily transported from exterior soil into household dust,and are IQcsly to be small, soluble, and highly bioavailabte.

2. These analyses, phis observation of Granite CUy neighborhoods closest to theNL site, show that:

(D there are many young children in tlte community,(ID children often play outdoors for much of the day,(111) residential yards often contain large bart areas without grass cover,(!T) adjacent yards are often not fenced and art rtadfly accessible to young

children,(v) the residential areas are surrounded by Industrial areas and by

transportation routes that contribute to die total environmental impact on these

TT,

1. Sito-spedfk parameters wsrt based on our judgement and analyses that theNL tit* had many points of shnflariiy to the calibration site, Mldvak, and that it Isappropriate to assume no mft<g?H»g factors tint may reduce childhood exposure to dustand soiL

2. Hie default parameters with an assumed soQ-to-dust coefficient of 70 percentprovided a very good fit to the blood lead data, in terms of geometric mean blood toad,percenters of the blood kad distribution, and ireasonabto correlation between observed

PRELIMINARY ASSESSMENT OF DATA FROM THE MADISON COUNTY LEADSTUDY AND IMPLICATIONS FOR REMEDIATION OF LEAD-CONTAMINATEDSOIL

Allan H. MarcusEnvironmental Criteria and Assessment Office

U.S. Environmental Protection AgencyResearch Triangle Park, NC 27711

1. INTRODUCTION

The data on blood lead, environmental lead, and family interviews for 490children in Madison County, Illinois, were provided to us by the Institute for EvaluatingHealth Risks (IEHR) through the U.S. Department of Justice. The study was carriedout in 1991 by U.S. EPA, Illinois EPA. the Illinois Dept. of Public Health (IDPH) andthe Agency for Toxic Substances and Disease Registry (ATSDR). We requested accessto the data from IDPH, which was provided by their contractor, IEHR. EvaluatingHealth Risks (IEHR). The data were sent to us on diskette In ASCH format. Weconverted these data Into a SYSTAT (WOkinson 1992) data file, from which allsubsequent analyses reported here were performed. Additional analyses will requirecreation of SAS data sets, Analyses of the data were reported by IDPH (February 1994)and in a more compact form by ATSDR (May 1994), U.S. EPA provided a criticalreview of these analyses (Marcus et aLff May 1994).

The purposes for our reanalyzes of the data are:

1. To assess the results described in (IDPH 1994) for use in erahiatinf childhood leadexposure in Madison County;

2. To provide site-specific information about relevant parameters in the EPA IntegratedExposure, Uptake, and Bloklnetic Model for Lead (IEUBK Model);

3. To evaluate the proposed soU remediation level of 500 ppm using this recentInformation.

This is only a very preliminary report of results. A more detailed report wfll beprepared, describing the methods used in the analyses, a complete set of mutts, and thebasli for our conclusions. This report is dMded into the following sections. Section 2report? the remits of the preliminary analyses. Complete technical details will follow ina subsequent report. Section 3 presents the bads for a sofl lead cleanup level. Section3.1 deescribe* the empirical bask for die findings, which combines personal observationof the she, results of data analyses, and professional Judgements about the tite. Section3.2 describes the soil lead results when input rabies for the IEUBK Model from Section3.1 are used, and describes the reevaluatton of the soil lead cleanup level for the she*

2. SUMMARY AND CONCLUSIONS FROM DATA ANALYSES

A Urge number of graphs using these data are attached. The following resultsare based on these figures. Note that with few exceptions, the logarithms of theenyironmental lead concentrations and blood lead concentrations were used since thedistribution of each variable it skewed to the right. AD are "natural1 base e logarithms,not common base 10 logarithms. The error bars on the graphs show the standard errorof the (geometric) mean. The exact distance of the household from the NL site b not hithe data set, only location within (approximate) 1/8 mfle rings around the smelter. Noinformation fa provided about the quadrant or direction from the smelter site, and noinformation b provided about the city or township where each child Urea.

We reanalyzed data on 490 pre-school age children in 351 households la MadisonCounty, Illinois. These data contain no information on location within the study area,apart from approximate distance of the child's household from the NI/Taracorp site("NL site"), and in particular do not identify the child's community of residence, so theresults cannot be ascribed to any locale such as Granite City, Madison* or Venice* Ouranalyses of these data showed that:

1. Sixteen percent of the children had blood concentrations of 10 ug/dl orgreater In the Madison County study area as a whole, but the percentage of lead-burdened children increased with decreasing distance from the NL anther site. TheIncrease was from 7 percent of children Bring at a distance greater than 0.75 mDss fromthe NL site to about 26 percent ef children in the study who lived In die area closest tethe site* The areas closest to the site would bo considered as appropriate forremediation, based on remedial investigations, by current EPA criteria. (Figure 1)

2. The percentage ef children in the study with blood lead concentration! of 15ug/dl or greater increased from about 1 percent of children living at a distance greaterthan 0.75 rnOes from the NL site to about 9 percent of children within 0.4 mOea of diesite. None of the children living at a distance of a mfle or more from the NL site had ablood lead of 20 ug/dl or grater, whereas about 5 percart of the pre-echool childrenwho UTW! within half a mfle from the NL site hid a bk^ Ablood lead concentration ef 20 ug/dl is wett almve EPA'i level of concern of 10 Bg/dl andIs associated with a substantially increased risk of permanent and irreversibleneurobehaTioral damage. Current CDC guidelines recommend individual environmentalor medical Intervention with children whose blood lead concentrations are at least 20ug/dL (Figures 2 and 3),

