Tobacco Documents Online

\'ol 32: No. 2 LONG-TER1`f EFFECTS OF EXPOSURE TO LEAD -NEEDLEMAN ET AL. 83 THE LONG-TERM EFFECTS OF EXPOSURE TO LOW DOSES OF LEAD IN CHILDHOOD An 11-Year Follow-up Report HERBERT L. NEEDLEMAN, M.D., ALAN SCHELL, Af.A., DAVID BELLINGER, PH.D., ALAN LEVrrON, M.D., AND ELIZABETH N. ALLRED, M.S. AbstL,act To determine whether the effects of bw-level fidence interval, 1.4 to 40.7) and of having a reading dis- lead exposure persist, we reexamined 132 of 270 young ability (odds ratio, 5.8; 95 percent confidence interval, 1.7 adults who had initially been studied as primary sdhool- to 19.7) as compared with those with dentin lead levels children in 1975 through 1978.1n the earlier study, neuro- <10 ppm. Higher lead levels in childhood were also sig- behaiiicxal functioning was found to be inversely related to nificantly associated with lower class standing in high dentin lead levels. As compared with those we restudied, school, increased absenteeism, lower vocabulary and the other 138 subjects had had somewhat higher lead grammatical-reasoning scores, poorer hand-eye coordi- leveisl on earlier analysis, as wea as significantly lower nation, longer reaction times, and slower finger tapping. 10 scores and poorer teachers' ratings of classroom be- No significant associations were found with the results of havior. i0-b 'ifier e ts neurobetiayro al fundjoing~Lead levels When the 132 subjects were reexamined in 1988, im- were tnversery related to self-reports of minor deGnquent pairme nt in neurobehavioral function was still found to activity. be related to the lead content of teeth shed at the ages We conclude that exposure to lead in childhood is asso- of six and seven. The young people with dentin lead levels ciated with deficits in central nervous system functioning >20 ppm had a markedly higher risk of dropping out of that persist into young adulthood. (N Engl J Med 1990; ...high sdhool (adjusted odds ratio, 7.4; 95 percent cort- 322:83-8.) ~r1;THIN the past three years, the Environmental Y V Protection Agency and the Agency for Toxic Substances and Disease Registry have concluded in policy statements that lead at low doses is a serious threat to the central nervotu systems of infants and childiren.t•2 These policy statements have been based on a;3rowing convergence of results from both epide- miologic and experimental studies of lead toxicity in the United States, Europe, and Australia." Whether the afY'r.cts on the central nervous system of exposure to low doses of lead that have been observed in infants and children persist has received limited attention. Only three follow-up studies have been published to date, and the longest follow-up has been five years.''tt No data have yet been reported on whether early dis- turbances influence functional abilities in later life. In 1979 we reported that first- and second-grade children without symptoms of plumbism, but with elevated dentin lead levels, had deficits in psychomet- ric intealigence scores, speeeh and language process- ing, attention, and classroom performance.3 When they were studied in the fifth grade, the children with high &ntin lead levels had lower IQ scores, needed more s?ecial academic services, and had a significant- ly higher rate of failure in school than other chiidren.9 «'c have now evaluated the neuropsychological and academic performance in young adulthood of 132 of Frwn tt e Schoe( of Mediciaee, Univasey of PittsbtuEh. Pittsbiugti (H.L.N.k Bosina utuvasny. eotton (A.S.): and the lkrtoep~demioio6y Unit. Oiildtcn•s Hospital and Harvard Medicat Schod, B® (D.B.. A.L. E.N.A.). Address reprtnt raluau to Dr. Needleman at the Ueirasiry of Pittsburgb School of Medicire. a'estera Psyctiatric Insbtute and C7inic, 3811 O'Hsn St.. Pittsburgh, PA 15213. SupQxted by a`rant (ES 4eA95) fromtbe National tnstitute of Enrironmenul Health Scierm. Df.' Betlinaer's wok .as svpponrd by a Research Gireer De- vefoprr.7. Award (ES 0013E) durint the eoeoc6x of this study. Pmeruex! in put u the annuai n==g d the Society for Pedianic ResearttY Amaiem Pediatnc Soeiety. WasAmaroa. D.C.. May 4, 1989. the original sample of 270 subjects, and we report the relation of their recent performance to their exposure to lead, as measured 11 years earlier. iAETFiODS Sampte The initial sample was chosen from the population of 3329 chil- dren enrolled in the first and second grades in the Chelsea and Somerville, Massachusrns, school systems between 1975 and 1978. Of this population, 7 percent provided at least one of their shed primary teeth for lead analysis. From this sample of 2335 children, 97 percent of whom were white, we identified 270 from English- speaking homes whose initial dentin lead levels were either >24 ppm or <6 ppm. These children (mean age, 7.3 years) underwent an extensive neurobeha.ioral examination. More teeth were subse- quently