Child: care, health and development, 1992, 18,321-337
Review: neurotoxicity of lead* WILLIAM YULE Department of Psychology, University of London Institute of Psychiatry, De Crespigny Park, London SES 8AF, England
Accepted for publication 22 April 1992
Summary 'Lead and its compounds are potentially toxic; the element has no known physiological function; it is widely distributed in nature and as a result of man's activities' {Lawther Report: DHSS 1980). Concern grew in the late 1970s that even low levels of exposure to lead caused adverse changes in children's adjustment and academic attainment. This paper traces the rapid developments of the past decade in investigating the relationship between body lead burden and relevant outcome measures. Research strategies moved from smallscale, clinical-descriptive studies to large-scale epidemiological studies; from cross-sectional studies to longitudinal ones. Agreement had to be reached on how to measure body lead burden. Decisions had to be made on which outcome measures to use. Measures of global intelligence were complemented by measures of academic attainment, emotional adjustment and hyperactivity, as well as by experimental measures of more basic psychophysiological functioning. Ways had to be found of dealing experimentally and statistically with socioeconomic factors which acted as confounds. The investigations were marked by tremendous collaboration between investigators, between national agencies and across countries. The methodological issues are relevant to the investigation of other toxins and they define a field of developmental behavioural toxicology. This case study can act as a model for investigating the effects of other environmental toxins and hazards.
INTRODUCTION It had long been recognized that lead at high levels was very toxic and •Paper presented at the Third Meeting of the International Neurotoxicology Association. 1-5 July 1991, Salsomaggiore Terme. Italy.
321
322 W. Yule
often had serious effects on children's health and development. However, until the publication of Byers & Lord (1943) it was not fully realized that there could be persistent neurotoxic impairments following milder episodes of lead poisoning that fell short of overt encephalopathy. Untiljhe 1970s, there was a general assumption that increased lead levels were of little clinical importance if there were no recognizable symptoms of poisoning (in terms of anaemia, peripheral neuropathy, renal damage or neurological dysfunction), and if blood levels were below about 50 or 60 [Ag/dl. These assumptions were challenged both by scientists and by environmental pressure groups from the late 1970s onwards. The studies undertaken have been fully reviewed elsewhere (Rutter 1980, Rutter & Russell Jones 1983, Yule & Rutter 1985, Lansdown & Yule 1986, Smith et al. 1989). In this paper, the main findings up until the early 1980s will be summarized and then attention drawn to the methodological advances over the past decade. STUDIES PRIOR TO 1980 Prior to 1980, published studies had mainly been of small, highly selected groups of children with, by today's standards, very high body lead burdens. By 1980, it was concluded (Rutter 1980, Lawther Report: DHSS 1980) that blood lead levels persistently raised above 60 (Ag/dl were associated with an average reduction of some 3 to 4 10 points. It was thought likely, but less certainly, that there might also be adverse sequelae in the 40-60 [Ag/dl range. There were even suggestions that there may be possible psychological risks with lead levels below 40 (Ag/dl, but the evidence was scanty and inconclusive. These early studies established that there was a case to investigate. It seemed biologically plausible that low levels of lead exposure might have subtle effects on children's development and functioning, but the case was far from being established. The studies, and their critics, also drew attention to a number of important methodological problems that had to be addressed (see Table 1). The main problems are outlined below. Selection bias in sampling Small studies of highly selected subjects may be useful in helping formulate hypotheses, but these need to be tested out in large samples
Review: neurotoxicity of lead 323
TABLE 1. Methodological issues 1. Selection bias in sampling 2. Assessment of body lead burden blood versus tooth . quality control 3. Assessment of psychological functioning global IQ theory-related problem of multiple tests 4. Confounding variables 5. Sample size 6. Low-level lead exposure
drawn from total populations. Even with large numbers, it is important to trace a very high proportion as it is well established that those who do not co-operate with epidemiological investigations are often more pathological than those who do (Cox et al. 1977), In particular. Smith et al. (1983) established that those children who donate teeth for a tooth lead study are more likely to be middle-class girls, introducing systematic bias before the study is done. Assessment of body burden of lead There was considerable debate about the most appropriate index to take of body lead burden, A single estimate from one blood sample was seen as unsatisfactory as it would reflect only recent exposure and be unable to identify children who had a past history of an acute exposure to high lead levels. The quality of blood lead measurement left a lot to be desired, before standard measures with quality assurance built in were developed (Vahter 1982), When it became more feasible to study levels of lead in shed deciduous teeth, this was hailed as a breakthrough, partly on the grounds that bones acted as reservoirs that integrated lead exposure over time and so would provide better estimates of long-term exposure. However, these could only be used with children up to the age of shedding the teeth (7-9 years) and were not available for younger or older children. Assessment ofpsychological impairment Early studies used poorly standardized tests of neuropsychological functioning. It was clear that there was a need to include better