3. Blood lead concentration, soil lead concentration, and house dust leadconcentration show very similar patterns of decreasing concentration with increasingdistance from the NL site, on average. Loadings of deteriorating lead paint Inside andoutside the house show little or no relationship to distance from the ML site. Thissuggests that deteriorating lead-based paint k not the most important environmental

Tao<e 3 —OstntKAoo o' BJooo le*a Levels 'or Me*»can Americans Aged 4 to 74 v«rs ay Age Category. Sei. and income L*vet: ?982 •» "9ft* (Hispanicand Nutrtxxi Examination Survey ;HHAN6SD ano '988 to '991 Pfase i of tr»e Tn.nd Natxx^ai Heartf) and Nutntion Exa/rw^tton Survwy NHANES HI pnase *;)

1982-1984

1988-1991

Ages 4-5 11982-1964

I96e-<99l

Ages 6-19 y1982-1984

1988-1991

Ages 20-74 y1982-1984

1988-1991

Males1982-1984

1988-1991

Females1982-1984

1988-1991

income level, towt1982-1984

1988-1991

incorr>e lev*, rrndt1982-1904

1988-1991

income wvei. r»g»iT1982-1984

1988-1991

No.

6682

3611

269

J49

2331

1188

3062

2074

2638

1797

3044

1814

2*60

1664

2032

1024

674

393

Geometric Mean.ymotA. (yfl/dL)'

0*1851

0141301

053(109)

017135)

039(80)

0 12(2.5)

042(8.7)

015132)

050(104)

0.19(40)

034(70)

0.11(22)

0.42(88)

016(33)

0.40(8.3)

0.13(2.6)

0.39(8.1)

0.11(2-3)

ConfidenceInterval,

pmoVL (pgVdL)

0*0-042[83-87)

012-017- 2 5 - 3 5 )

050-058(103-11 5)

0 '4-021(2M3)

038-040(7 8-82)

0 10-0-15(2 0-3 2)

0*0-043(83-90)

0.13-0 18(2.7-3.7)

0*9-051(102-105)

0 16-023(33-48)

032-035(67-72)

009-013(1 8-2.7)

0 42-0 44(66-91)

013-0 19(2.7-40)

038-042(79-8.7)

0.11-0-15(22-3.1)

037-039(7 6-8.6)

0.09-0.14(1.8-2.9)

t5th

0 19(40)

(<! 0)

024(50)

<<1 0)

0-19(40)

<0.05

0.19(40)

<005(<10)

024(5.0)

0.06(12)

0.14(3-0)

<005

0.19(40)

<0.05(<10)

0.19(40)

<0.05

0.19(40)

<0.05

10*

024(50)

005(1 D

029(6.0)007(1 *)

024(50)

<0.05

0.24(5.0)

005(11)

079(60)

0.08(18)

o.ie(40)

<005(<1 0)

02*(50)

0.08(171

074(5-0)

<oos

074(5.0)

<0.06

Percent

25*

079(60)

009(1 9)

039(80)

012(25)

0.29(6.0)

0.08(16)

0.29(60)

0.10(2.0)

0.39(8.0)

0.12(2.5)

074(5-0)

0070.4)

079(60)

0.10(2.0)

079(6.0)

0.08(16)

079(6-0)0.00(16)

Hie*, pmott.

5001

043190)

016133)

053(11 01

0 18(361

0.39(80)

0.14(28)

043(90)

016(34)

0.48(100)

070(42)

0.34(7.0)

(24)

0.43(90)

0.17(3.6)

0.43(90)

0.14(2.91

039(9-0)

0.12(2-5)

.(pg*u75*

058(120)

02fl

068(1401

078(59)

053(110)

073(47^

058(12.01

028(57)

088(140*

031

043(901

019(3.91

0-58(12-0*

0-26(5.0)

0.58

073(48)

053

0.19

90th

077(160)

040(63)

092(190)

0.40(83)

0.68040)038<74)

0.77(160)

0.42(86)

0.87(180)

0.45(9.4)

058(12.0)

031(6.4)

0.77(16.0)

043(90)

0.77(16.0)

0.36(7.5)

0.72(tS.O)

070(5-71

95th

087(180)

051HO 6)

t 11(230)

0.4»(99)

082(170)

047(98)

092(190)

054(11 1)

1 01(210)

057(11 6)

066(140)

0.41(8.4)

0 92(190)

0.54(11 1)

087(180)

0.44(92)

0.82(170)

0.35(7.3)

•For eacfi groupmg, tr» geometnc meem from KHANES w) NHANES HI pnase i are subs&caffy different (Pt01).tincome wvet »*a defined by poverty-«xome ribo (PIR) categonzed u low (CKPiFki 30). rad (1 30iPiB<3.00). end Ngh (P1R23-OO).

housing. The distribution of blood leadlevels in the NHANES reflects expo-sure in the general population, whereasstudies focusing on high-risk populations,such as persons living in older, deterio-rating housing, may find a different bloodlead distribution. Data from nationalhousing surveys indicate that in 1980about 242 million (30,3%) occupiedhouses in the United States were builtbefore 1940 when lead-based paint wasin common use. By 1989. this numberhad decreased by 3.4 million to 20.8 mil-lion (2L2%), suggesting that populationexposure to lead-based paint may havedecreased slightly.21-2* On the other hand,the continuing deterioration of lead-based paint in existing houses could in-crease the likelihood of exposure for per-

sons in the 20.8 million households whoremained in these older houses. On apopulation scale, it U not clear whetherthe net effect is an increase or decreasein exposure to lead-based paint.