collected and analyzed, and the subjects whose teeth were discordant with respect to lead level according to a prion aiteria were exduded from the data analysis. Also excluded from the analy- sis were children who had not been discharged from the hospital after birth at the same time as their mothers, who had a noteworthy head injury, or who were reported to have had plumbism.s In a Later reanalysis, conducted in response to suggestions from the Environmental Protection Agenry,t= the tooth lead level was treated as a eentinuous variable. A mean dentin lead level wu computed for each subject from all the teeth collected. The edu- sionary factors previously used were found not to be related to outcome scores. The subjects initially ezduded were therefore not excluded from this follow-up sampte. The 270 subjects tessed from 1975 to 1978 constitute the base population for this repot. From old research records, telephone directocies, town records, and driver's-license rolls, we lacated 177 subjects. Of these, 132 agreed to partiapate, and the re- maining 45 dediaed.'I"he subjects were paid $35 each and received travel expenses. Ten subjects tested in 1988 had been excluded from the analysis reported in 1979 because their parents stated at the time of testing that the children had elevated blood lead Icvcls or had undergoneehelatioa for lead poisoning. This group is discussed separately in this report. The mun age of the 132 subjects at the 1988 reexarnination was 18.4 years; the mean length of time between the t.ao examiriaions was 11.1 years. All but four subjects in the eurrent follow-up study were white. No dininl manifata- uons oC lead exposure were recorded in the earlier intcrhews for the 122 subjccts who were not treated with chelating agenu.

84 THE NEW ENGL,lND JOURINAL OF MEDICINE The research protocol and informedtonsent procedurn w"e ap- provcd b} the tnstttuttonal review boards of the Children's Hospital of Pittsburgh and the Children's Hospital, Boston. Informed con- scnt was given by all the subjects or their parents. Classification of Lead Exposure All the dentin lead levels measured from 1975 through 1977 were used to compau an arithmetic mean lead concentration for each subject. The lead burden was treated in two ways: as an iaterval variable in finear regressions and as a categorical variable - i.e., high (>20 pprn), medium (10 to 19.9 ppm), and low (<10 ppm)- in the logistic regressions described below. Lead levels in .ertous blood were :measured at the time oC the rt:examination to estimate current apixure. This practice was discontinued after the fsrst 48 subjects wac tested, because none had a lead level exceeding 0.34 µmol per liter (7 Ecg per dealiter), well bdow the Centers for Dis- use Control's definition of undue lead exposure of 1.25 µnod per liter (25 µg pt.r deciliter). Behavioral Evalt:atiot The subjects were evaluated individually by a single tustiner, who remained, blinded to their lead-exposure status until all the data had been coded and entered into a computer data ba.se. All assessments were carried out in a fixed order; the duntion of the testing was about two hours. NeurobeHaviora/ Evaluatan System The subjeets completed an automated assessment battery in which they iutd a personal computer, joystick, and response key.'s We selected the following items from the battery for evaluation: Coatituurtu Jrrfonnatea lat.t s Syrnbehdiht nrbJtituti~t, an adaptation of the Wechsler itunrt for computer administration. Xand-,~;K raordinatioa. Using a joystick to move the etasor, the subject traced over a@arge sine wave generated on the esonitor screen; dc.iatioru from the line (rnot mean square error) were recorded. Sunpt< oi+aa/-rcartioa tuse. Subjects pressed the response k., when an O,appe:red on the s=ern; the interval before the stimu- lus was vuied randomly. Finln tapDu3. The subject pressed a response button as many times as pc~ssible during a 10-setond period; both hands were tested. Pattmt aumorr. The subject was presented with a eomputer- genented pattern formed by a 10-by-10 array of dark and bright elements. After a brief exposure, the subject was presented with three patHUru, only one of which was identical to the original pattern.'I?te numbcr:of correct responses and the lengtsof time to the correct choice were recorded. Patura m,npartsort. The subject was presented with three eom- puter-generated patterns on the l0-by-10 arny. Two were identi- cal, and one dt(iered slightly from the other two. The subiea was required to select the nonmatchirtg pattern. Seria!