324 W. Yule
measures of cognitive performance and of behaviour. There were no well-worked-out theories of which brain functions were most affected by the action of lead on the CNS, so bigger and bigger batteries were being used. This brought the added statistical problem that a large number of univariate statistical comparisons between high and low exposure groups were bound to produce some 'significant' results by chance alone. Better statistical handling of the data involving multivariate statistics and much more replication was called for. Confounding variables It was becoming well established that there is a substantial overlap between social disadvantage and high level lead exposure (Rutter & Russell Jones 1983). For reasons unconnected with lead exposure, social disadvantage has an adverse influence on children's psychological development. Put crudely, some people argued that there was a relationship between lead and intelligence, say, that was solely due to inadequate mothers living in socially disadvantaged situations both failing to stimulate their children and allowing them to get contaminated in dirty houses or streets. Thus, if the effects of lead are to be properly assessed, it is essential to take account of the non-lead effects of the social disadvantage with which lead tends to be associated. Sample size As it came to be accepted that the effects of lead were likely to be small and because there was a need to control adequately for confounding variables, it was clear that large groups were needed to be able reliably to observe statistically significant results. Low level lead exposure With the focus shifting to effects of lead exposure in the general population and with the publication of data in Europe and North America suggesting that, with the exception of a few polluted areas, few children had blood lead levels above 35 ng/dl, research in the 1980s had to concentrate on children with levels below this.
STUDIES PUBLISHED 1980-1985 Studies published between 1980 and 1985 have been reviewed in Yule & Rutter (1985). The landmark study was that of Needleman et al.
Review: neurotoxicity of lead 325
(1979) from Boston. Of a total population of 3329 eligible children, 70% donated teeth for analysis: 58 children were recognized as having high tooth lead levels and 100 with low tooth lead levels. These were given 4 hours of detailed neuropsychological testing. For most children who gave teeth, teachers completed an U-item questionnaire on the children's behaviour. The average IQ of the low lead group was 106-6 while that of the high lead group was found to be 102-1, after correcting for social confounds. This difference was statistically significant. At this point it may be useful to comment on the significance of such a difference between the groups,. It is known that statistical significance and clinical significance are not always the same. It was also known that IQ tests, like all measurement, are subject to random error, and that they are sufficiently unreliable that any one person's IQ may fluctuate by more than 3 or 4 points over time. Wrongly, this sort of argument has led people to dismiss the significance of such findings as those of Needleman, The error lies in confusing the significance for an individual with the significance for the total population. If the group findings can truly be generalized to the total population, it is saying that the effects of exposure to lead is to shift the distribution of IQ scores down to the left by 3 to 4 points. This has little impact on the vast majority of people, but it has the effect of increasing the proportion of individuals who score below an IQ of 70 — that is, the children who often require special educational provision — by a factor of three or four. That would be a very significant social and educational impact, Needleman et al.'s (1979) study showed impressive dose response relationships between tooth lead and teacher's ratings in over 2000 children, although in this large group they could not control for social confounds, A surprisingly similar pattern was reported by Lansdown et al. (1983) and Yule et al. (1981) in their first London study. Again, social factors were not controlled, but the findings were extended both by using other better-validated scales that showed that lead was particularly associated with activity levels, and in finding that effects on intelligence were more pronounced in children coming from disadvantaged backgrounds. This period also saw the publication of Winneke etal.'s (1982, 1983) impressive tooth lead studies in Germany and Smith et a/,'s (1983) tooth lead study in London, All the studies in this period were crosssectional and were beginning to show increased sophistication, both in
326 W. Yule
the dependent and independent variables measured and in the statistical procedures used to control for social factors. By 1985, Yule & Rutter were able to conclude that the neuropsychological effects of lead within the range seen today in ordinary environments without special high risks are quite small . . , lead accounts for a tiny proportion of the variance in cognitive performance (probably no more than 2 or 3%) , . . even that minor effect is sufficient to be of importance from the viewpoint of both the individual and the policy maker.