The consistent decline in blood leadlevels across broad population catego-ries of age, sex, race/ethnicity, urbanstatus, and income level most probablyreflect changes in exposure to majorpopulation-wide lead sources. In addi-tion, selected population groups withinthe United States are likely to have ben-efited from other changes in exposure,such as reductions in lead in communitywater supplies and reduction of leademissions from local industry.

The public health impact of the ob-served decline in blood lead levels of the

US population is dramatic, especiallyfor children. The change in the propor-tion of children aged 1 to 5 years withblood lead levels 0.48 umol/L (10 ug/dL)or greater was at least 70% for non-Hispanic whites, non-Hispanic blacksand Mexican Americans. Although thedecline in blood lead levels is encourag-ing, the number of children with leadlevels 0.48 uraoI/L (10 ug/dL) or greaterremains substantial and disproportion-ately higher for noo-Hispanic blackchildren (one in five children), as dis-cussed in the accompanying article inthis issue.*

At least 99.8% of lead in gasoline hasalready been removed; and domesticallyproduced cans are no longer lead sol-dered. Therefore, to achieve additional

290 JAMA. July 27. 1994_VOI 272. NO 4 Decline r Stood Lead Levels—Pirkle et aJ

Tawe 4—Percentage of Mexican Americans *g«0 4 to 74 Yean at or Above Selected Blood Lead Level Cutoffs by Age. Sex. and income Lever 1982 to 1984(H.spanic Heattft and Nutntwn Examination Survey) and 1988 to 1991 (Phase 1 of the Third National Heaitn and Nutntwn Examination Survey)

Blood L«ed Laveta of Population Group, %\

All persons1962-1984

1968-1991

Ages 4-9 y1962-1964

1964-1991Ages 6-19 y

1982-19641966-1991

Ages 20-74 y1962-1964

1966-1991

Males1962-1964

1986-1991FsmaJee

1962-19641966-1991

income tevei. sow*1962-1964

1966-1991Income level, ma*

1962-1964

1966-1901

Income level, high*1982-19641986-1991

No.

5682

Mil

269

349

23311168

30832074

2636

1797

3044

1514

24601664

20321024

874

303

i21.48 wmoVL<230M«m)

0.4

0.0

24

00

03

o.o

04

0.0

0.6

00

0.1

0.0

0.5

0.0

0.4

0.0

0.2

0.0

21 J1 umoVL

1.30.2

49

00

050.1

1502

2.10.2

0.4

0.1

1.4

0.1

1.30.0

0.8

0.0

20.97 timoVL(22(> pg/dL)

360.4

88

0.0

2.0

0.4

42

0.4

«2

0.5

1.1

0.3

4.1

0.4

34

0.3

2.7

0.0

20.73 timoM.(21 S P9/4L)

128

15

247

0.1

90

09

14.1

19

21.122

4 5

0.7

14.61 7

11.41.1

10.0

0-0

20.41 timoVL(210 ygrtfl.)

41259

«1 5

49

356

45

429

66

584

8.7

23 S

28

45.2

73

38042

358

16

I20 J4 pmoVL

912289

964

327

901

238

91 5

310

966

405

856164

922

33.5

90.123-6

91.2147

•income *v« wu defined by povettyHncome ratto (Ptfl) categorized u low (0<PIRS1.30). nmd (l.30sPiR<3.00), and high

reductions in blood lead levels in the USpopulation, sources other than lead ingasoline and lead in solder need to be

Reference*)L Agency for Tone Substance* and Diaeaac Reg*istry. The Nature and Extent of Lead Pmeoning mChildren in the United States, A Report to Cow*yreu. Atlanta, Ga: US Dept of Health and HumanServices. Public Health Service; 1968.2. Mahafley KR, Annest JL, Roberta J. MurphyRS. National estimates of blood lead level*: UnitedStatea 1976-1960. .V Engl J Med. 1982307:573-679.1 Aw^JUFIitfcJUMakucD.NeeaeJW.BayseDO. Kovar MG. Chronological trend in blood leadlevels between 1976 and 1960. N Engi J Med. 1988;306.-137S-1377.i. National Center for Health Statiatica. Plan andoperation of the second National Health and Nu-trition Survey, 1976-80. Vital Health Stat i. 1961;No. 15. US Dept of Health and Human SerrieaaPublication PHS 81-1317.S. National Center for Health Statistics. Plan andoperation of the Hispanic Health and Nutrition El-imination Survey, 1982-84, Vital Health Stat 1.1986;No. 19. US Dept of Health and Human Ser-vicee publication PHS 86-1381.C. National Center for Health Statiitka,E»tiTM,MasaeyJT, WataburgJ, Chu A, Maurer KR. Sampledesign: third National Health and Nutrition El-imination Survey. Vital Health Stat t. 1992£io.113. US Dept of Health and Human Services pub-lication PHS 92-1387.7. National Center for Health Statiatka, AnnestJL, Mahafley K. Blood lead levels for persons 6months-74 year* of age. United State*. 1976-80.Vital Heal* Stat 11. 1984;N<x 223. US Dept ofHealth and Human Services publication PHS 84-1683.

addressed further. The major remain-ing sources are lead in paint and leadthat has already accumulated in dust