-dipt tcarni,t3. The subject was presented with a srirtg of 10 digits, then asked to enter the string into the eomputer. After an error, the same stimulus was presented, and the scmod trial began. IorabYLm. For each of 25 words, the subject chose tts word most ncaiiy synonyrnous from a list of four choices. Crmnimuu:a1 rcaroau,S. The subject was presented with a pair of letters, A and B, whose relative position varied. Then the screen deared, ani'~ the lette=s were replaced by a sentence that described the order of the letters. The sentence might be active or pas- sive, atfirnt:ntive or negative, tnu or false (examples are 'A fol- lows B" ard 'B is not followed by A"). The subject had to choose the correct sentenms, and the number of err3rs was rccorded. Jan. 11, 1990 Sloitckia3 ettrnteoa. The subject was required to ehoose which key to press in response to three dt(ferent instructioru. In the "side' trials, the subject had to press the key on the same side as the stimulus. In the "directton" trials, the correct choice was the direction in which an arrow pointed. Before each trial'ct the third set, the subject was told whether to choose the side the arrow was on or the direction in which it pointed. M.ed su4s. This test was derived from the Profile o( Mood States.'s Five scores were computed for tension, anger, depres- sioa. fatigue, and confusion. The following tests were also used to evaluate neurobcitavioral functioning: Cakfornia Verbal Learning Test The California Verbal Learning Test" was used to assas multi- ple strategies and processes involved in verbal learning and mern. ory. Scores for immediate and delayed recall were also obtairted. Boston Naming Test In the Boston Naming Test," the subject was presented with 60 pictures in order of increasing difficulty and asked to name the objects shown. Rey-Osterrefth Complex Figure Test The Rey-Osterreith Complex Figure Test" was used to evaluate visual-rrtotor and visual-spatial skills. The subject was asked to copy an abstract geometric figure and then to draw it 6rm memory both immediately and after 30 minutes. Accuracy and or=artiution scores wese talculated. avorti-edentrtcation Test Form B from the Woodcock Reading Mastery Test.ru used to evaluate rrxding skill. Gnde-equivalency scores were calculated from raw scores. Reading disability was defined as iadinted by scores two grade levels below the score expected on the basis of the highest grade completed. Self-Reports of Delinquency The subjects completed a structured questionnaire frnm the Na- tional Youth Survey's that included scales for minor atstis®eial be= havior and for violent crimes. Review of School Rec+ords High-school records were obtained for all but two of the subjects tested. Class size and rank, the highest grade completed, and the number of days absent and tardy in the last full semester were recorded- Students who were still in the 11th grade at the time of testing were not induded in analyses of the highest grade complet- ed. Class rank was computed as I - (class nnk(dzss size). Ststistksl Analysis To evaluate whether the participants in this follow-up evaluation werc reprrsentative of the original cohort, subjects who.rere tested and not tested in 1988 were compared in terms of variabits reported in 1979, iadudinE dentin lead levels, covariates tsa related to lead exposurc, teuchers' ratings of el.assroom behavior, and IQ scores. In addition, we carried out separate regressions of deatie lead level against IQ score as measured between 1976 and 1978 br subjects tested and not tested in 1988. We then performed a regression on both groups taken together, entering both a dummy tam for partia- pation in the current follow-up (yes or no) and a lead-levd-by- partieipation status term. To evaluate the relation between early exposure to Irad and each of the continuously distributed outcome variables, subjects were drssified according to dcntin Icad-level quartiles, and ecsn scores, adjusted for covariates, were computed. Ordinary latu-squares lin-

322 No 2 LONG-TERM EFFECTS OF EXPOSURE TO LEAD -NEEDLE.MAN ET AL ., rceression, N.nh the mean or log-mean dentin lead level as the rrs.n c(icct. %.as used to estimate the significance of the relation. G_:comes th,at were signifiantly associated with lead exposure in r.<s< bi~anacc analyses tvere funher evaluated by multiple regres- -:- anaksts. Ten covariates were included in the model. They were c moihcr's at;e at the time of the subject's birth, the mother's cr _-_auonal levcl, the mother's IQ, family size, socioeconomic sta- r_s a tNo-facter Hollingshead index), sex, age at the time of test- .n<, birth order, alcohol use, and whether the subject and the moth- e-. icft the hospitat together aRer the subject's birth. The lead -easure (the mean or the log of the mean) that produced the best- 4.::ca model (highest R2) is reported. Five of these eovariates were er.:pioyed in the first study of these subjects and shown to be inAu- e:c:a!. Five otha~n (sex, age at testing, prolonged hospitalization as a r.eonaee. birth order, and current alcohol use) were added to the mxcl on the basis of prior knowledge of their effects on psyehomet- nc iunction. l,ogistic-regeession analysis was used to model the as- sxiation of lead level and two outcomes treated eategorinlly (fail- t:re to graduate from high school and reading disability). In this a:.aivsls, we eonurolled for the eovariates listed above. Two indiea- tor variables were used to represent the three exposure groups. 