8. Carter-Pokna 0, Pirkle JL, Chavez G. GunterE, Blood toad tevels of 4-11 year old Mexican- Ameri-can, Puerto Rkaa, and Cuban ctuldren. PublicHeaU* Rep. 1990; 106 88 98.I Brody DJ, PtrUe JL, Kramcr ElA, et aL Bkxidlead kveta in the US population: pfaaae 1 of the thirdNational Health and Nutrition Examination Sur-vey (NKANES III, 1988 to 1991). JAMA. 1994;272277-281.II Centen for Diaeaae Control laboratory pro-cedure* Vied by CA* Clinical Ckfrniftry Diwvm,Centen fa /XseaM Control, for Uu Second Na-tional Health and Nutrition Exatninatum Survey(NHAXES W 1979-1990- Atlanta. Ga: Centers forDisease Control and Prevention; 1981.1L Grater EW.MflkrDT. Laboratory rVxa&TwUted fry tiu Divinon of Envrnnmental HealthLaboratory Science*, Center for EnvtrmrnentalHeattk, Centen far Diteaet Control, for the Hie-panic Health and Nutrition Examination Survey(HHANKS) l98t-»4. Atlanta. Ga: Centers for Dis-ease Control and Prevention; 1986,11 Mffler DT, Paschal DC, Guntaf EW, StroudPE, D'Angeio J. Determinatioa of lead in bloodusing electrothermal atonuaattoc. atomic absorp-tion spectrometry with a L'vov p4atform and ma-trix modifier. Analyst 1967:112:1701-1704.11 SAS Institute Inc. SAS La*g-**e* Vemo* t,Cary, NO SAS Institute Inc; 19SO.14. Shah BV, BarnweU BG, Hast PN, Lavang*LM. SUDAAN U**r'» Manual. Releate 5.50. Re-search Triangle Park, NC: Resenren Triangle In-stitute; 1991.14. Environmental Protection Aipency. Quarteriy

and soil Without efforts to reduce theseexposures, population blood lead levelsare unlikely to continue to decline.

Summary of Lead Pkaiedovn Reporting Data.Washington. DC: Office of Mobile Sources, Office ofAir and Radiation, US Environmental ProtectionAgency. 199L1C Centen far Diaeaae Control Preventing LeadPottontng ta Young Children. A Statement by theCenten for Dieeaee Control Atlanta, Ga: US Deptof Health and Human Serricea, Pubhc Health Ser-vice; 199L17. Environmental Protection Agency. Air QuoJ-Uy Cntena for Lead. Research Triangte Park, NC:Office of Health and Environmental Assessment;1966. Environmental Protection Agency report

18. Can Maaufacturers Institute, ^ood and SoftOrvnk Can Shipment!. Washington. DC: Can Manu-facturer* Institute; 1992.19. Bolger PH. Carrington CD. Capar 9G. AdamsMA. Redactions in dietary lead exposure in theUnited Suus. Chem Speciation B+oavadalnlity1991 31-3120. Adams MA. PDA total diet studr dietary in-takea of lead sad other chenucala. C*»« SptciotumBioavailab&ty. 1991 37-41.21. US Dept of Commerce. US Dept of Housingand Urban Development. Annual Houting Survey.1990: Part A: General Housing Chcrattenttie*.WashmgtQK, DC: US Dept of Commerce; 1982. Cur-rent Housing Report* series H-150-80.21 US Dept of Commerce. US Dept of Housingand Urban Development. American Houfing Sur-vey for the United State* m 1989. Washington, DC:US Dept ef Commerce; 1991. Current Housing Re-ports series H- 160-89.

JAMA. July 27. 1994—Vol 272. No 4 Dedtne n Stood Lead Levete—Pwvie « ai

'awe 2 —Percentage o* Persons *^eo i -o "4 *«ars at or Aixjve Setecieo Bkxxj Lead Lev* Cuioflj Cry Age. Se«. Hace/EtJvucrty. Uffcan Status and income Lev*:yn.iM States '976 10 1580 .Second National Heartri and Nutntxxi Exarmnaoon Survey) and 1988 to 1991 (Phase 1 o< me Third NaoonaJ Hearth ana Nulnoon Ex-amination Survev)

Stood L**d Laveta at Population Group, %

1976-1960

1968-1991Ages 1-5 y

1976- 1980

1988- '991Ag«$ 6-19 y

1976-1980'988-1991

Ages 20-74 y1976-1980

'968-1991Males

1976-1980

i 988- i 991

Females1976-1980

1988-1991

1976-1980'988-1991

Non-Hispantc blacks1976-1980

'983-1991Non-ceoiral Crty

1976-19801988-1991

Ceriirai crty, <1 melton1976-1980

1988-1991CeniraJ Crty >1 mrflion

1976-19601986-1991

income level, low*1976-1960

1968-1991income levet. mid*

1976-1960

1968-1991income level. h*gn*

1976-196O

1988-1991

NO-

9632

12119

2271

2234

2024

2963

5537

6922

4895

6051

4937

6068

68164307

12593274

7112

7495

1612

2909

11001379

2548

4106

4176

4050

2764

2781

1>1 45 ymo*/L(>30 (j dL)

i 9

02

4 1

04

06

00

2 3

03

330 4

06

0 1

1 7

0 2

28

02

1 9

0.1

1 9

03

181 1

29

0.5

1 7

02

1501

21-21 JJfTVOrt.(>25 po'dL)

52

0 4

93

05

2 4

02

590 4

900.7

1 6

0.1

480 4

8.4

04

4 9

03

61

03

60

1 4

6.809

46

03

5.1

03

SSS35-149

08

247

1 1

8-2

0-4

16.7

0.7

241

1.1

82

0-2

140

0.6

22912

13.9

06

171

0.6

18.4

1.9

16.0

1.6

13.8

0.5

14.50.4

*0.72 ymoi/1.{215 p^dL)

377

1.1

526

2 7

277

08

403

1.1

53.1

1.9

23.00.4

360

o.»