0~dds ratios and 95 percent confidence intervals, adjusted for eovar- ia:es. were computed for. the high-lead-level group, with the low- ksd-level group used as the reference group. RFSVLTs Selection Bias The 132 subjects who were retested in 1988 (Table 1; were not rt.presentative of the group of 270 subjects tes:ed in 1979. The subjects we retested tended to have sl:zhtIv lower dentin tead levels, more highly educated fa.;tilies of higher socioeconomic status, and mothers <<::h higher IQs and better obstetrical histories; a higher proportion of the retested subjects were girls. Irn addition, they had had fewer head injuries and had sicnificantlv higher IQ scores and better teachers' ra:ings as reported in 1979. The slope of the regres- sion of childhood IQ on dentin lead level was steep- er in the group not tested in the follow-up study, al- t::sugh the difference from the slope in the group we re-ested was not statistically signifi- ca-t (F = 1.82, 1,196 df; P = 0.18). Academic and Neurobehavioraf Outcome Table 2 shows the covariate-ad- ;~sted scores of the 122 subjects t• ,no did not have clinical plum- bism, according to their dentin lead ccricentrations. Table 3 summa- rizes the resul ts of modeling the re- lation between early exposure to lead and outcome by multiple re- gression. Earlier exposure to lead was significztntly associated with di- minished academic suceess. Among children with dentin lead levels >20 ppm, as compared with those whose dentin lead levels were <10 ppm, the unadjusted odds ratio for failure to graduate from high school was 4.6 (95 percent confidence in- terval, 1.2 to 17.4). fldjustment for Table 1. Comparison of Subjects Tested and Not Tested in 1988.' CM.u.CTEasnc TESTED (N - 132) NOT TESTEO fN - 1381 P v'.ut Lead•kvel group (g/ Low 50 47.8 Middle 22.7 16.7 - High 27.3 35.5 0.7t Birthader 2.3_1.6 2.8_1.9 0.016 No. of live hitths 2.111.5 3.2_ 1.6 0.05 Fadter's education (yr) 12.2=2.6 11.4_2.6 0.009 MaRhrs's educatae (yr) 12.0x2.2 11.1 s2.1 0.0005 Modcr', IQ 112st5 108x15 0.017 Motlxr's age at subyett's 25.5x5.9 25.3t5.8 0.7 binh (yr) Father's age at subjcn's 28.3=7.8 28.8_7.9 0.6 b+rth (yr) Crazation (wk) 39.9z2.0 40.0s 1.7 0.7 Birth veight (g) 3776x608 3712s600 0.40 sea (S) Female 55.3 42.8 Mak 44.7 57.3 0.04 Had injuries (56) 3.8 8.7 0.09 Teachers' ratings (1979 sum scnre) 9.3=2.8 8.2:3.6 0.009 Full-scale IQ (1979) 1073=14 99.5_15 0.001 •tim-minu+ alun me mean :SD. ?By du-wqure bw ta all lead-ie.cl traqs covariates increased the odds ratio to 7.4 (95 percent confidence interval, 1-4 to 40.8). Higher dentin lead levels were also associated with lower class rank, increased absenteeism, lower scores on vocabulary and grammatical-reasoning tests, significantly slower finger-tapping speed, longer reaction times, poorer hand-eye coordination, and lower reading scores. In subjects with dentin lead levels >20 ppm, the unad- justed odds ratio for having a reading disabilit.•, de- fined by a score two grades below that expected for the highest grade completed, was 3-9 (95 percent confi- Table 2. Outcomes in Young Adulthood /'`Ccording to Dentin Lead Concentration in Childhood.' OvrcoKt Va!.tsu ts.a CoMCa.Mta..noM No. of subjorts LpYQi (G7.9 pQm) 30 LOW WA-L<s ppm) 31- • MIGM 0}TL2 pRn) 30 MI6MLTT (>22.: Rlal 31 Reading score (words trad eorreetly) 143.8 142.7 140.2 135.2 Reading grade equivaknt (gade level) 12.2 11.9 11.2 10.1 Highax grade achie.ed (grade level) 11.7 11.9 11.5 11.3 Class stardin8 (peroendk) 0.60_ 0.59 0.48 045 Absertce hocn school (no. of days/ 12.0 12.0 17.9 20.8 semcster) vocabutary (words correR) 18.0 16.4 17.6 14.6 Gramrnuieat reasoning (no. 13.1 13.0 12.9 15.8 inconea) Hand-cyc c®adinatiaot 5.1 5.4 5.5 6.2 Reactiort time (msoc) Preferred lsand 246.6 255.5 267.3 275.1 Noaprzfrncd hand 241.2 238.2 258.4 261.2 Finger uppint (noli0 sec) 46.6 47.2 45.9 43.5 -tAe wbjeQ, .ae di.;eed :o Qa+ps .cmdss o lod-k.el..aka. 7be ra,urs rhw. ,:< lewsq,ure stun 8ma- aher ad juSMma rQ av.ram. s.D;nCS .d+ dr.ai rlMMbne. !,r•e Ym ad+ded. trbr rs-eyc oecau,om- hner "=be" saa mc e~

THE NEW E\CLAND JOLR\AL OF \tF:DICI\E Jan 11, 1990 Table 3. Regression of outcomes in Young Adulthood on Dent n Lead Levels in Childhood.' .1. TCOUt ~ •o:•fLE . @rv~E1nTE REGEC3S10N ASUETnnt Rtutss-0n R= I.lsM[it9 uT/M.Tt SE P ..LtZ R" P.l.METtt ESTIMATE $E P vAIUE t;~chcst En:c ach/cyed 0.061 -0.027 0.009 OOOE 0.319 -0.027 001 0.013 RcamnE l:ax equrvalent 0.121 -0.07 0.01$ 00001 0.229 -0.072 0 021 0 001 ("3ss sunc~:g 0039 -0.006 0.003 0.049 0.24: -0 006 0 003 0048 4osence fra= school' 0.071 4.8 1.7 0.006 0.209 4.73 1.8 0.01 Gnmrnauca, reasoning 0.051 0.159 0.062 0.012 0.197 0.178 0 06i 0.011 k ocabular. 0,108 -0.124 0.032 0.000 0.324 -0.122 0.033 0.001 Flneer cap,.se 0031 -0.104 0.05 0.05 0.336 -0.133 0 05 0 01 Hand--c.e ccrotdtnaaon 0.043 0.041 0.018 0.02 0.195 0.048 0 019 0.01 Reaction eme" Prefer:e: Isand 0.025 11.3 6.66 0.08 0.242 12.9 6 3 0 042 .I;onpreier~ed hand 0.03 11.5 0.05 0.056 0.229 10.3 5.5 006 `.ttnor antaa:ul behaaort 0 025 -0.639 0.36 0 082 0.306 -0.739 0.35 0 038 •7he folk-r co•anaus re*e cootnalkd fo. In 11r nwhlple rcEresslon analysls aee. sea. bink nrder. famJ7 ue. a+ahef's are a+ 41c ssOXet•s Mrth. knrA or the seor,asak sa% rthe hospual. ed+aa: x le.c~. r•,+~•''+ IQ .x•m:ommK sulu.. and evzvent akoAd rx. 'The natua to~ of the mun dentin kad k.tl ras wed as tl+e matn effect. dence ix:terval, 1.5 to 10.5). Adjustment for covariates increased the odds ratio to 5.8 (95 percent confidence intervaP. 1.7 to 19.7). For most outcomes, neither the size of che lead regression coefficients nor their stand- ard errors %:ere substantially changed by adjustment for covariates. Of t'ht, 10 children with clinical plumbism (who either underwent chelation or were reported to have had elevated blood lead levtlc), 3 of 7 (43 percent) dropped out before graduating from high school (3 others ane still in school), and 5 of 10 (50 percent) have rcading disabilities. W'hen the children with plumbism were grouped with the other subjects ac- x 4.01. 3:- ~~- 25 - /7 r~- ~ hE.GRTtD rEUY(•11Y ts- I 7/28 z : i 5/31 I 2/27 ; 0/27 f <: 95 5 95-a 2 8.3-22.22 >2222 Dentin Lead Level (ppm) Figure 1. Th > Proport on of Subjects Who Did Not Graduate from High School, Classified According to Their Past Exposure to Lead. Asymptomatic subjects are classified according to lead•level quamles Seven of the 10 subjects who were earlier reported to have c1m cal plumbism are shown in a separate column. No schoot records were found for two subjects. One subject was not tested but reported that she had graduated from high school. (There are therefore 121 subjects represented in this figure.) Ten subjects (three with reported plumbism and seven asymptomat,c subjects) are still attending high school and are therefore not shown here The numbers in each column indicate the number who did not graduate and the total number in the category. cording to quartiles for dentin lead le.•els. a dose- response relation was evident for both outcomes (Fig. I and 2). Early exposure to lead was not significantly associ- ated with performance on the symbol-digit or serial- digit tests, the continuous-performance test. pattern memory or pattern comparison, switching attention, the California Verbal Learning Test, the Rev-Oster- reith figures, the Boston Naming Test, or mood scores. The lead level was inversely related to the summed score on the se1C report of delinquency questionnaire, which consisted primarily of reports of minor antiso- cial behavior. When subjects were divided into two groups ac- cording to their dentin lead levels (<10 ppm vs. =10 ppm), high dentin lead levels predicted future failure to graduate from high school with a sensiti%it.• (=SE) of 0.71 =0.12 and a specificity of 0.61=0.05 (Table 4). DISCUSSION In this extended follow-up study, in which the mean length of follow-up w2s 11.1 years, we found that the associations reported earlier between lead and chil- dren's academic progress and cognitive functioning persisted into young adulthood. The persistent toxici- t.• of lead was seen to result in significant and serious impairment of academic success, specifically a seven- fold increase in failure to graduate from high school, lower class standing, greater absenteeism. impairment of reading skills sufficiently extensive to be labeled reading disability (indicated by scores two grades be- low the expected scores), and deficits in..•ocabularv. fine motor skills, reaction titne, and hand-e.e coordi- nation. A number of issues require consideration when one is interpreting the data reported here. The first is the influence of sclection bias on the associations .:e ob- served. The subjects retested in 1988 had more favor- able characteristics than those v.•ho could not fx locat- ed or who declined to participate. The subjects who wcrc not retested tended to have had 1Tighcr lead Iev-