50.9

2.8

353

0,9

43.1

1.8

444

29

306

2.8

36.30.9

3630.6

20.46 ixnoVt(210 M dC.)

778

43

M89

71 7

26

794

4 4

89668

66.7

1 8

76.9

36

66.4

8.5

75.7

35

62.1

5-9

648

9.8

78.4

8.8

76,134

798

2.7

i>0.24 ymoVl.

992

233

998

332

99.1

12.2

992

255

998

335

98.7

132

992

21 1

997

33.7

99.0

21 7

998

269

999

360

902

326

993

22.9

994

18.4

-income level «u defined by pov*ry-<naxne ratio (PiR) categorized as tow i.O<PlR<i 30). mid (1 30iPlB<3 00), and hi£i (P1R23.00).

dL) or greater between NHANES IIand NHANES III phase 1 was con-sistent across the entire age range(Fig 4).

The percentage of the population withblood lead levels at or above selectedvalues is presented in Table 2. Theselevels were chosen in part because oftheir prior or potential use in publichealth policy. For those aged 1 to 74years, the prevalence of blood lead lev-els 0.48 ujnol/L (10 ug/dL) Or greaterdecreased from 77.8% in NHANES IIto 4.3% in NHANES III phase i. Forchildren aged 1 to 5 years during thesame time frame, the prevalence of bloodlead levels 0.48 ujnol/L (10 ug/dL) orgreater decreased from 88.2% to 8.9%.

The change in percentage of children ator above selected lead levels fromNHANES II to NHANES III phase 1is presented in Fig 5,

Separate analysis by race/ethnicity re-vealed that geometric mean blood leadlevels declined by 77%, from 0.66 to 0.15umol/L (13.7 to 3.2 ug/dL), for non-His-panic white children and by 72%, from0.97 to 0.27 umol/L (20.2 to 5.6 ug/dL),for non-Hispanic black children. Theprevalence of blood lead levels 0.48timoUL (10 ug/dL) or greater for chil-dren in this same age group declinedfrom 85.0% to 5,5% for non-Hispanicwhite children and from 97.7% to 20.6%for non-Hispanic black children.

Mean blood lead levels decreased from

0.73 to 0.20 nmol/L (15.2 to 4.1 ng/dL)for children aged 1 to 2 years and from0.71 to 0.16 ^imol/L (14.8 to 3.4 ug/dL)for children aged 3 to 5 years. Duringthe same time period, the prevalence ofblood lead levels 0.48 umol/L (10 ug/dL)or greater also decreased from 88.3% to11.5% for children aged 1 to 2 years andfrom 88.1% to 7.3% for children aged 3to 5 years.

Mean blood lead levels decreased by60% (1.16 to 0.47 umol/L [24.0 to 9.7ug/dL]) for non-Hispanic black childrenfrom low-income fiunflies living in thecentral cities with populations of 1million or more. This compares withan overall decrease in mood lead lev-els of 75% (0.72 to 0.18" umol/L [14.9

288 JAMA. July 27 1994—Vd 272. No 4 Decline in Blood Lead Levets—P<rkie et al

to 3.6 ug/dL]) for all children aged 1to 5 years.

From HHANES (1962 to 1984) toNHANES III Phase 1 (1988 to 1991)

The HHANES was conducted from1982 to 1984. between the second andthird NHANES. Geometric mean bloodlead levels were also found to be inter-mediate between the estimates of thetwo national surveys. The blood leadlevels of Mexican Americans fromHHANES were lower than overall lev-els observed in NHANES II, but not aslow as levels of Mexican Americanssampled in NHANES III phase 1.

Geometric means, 95% confidence in-tervals, and percentiles of the blood leaddistribution of Mexican Americans be-tween HHANES and NHANES IIIphase 1 are presented in Table 3. Mexi-can Americans showed an overall de-crease in geometric mean of 60%, from0.41 to 0.14 mol/L (8.5 to 3.0 ug/dL).The geometric mean for children aged 4to 5 years declined from 0.52 to 0.17umol/L (10.9 to 3.5 ug/dL). As demon-strated in the comparison of NHANESII to NHANES III phase 1 estimates,the size of the decrease in blood leadlevels was similar in both sexes andacross age groups and income levels.

The proportion of the Mexican-Ameri-can population at or above selected bloodlead levels is shown in Table 4. Overall,prevalence of blood lead levels 0.48umol/L (10 ug/dL) or greater amongMexican Americans decreased from41.2% to 5.9%. The percentage of chil-dren aged 4 to 5 years with blood leadlevels 0.48 umol/L (10 ug/dL) or greaterdecreased from 61.5% to 4.9%. Theseresults demonstrate that one in 20 Mexi-can Americans aged 4 to 5 years con-tinue to have blood lead levels of healthconcern.COMMENT

The data from two national surveys ofthe US population, conducted more thana decade apart, demonstrate a substan-tial decline in blood lead levels. As theconsequence of a shift in the overall dis-tribution of lead levels, fewer personshave blood lead levels in the upperranges. The decrease in mean blood leadlevels was observed for the total popu-lation and within all race/ethnicity, sex,urban status, and income level subgroupsexamined in this article. The prevalenceof blood lead levels 0.48 umol/L (10 ug/dL) or greater also decreased sharplyfrom 77.8% to 4.3%.

As discussed herein, exposure to leadfrom major population-wide lead sourcesdeclined between 1976 and 1991. Con-sistent with this decline, the blood leadlevels observed in HHANES (1982 to

I5

>1.21 21 45(>25) (>30)

Blood Lead Level Cutoff, nmoYL

Fig 5,—Perwntag* of children aged 1 to 5 years at or above selected Wood lead levels: United Slates, 1976to 1980 (second National H«arth and Nutrition Examination Sun/fly [NHANES il]) and 1988 to 1991 (phase1 of the inird National Health and Nutrition Examination Survey [NHANES Hi prase 1]).

1984) were intermediate between levelsfound in NHANES II (1976 to 1980) andNHANES III phase 1 (1988 to 1991).Consequently, the magnitude of de-crease from HHANES to NHANES IIIphase 1 (65%) was less than fromNHANES II to NHANES III phase 1(78%). The percentage of MexicanAmericans with blood lead levels 0.48umol/L (10 ug/dL) or greater declinedfrom 41.2% to 5.9%. In NHANES IIIphase 1, both mean blood lead levels andthe prevalence of blood lead levels 0.48umol/L (10 ug/dL) or greater of Mexi-can Americans were closer to those ofnon-Hispanic whites than to those ofnon-Hispanic blacks.

The decline in blood lead levels seenin these national surveys is consistentwith the results of other studies of en-vironmental lead levels,1 which indicatethat a continued reduction in exposureto lead sources began in the late 1970sand continued throughout the 1980s. Be-tween 1976 and 1991, the three majorsources of lead exposure common to thegeneral population were lead in gaso-line, soldered cans, and paint In 1976, atotal of 186.47 million kg (205810 tons)of lead was used in gasoline in the UnitedStates." In 1983, this amount haddropped to 51.59 million kg; (56 940 tons),and in 1990, lead used in gasoline hadbeen reduced to 0.47 million kg (520tons).1* From 1976 to 1990, the amountof lead used in gasoline decreased 99.8%.The reduction of lead in gasoline La mostlikely the greatest contributor to theobserved decline in blood lead levels dur-ing the period of the national sur-veyi.1*1*"

Lead from gasoline and soldered canscontribute to lead in food. Since gasolinelead enters food through multiple path-ways,1 UJ* it is difficult to make a quan-

titative estimate of the reduction in foodlead that resulted from decreasing leadin gasoline. The amount of lead used insoldered cans decreased markedlythroughout the 1980s. In 1980, 47% offood and soft drink cans were lead sol-dered. By 1985, this figure had droppedto 14%, and by 1990. only 0.85% of foodand soft drink cans were lead soldered.l*As of November 1991, lead-soldered foodor soft drink cans were no longer manu-factured in the United Slates.18

The Food and Drug Administrationuses "market-basket" surveys to esti-mate the average daily intake of leadfrom food for various population groupsin the United States.1* For 2-year-oldchildren, these surveys estimate thetypical daily intake of lead to havedropped from 30 ug/d in 1982 to 1.9 ug/din 1991.l1JDThe Environmental Protec-tion Agency estimated in 1986 that about42% of lead in food came from lead-sol-dered cans,1 Thus, reducing the amountof lead used in soldered cans has likelybeen a major factor in reducing foodlead levels. Although it is difficult toquantitatively determine the decreasein blood lead levels attributable to re-duced amounts of lead in soldered cans,the decline in the amount of lead used inthis source probably contributed sub-stantively to the observed decline inblood lead levels.

The manufacture of lead-based paintwas limited to less than 0.06% by weightin 1978 by the Consumer Product SafetyCommission.1 Individuals who haveleft housing with lead-based paint orwho reside in lead-abated homes havereduced their lead exposure. Still, lead-based paint remains a problem, pre-dominantly in oWer, deterioratinghousing.1-*" The NHANES do not spe-cifically target persons who live in such

JAMA. July 27, 1994—Voi 272. No 4 Decline m Blood Lead Levels—PirWe et al 289

—Oistntxrtxxi o* Stood Lead L«v**a (or Persona Aged 1 to 74 vears by Age Category Sex RacevEthnerty. urban Status and income Level: United States'960 'Second NaoooaJ Health and Nutrrtion Ejamtnabon Survey [NHANES H|) and i98fl to 1991 iP*ase 1 o< the Th*rd National Hearth anc NuVrtion £j-

Survey ;NHANES ill Phase 1J)

Al 3*r$ons•376-1980

'368-1991

*5Mi-5y•376-1980

•968-'99l

Agei6-I9y•376-1980

•968-1991

AOM 20-7* y•376-1980

•388-1991

Ua«s'376-1980

•988-1991

Fe -aies'976-1980

'988-1991

Non-Hisoa/nc wtntes'376-1980

1988-1991

Norwrtisoan«c t*ac*s1976-1980

'968-1991

Non-<emraJ city1976-1980

'988-1991

Central city, <1 m*ion1976-1980

1968-1991

Central cay. 21 rraann1976-1980

1968-1991

lncom« i*v«r lowT:976-1980

•988-1991

incorw !*ve<. nw»T••976-1980

1966-1991

lncom« lev*, htgtrf1976-1980

1968-1991

NO.