\'oI 322 No 2 LONG-TERM EFFECTS OF EXPOSURE TO LEAD-•NEEDLE'`tAN ET AL. 87 4/30 4/31 6/30 12/31 0 1 >2222 <:595 59542 8.3-2= Dent,n Lead Lev®l (ppm) 5/10 r[.011TlM /LVY.1s1q Figure 2. 1-he Proportion of Subjects with Reading Disabilities. Classified Acoording to Their Past Exposure to Lead. Asymptom,atic subjects are classified aocording to lead-level ~quart le, and 10 dliLdren with a history of clinical plumbism are shown sep arately. Reading disability is defined as indicated by a reading level two or, more grades betow the expected level. The numbers in each column indicate the number with a reading dis- a bility and the total number in the category. els, lower socioeconomic status, and lower IQ scores and teachers' ratings of classroom behavior. The in- verse relation between dentin lead levels and IQ re- ported in 1979 was stronger for the subjects who were not retested in 1988 than for those we retested, al- though the difl'erence did not reach statistical sig- nificance. This finding is in agreement with the obser- vation, made by us and others, that children from families in lower socioeconomic groups are more vul- nerable ta the effects of lead than children from more favored economic backgrounds.0 We infer that the estimates made on the basis of the data on the 132 subjects we restudied are likely to be conservative. Indeed, had all the original subjects been located and retested, the magnitude of the effect of lead exposure might have been even greater. Is the nature of the relation between lead and later outcome causal, or does it result from confounding by other va.riables? The association between lead and outcome reported here meets six criteria for valid causal inference: proper temporal sequence, strength of association, presence of a biologic gradient, non- spuriousness, consistency, and biologic plausibility?' In this study, the exposure to lead preceded the school failure and the reading disabilities measured. The strength of the association, as measured by ad- justed odds ratios of 7.4 and 5.8, was substantial. A dose-response relation has been demonstrated be- tween exposure and numerous outcome variables (Ta- ble 2, Fig. 1 and 2). "Nonspuriousness" indicates that the association observed is not due to confounding. In this analysis, we controfled for both the covariates that were identified in 1979 as potential confounders and others we ;suspected were important. The magnitude of the eifeet of lead was reduced only slightly, if at all, by this procedure. The zero-order correlation between sccioeconomic status and dentin lead levels in this samplc kas not great (r = 0.04). Many covar- iatcs that were important contributors to performance in the early grades (e.g., the mother's IQ and the mother's educational level) had less efTect on the subject's performance in young adulthood. The re- sults, morcover, are consistent with those of several other studies by workers who have reported lead- associated deficits in reading`•22" and carly classroom behavior."•2S The lead-related deficits in IQ, speech and language processing, and attention reported in 1979 provide plausible mechanisms by which lead could impair performance in class and produce even- tual failure. Similar effects on learning have been demonstrated in the experimental studies by Gilbert and Rice of subhuman primates.' In these investiga- tions, rhesus monkeys, administered lead only in the first 100 days of life, had impairments in learning as adolescents. In adolescence, the mean blood lead level of these monkeys was 0.73 µmol per liter (15 ~.eg per deciliter). The value accepted as the threshold for lead-engen- dered neurotoxicity in children has declined steadily over the past decade as more sophisticated population studies, with larger samples, better designs, and better analyses, have been conducted.`-s,lla2,2,,z6z9 When this study was begun in 1975, the toxic level of lead in the blood was defined by the Centers for Disease Control as 2.0 µmol per liter (40 Ag per deciliter). In 1973, the mean blood lead level in a subsarr^t-_ of 23 chil- dren chosen from among those with the highest dentin lead levels in an earlier study was 1.7 µmol per liter (34 µg perdeciliter).3 None of our subjects were symp- tomatic. That these subjects were exposed to high doses of lead after the original study H••_s completed is unlikely. Lead exposure, the incid,.,ice of pica, and hand-to-mouth behavior diminish after the fifth year of life. The low blood lead levels found in these sub- jects in young adulthood (all <0.034 µmol per liter) provide convincing evidence that their later exposure to lead was not excessive. The consensus on what level of lead is toxic has changed in recent years. After reviewing the studies published up to 1987, the Agency for Toxic Sub- stances and Disease Registry defined the threshold for neurobehavioral to)ucity as 0.5 to 0.7 Er.mol per liter Table 4. Sensitivity and Specificity of the Dentin Lead Level in Childhood as a Predictor of Failure to Graduate from High Schoo/.• HY'N-so+ool ca.cu.now tL.w L1rnL alo n,. <lo rrv No 10 4 Ya 39 61 Semitivity = t0'(to+4) - 0.71 Specificiry - 6t/(61+39) - 0.61 `Or thc 12.2 asympxoevec sO-ysty eudied. 7 wb,ens uM .ve wU .aeadmt iduet r U+e nme or tt,b vulr.u .vs encludra. Oae %+rbjaa'i uhod raroe~ .ae .a fowd. Of U,e 132 aa)eeo rctcurd u I962, tfx 10 .,b N,tiu! pi.aRals r..e eoes -•d•,••,n.

88 THE NEW ENGLAND JOURNAL OF MEDICINE Jan. 11, 194Q (l0 to 15 Feg per deciliter).' The agency estimated that 3 to 4 million American children have blood lead lev- els in excess of 0.7 µml per liter. The mean blood levt:l among our subjects with high tooth lead levels, estimated in 1979 from a limited lead-screening pro- gram, was 1.6 fcmol per liter (34 ~cg per deciliter) (ranc;e, 0.87 to 2.6 Amol per liter [18 to 54 µg per decil,iter)). For subjects with low tooth lead levels, it wa,s 1.2 µmol per liter (24 Ecg per deciliter) (range, 0.513 to 1.7 µmol per liter [12 to 36 µg per deciliter)). Thus, the lead levels in the reference sample used in the calculation of the odds ratios for one high-lead- level group were relatively high by contemporary standards. ' The data presented here indicate that exposure to lead, even in children who remain asymptomatie, may have an important and enduring effect on the success in life of such children and that early indicators of lead burdt:n and behavioral deficit are strong predictors of poor school outcome. For the small group of 10 sub- jects who were diagnosed earlier as having plumbism, the outcome was espedally dire; half of these young peo;ple have reading disabilities, and almo,;t half left high :9chool before graduation. Given the federal esti- mate;, that 16 percent of children in the United States have elevated blood lead levels (>0.7 µmol per liter [15 µg per deciliter)), the implications of these find- ings fDr attempts to prevent school failure are intrigu- i: _. The practical importance of early detection and abar.ement of lead in the environment, before it enters the fx>dies of children, is borne out by these long-term findings in young adults. «'e atre indebted to Drs. Richard Frank, Constantine Gatsonis, Alan Virskv, and Rolf Loeber for their careful review and critiques of the rnanuscript and to bfs. Pat Hadidian for her careful work in finding subjects and reviewing records. RfF'ERENCFS I Agrncv for Toxic Substances aad Disease Registry. The nature and extent of k:acl poisoning in children in the United Stattx: a report toCongrest. Atlanti Ckpanment of Health and Human Services. 1988. 2. Air quality cntena for lead. Rescarch Triangle Park. N.C.: Eavironmental Pro ecuon Agency. 1986. 3. Needleman HL. Gunnoe C. Leviton A. et al. Deficits in psyetaio;ic and ciassroom performance of chdd:en with elevated dentine lead levels. N Engl I Med 1979: 300:689-95. 4 t FuNon M. Raab G. Thomson G. Luen D. Hunter R. Hepburn W. tnfluence of blood lead on the ability and saainment of children in Edinburgh. Lancet t98'7: 1:t221-6. 5. Hansen ON. Trillingsgud A. Beese 1. Lyngbye T. Grandjeaa P. A neuro p;ycholoc cal study of children . th elevated dentine lead kvel: assesement o1' dx effect of lead in differrnc socioeconomic groups. In: Lindber8 SE. Hutchinson TC. eds. Heavy mcrals in the environmenc Intemariarul Con- Icrnxe. New prleans. Edinburgh. Scotland: CEP Consultants. 1987:54. 6. Betlinger D, tevison A- Wfe^m-%'* C. Needleaun H, Rabinowitz M. Loo- gitudinal analyscs of prca2ul and posfiatal kd esposure and earty cogni- tive developrncnt. N En;l I Mcd 1987; 3t6:I03743. 7. Gilbert SG, Rice DC. Low-kvd lifetime kad expowso produces behaviora) toxicity (spatial disaiminuion reversal) in aduY rnookeys. Toxicol App( -15harmacd l987:9I:4bt-90. S. Cory-Siechta DA. Weiss B, Co: C. Delayed behaviot5) toxicity of kad with increasing exposua eoeoerwuiots. Toxieol App! Phamsaeol 1983: 71:342- 52. 