9632

12119

2271

2234

2024

2963

5537

6922

4895

6051

4937

6066

6616

4337

1259

3274

7112

7495

1612

2909

1108

1379

2546

4106

4178

4050

2784

2781

Geometric Mean,pmoiA. (ng/dU"

062(128)

0.14128)

071(150)0.17(36)

056(11 71

0.09(1 9)

063(131)0 14(30)

0.72(150)

0.18(3.71

054(11 1)0.10(2.1}

0.61(12.6)0.13(2.71

070(1*5)017(3.5)

060(125)

013(2.7)

0.66(13-6)0.14(2.9)

0.67(13.9)

0.19(39)

0.03(13-1)0.16(3.4)

0.61(12-6)0.13(2.7)

063(13.0)0.12(2-5)

95%Confidence

Interval,ymolrt. (pg/dL)

060-065(124-133)OU-0 .15(27 -30 )

0 67-0.75(142-158)

0 16-0.19133-40)

054-060(11 2-12.4)

006-011(1 7-22)

061-066(127-13.7)

0 14-0.15(28-32)

070-075114 $-155)

0.17-0-19(3 5-3.9)

051-055(106-11 5)

0 10-0 11(2.0-2.2)

056-0.63(12.1-13.1)0.12-0 14(2.2-2.8)

066-0.74(13.7-15-5)

0.16-019(33-3.9)

0.56-0.64(12.0-13.1)0.12-0.14(2.5-2.6)

061-0.70(12.7-14.5)

0.12-016(2.5-3.4)

0.61-0.73(12.7-15.1)0.1 7-0 .21(3.6-4 3)

060-0-67(12.4-13.8)

0.15-0.18(3.1-3.8*

0.58-063(12.1-13.1)0.13-0.14(2.6-2.9)

0.60-0.65(12.5-13.5)

0.12-0.13(2.4-2.7)

Ptrcentlte*. ymoWl. (MQ/dL)i

5th

034(70)

<005(<10t

039(8.0)

0-05(1.1)

0.29(60)

<005(<1 0)

034(70)

<005(<1 0)

039(8.0)008(12)

029(8.0)

<0.05(<1 0)

029(6.0)

<0.05(<10)

0.39(8-0)

<O.OS<<1.0)

0.29(6.0)

<oos<<VO)

0.34(7.0)

<0.06<<10>

0.34(7.0)0.08(13)

0.29(6.0)

<0.05(<1 0)

0-34(7.0)

<O.OS(<1.0)

0.34(7.0)<005(<VO)

10W

039(8.0)

005(1 0)

043(9-0)0.07(15)

0.34(70)

<0.05(<1.0)

0.39(8.0)

006(1.2)

0.43(9.0)

0.06(16)

0.34(70)

<0.05(<10)

039(6.0)

0.05(1.0)

0.43(9-0)

0.08(13)

0.34(70)

0.05(10)

0.39(8-0)o.os(10)

0.43(9.0)0.09(1.8)

034(70)

0.06(1 3)

0.39(8.0)

0.08(1-0)

0.39(8.0)

<O.09<<10)

25th

048(100)

009(1 8}

058(120)

0.11(22)

0.43(90)

006(13)

048(100)

0.1012.0)

056(120)

0.12(24)

0.43(90)

0.07(1-4)

0.46(100)

0.08(1.7)

0.53(11.0)0.1 1(2.2)

0.48(10.0)

0.09(18)

053(110)0.09(1.8)

0.53(11-0)0.12(2.5)

0.46(10.0)

0-10(2.1)

0.48(10.0)

008(1.7)

0.48(10.0)

0.08(1.7)

50th

063(130)0 1 4(30)

072(150)

0.18(3-71

056(12.0)

0.10(2.1)

063(130)

0.15(3.2)

072(150)

0.18(38)

0,53(U 0)Oil(2.3)

063(130)0.14(2.9)

0.72(15.0)

0.18(3-7)

0.63(13-0)0.14(3.0)

0.68(14.0)0.14(3-0)

0.68(14.0)

0.19(40)

063(13.0)0-17(36)

0.63(13-0)0.14(2.9)

063(13.0)0.14(2.8)

75th

082(170)

023(48)

092(190)

028(59)

072(15.0)

0.17(35)

062(170)0.24(5.0)

092(190)

028(58)

068(140)

0 18(3.8)

0.77(160)

022(45)

092(190)

0.28(59)

0.77(160)022(4.6)

087(18-0)

0.25(5.2)

0.87(18.0)

0.29(8-1)

0.82(170)

028(5.8)

0.77(16.0)

023(4.7)

0.82(170)0.21(4.3)

90th

1 01(210)

035(73)

1 16(240)

046(96)

092(190)

026(54 )

1 06(22.0)

0.36(74)

1 16(240)042(8.7)

087(180)

0.28(5.7)

1 01(210)

033(6.8)

1.11(230)

0.45(9.3)

1 01(21.0)

0.33(6.9)

1 06(220)

040(8.3)

1 08(22.0)

0.48(99)

1 11(230)

045(94)

101(210)0.34(7.1)

1.01(21.0)

0.30; (6.3)

95th

i 21i250)

0*5( 9 4 )

1 351280)

059(122)

1 06(220)

036( 74 )

1 25:260)

046i95)

1 30(270)052

(109)

097(200)

036(74)

1 16(240)

0.43i89)

1.30(270)

058(12 1)

1 16(240)

0*3(89)

125(.260)050

(10.4)

1.21(250)064

(132)

i 25(26.0)057

(H 8)

1.18(240)

044(9.1)

1 21(250)

0.39(8.0)

•For eac* grouping, me geometric meane from NHANES I and NHANES III phase i are stataocalty different (PcOl).tevel was defined by poverty-income ratio (P1R) cwgonzed u low (0<P1R<1.30). rmd (1.3OsPlfi<3.00). and high (P1R*3.00).