9. Bellinger D. NeaQkmat; HL. Bromfield R. Mintz M. A follow-up study of the aademic anaiameat and classroom belsavior d childrsa with ekvated dauine lead kveJs. Brot Trace EJem Res 1986; 6:207-23. 10. Ernhart C. Landa B. Schel! NB. Subclinieal kxh of kad and develop• rneraal de6cit n trsuhiivariase follow-up remasmmt. Pediaaies 19111; 67: 911-9. I t. Schroeder SR. Hawk B, Ono DA, Mushat P. IFiCts RE. Sepantint the efieets of lead and social factors on IQ. Eav'voa Res 1985: 38:144-5I. 12. Needleman HL. Geiger SK. Frank R. Lead and IQ seoecs: a neinalysis. Science t985: 227:7014. 13. Baka EL, Letz RE. Fidler AT. Shatat S. Ptuuannrs D. A eomputer-based ncurobehavioral and evaluation system for occupr.ional and environmental epidemiolo=y: nxtfiodotop rtnd validatioe smdxs. Neutebehav Toxico! Teruol 1985:1:369-77. 14. Rosvold HE. Mirsky AF. Sarason I. Bransome ED Jr, Bedc LH. A continu- ous peiormance test of brain darruge. J Consuh Psycho! 1956: 2t}.3d3- 50. 15. MeNair DM. Lar M. Droplemm LF. Errs eunasl - profik of awod states. San Diego: Educational rtnd Testing Service, 1971. 16. Delis DC, Knrner JH. Kaplan E. Ober BA. The California verbal karn. ing test - research editioa. San Antonio: The Psychological Corporation. 1986. 17. Kaplan E, Goodglass H. Weintraub S. Boston naming test. Philadelphia: Lea & Febiger. 1983. 18. Rey A. L'examew psyclwlogique dans ks e>s d'etcephalopuhie tnuma- tique. Arch Psychof 1941; 28:286b340. 19. E)Iiot DS. Huizinp AD. Atetoo SS. Explaining delinquency and drug use. Beverly Hills. Cafif.: Sage Publicadoas, 1985. 20. Bellingcr D. Leviron A. Watunaux C. Nesdkraaa H. Rabinowitz M. Low level lead expossue. social class and infant developrrent. Neurotoxicol Tera- to! 1988; I0:497•503. 21. Hill AB. The envimnn+eru and disease: assoeiarioe or causation? Proc R Soc Med 1965; 58:295-300. 22. Fergusson DM. Fcrjusson JE. Horwood U, Kiama NO. A longitudinal study of dentine lead kvels, intelligence. school perfornsance, and behav- iour. II. Demine kad ud cognitive ability. J Chi)d Psycfwt Psychiatry 1988: 29:793-809. 23. Yule Q. Lansdowa R. Millar IB. Urbuawicz MA. The relationship tx- tween blood lead coacauruioas. intelligwa and utaitunerx in a school population: a pilot study. Dev Med Child Neurol 1981; 23:367-76. 24. Hataakis A. Kokknvi A. Katsouyanni K. et al. Psyehomesric intelligence and anentional petiornunee deficits in lead-exposed children. in: Lindberg SE. Hutehinsat TC. eds. Heavy meals in the enviroenxnc Intematior} al Conference, New Orleans. Edinburgh. Scotland: CEP Consultanrs. 1987:204-9. 25. Yule Q, Urbanowies MA. hnsdowts R. Mil1u IB. Teachers' ratings of children's behaviour in relatioa to blood lead levels. Br1 Dev Psychot 1984; 2:295-305. 26. Hawk BA, Schroeder SR. Robinson G, et at. Relation of lead and social factors to IQ of low-SES childn:a: a partial rspliatioo. Ans J Mem Defic 1986: 91:17843. 27. Winneke G, Hrdina K-G. Broekhaus A. Neueopsychobjiea! studies in chib dn:n with ekvited tooth-tead concenerations. I. Pibt study. Iru Arch Occup Environ Health 1982: 51:169-83. 28. Diarich KN. Knffi KM. Bamschein RL, et: al. Low.kvet fetal exposure effect on neurobehavioral development in eatty-infaney. Pediatncs 1987: 5:72 t-30. 29. McMichael Al. Baghurst PA. Wigj NR. Vimpmi GV. Robertson EF. Roberts R1. Poct Pirie Cohort Study: enviroamenul exposure to lead andN children's abilities at the agc of four years. N EnEl J Med 1988; 3I9:468-75.0 i~W~ 1~i/ ~1 V1 yti ~V '1iL.J ~