286 JAMA Jury 27. 1994—Vol 272. No 4 Decline m Blood Lead Levels—Pirkie et al

Ail Persons ~|

Age* 1-5 yAges 6-19 y

Ages 20-74 y

Mates -|Females

Non-Hispan»c WhiteNon-Hispanic Black

Non-Central CityCentral City, <1 MillionCentral C'ty. 21 Million

Low PIRM.d PIR

H.gh PIR

20 40 60 80Decrease in Geometric Mean Blood Lead Levels. %

(1976 Through 1980 to 1988 Through 1991)

100

Fig 3.—Percentage decrease <n geometnc mean Wood lead levels for persons aged 1 to 74 years Dy agecategory, sex. '•aceetftnicrty. urban status, and income level: United States. 1976 to 1980 (second NabonalHeaith and Nutnbon Examination Survey) to 1988 to 1991 (phase 1 ol the third National Heaim and Nutn-tion Examination Survey). Income fcvW defined by poverty-income ratio {PIR) as tow ((kPiR<! 30). rmd(1 3CKPtR<3.00). and high (PIR23 00).

0.77(16) -,

0.68(14) \

3 0.58(12)I3 0.48(10)

NHANES II ; NHANES HI Phase 1

I

n

1-2 3-5 6-11 12-19 20-29 30-39 40-49 50-59 60-74

Rg 4.—Geometnc mean Wood lead ***** tor persons aged 1 to 74 yean by age UnrtwJ Staia*. 1976 toi960 (second Natonal HeaJtti and Nutrition Examination Suvey (NHANES II]) and 1988 to 1991 (phase 1of tn« third Nator* Health and Nutrition Exammatkvi Survey [NHANES Ml pnase 1D-

veys, the blood lead measurements werecalibrated using standards preparedfrom lead nitrate Standard ReferenceMaterial 928 obtained from the NationalInstitute of Standards and Technology,Gaithersburg, Md< The consistent use ofStandard Reference Material 928 for cali-bration assured a common accuracy baseacross surveys.Demographic and SocloeconomtcCovarlatM

The trends analysis included stratifi-cation by five sociodemographic vari-ables: age, sex, race/ethnicity, urban sta-tus, and income level. Age was defined

in years and categorized as 1 to 5 years(4 to 5 years for Mexican Americans). 6to 19 years, and 20 to 74 years for analy-sis. Race/ethnicity was allegorized asnon-Hispanic black, non-Hispanic white,and Mexican American. Because of smallsample sizes, persons not defined bythese three largest US race/ethnicitygroups were included only in the overallestimates.

Definitions for income level and ur-ban status were based on those previ-ously determined by the US Bureau ofthe Census. Income level was definedby the poverty-income ratio (PIR): thetotal family income divided, by a poverty

threshold. The PIR wa> divided intothree categories: low i(kPIR< 1.30), midU.30<PIR<3.00), and high iPIR>3.00)Urban status was categorized as non-central city, central city with populationless than 1 million, and central city withpopulation 1 million or greater.Statistical Analysis

Survey-specific sample weights wereused in all statistical analyses. Geomet-ric means and percentiles of blood leadwere calculated using SAS.'3 Log10 trans-formed blood lead levels were usedto normalize the distribution of bloodlead levels. Geometric means were cal-culated by taking the antilog of themean logic blood lead levels. SUDAAN.14

a statistical software package that in-corporates the sample weights and ad-justs for the complex sample design ofthe survey, was used to estimate theSEs.

RESULTSThe results of the trend analysis in

blood lead levels are presented in twoparts: first, the change in blood lead lev-els from NHANES II (1976 to 1980) toNHANES III phase 1 (1988 to 1991),and second, the change in blood leadlevels for Mexican Americans fromHHANES (1982 to 1984) to NHANESIII phase 1(1988 to 1991).

From NHANES II (1976 To 1980) toNHANES 111 Phase 1 (1988 To 1991)

The different distributions of bloodlead levels for those aged 1 to 74 yearsfrom NHANES II and NHANES IIIphase 1 are presented in Fig 1. A declineof approximately 0.48 umol/L (10 ng/dL) occurred in the geometric meanblood lead level as well as a clear changein the shape of the distribution. Whenthe sample was limited to children aged1 to 5 years, similar results were ob-served (Fig 2).

The geometric means, 95% confidenceintervals, and percentiles of the bloodlead distribution for the total populationand stratified by the five sociodemo-graphic factors are presented by surveyin Table 1. For the total population, thegeometric mean decreased by 0.48umol/L (10 ng/dL). Stratification of thedata showed that the size of the de-crease was fairly constant across sex,race/ethnicity, age groups, urban sta-tus, and income levels.

The decline represents an overall de-crease in blood lead levels of 78% forpersons aged 1 to 74 years and a de-crease of 70% or more for selected sub-groups (Fig 3). Children and youthsaged 6 to 19 years showed the great-est decline in blood lead levels. How-ever, a decline of 0.48 umol/L (10 ug/

JAMA. Jufy 27. 1994—Vol 272, No 4 Decline m Blood Lead Levels—P<rxte et a: 287