They were less certain that this was a true causal effect but thought on balance that it was. The causal effect might largely apply within socially disadvantaged sections of the population, as is found with perinatal complications and malnutrition. It was more difficult to reach clear conclusions on the relationship between lead exposure and behaviour as findings were far less consistent across studies. Further, uncertainty remains on the possibility of a threshold below which lead effects are minimal or nonexistent . . , it is possible (although not demonstrated) that lead effects are negligible with blood levels below 20 )ig/dl.
Relevant to this important issue of a possible threshold, the (British) Royal Commission on Environmental Pollution (1983) observed: At present the average blood lead concentration of the UK population is about one quarter of that at which features of frank lead poisoning may occasionally occur. We are not aware of any other toxin which is so widely distributed in human and animal populations and which is also universally present at levels that exceed even one tenth of that at which clinical signs and symptoms may occur.
They concluded: TTie main tasks that remain for the future concern the biological mechanisms that underlie the neuropsychological effects; the determination of the neuropsychological processes affected by lead (and in particular whether these involve attention and behaviour); and, most of all, the elucidation of possible interaction effects by which lead exposure and social disadvantage may potentiate each other's ill-effects.
STUDIES PUBLISHED 1985-1991 The commission of the European Communities and the US Environmental Protection Agency jointly sponsored an international meeting in Edinburgh in September 1986 to undertake an up-to-date assessment of the area, paying particular attention to findings from longitudinal studies (Smith et al. 1989), In this paper, there is only
Review: neurotoxicity of lead 327
space to refer to some of the main findings from this impressive body of work (Table 2), In the Port Pirie smelter study of 831 pregnant women, McMichael et al. (1988) reported that women with elevated blood lead levels have a significantly increased risk of having a preterm delivery. In turn, the preterm infants will have an increased risk of developmental problems. Even in this smelter town, the average blood lead level in the women was only 9-9-10-4 [Ag/dl, When the children reached 4 years of age, they were tested on the McCarthy Scale of intelligence. After controlling for social confounds, a difference of 7-2 IQ points was found between those with very high lead levels (1-50 pimol/I) and those with average concentrations (0-50 M.mol/1) [1-0 fxmol/1 = 21 [xg/dl]. The investigators had used an integrated postnatal measure of blood lead concentration, and suggested that the effects may be cumulative. They noted a change in blood lead levels between the first and second 2 years of life and found the later estimates related less strongly than the earlier ones, Bellinger et al. (1987, 1990) also report a longitudinal cohort of children whose mothers were recruited during pregnancy. They also found that prenatal lead levels relate more strongly to children's later functioning than the children's later, postnatal lead levels. Two hundred and forty-nine infants were assigned to one of three prenatal exposure groups based on umbilical cord level of lead. The low exposure group (< 3 \ig/d\) scored nearly 5 points higher on the Bayley Mental Developmental Index at 2 years of age than did the high exposure group (> 9 jxg/dl). Again, infant outcome scores were not related to their postnatal blood lead levels.
TABLE 2. Prospective studies Principal investigator
Study
Bornschein Vimpani Bellinger Moore Graziano Needleman Silva Fergusson
Cincinnati birth cohort Port Pirie smelter "Boston (cord blood lead) Glasgow Yugoslavia (smelter/control) Boston 11-year follow-up Dunedin longitudinal Christchurch longitudinal
700 mothers 595 children 249 children 151 children 1000 children 132 children 579 children 1035 children
328 W. Yule
By 5 years of age, the children had compensated for the early exposure. They were reassessed on the McCarthy Scales and venous blood samples were analysed. Data on social factors such as quality of interactions were also gathered. Children from higher socio-economic backgrounds, with better scores on the HOME inventory (a measure of good parent-child interaction and stimulation), with brighter mothers, and with lower current blood lead levels, fared better. Gender was also an important risk factor in that girls fared better than boys. Thus, although the study shows some evidence of catch-up, the degree of progress does not apply equally to all children who experienced high exposure to lead in uterus. Even these two sets of studies are sufficient to illustrate that prospective studies clarify some issues and complicate others. It is certainly not the case that better, serial estimates of body lead burden facilitate the search for effects on dependent variables. Of course, here one runs into the reality of trying to measure anything in rapidly developing children. The Bayley test in infancy is different in many ways from the McCarthy test used at 3-5 years, and these differ again from the Wechsler Intelligence Scales appropriate for older children. The tests are age-related and while they all purport to measure underlying cognitive ability, it is difficult to disentangle differences across tests from real differences in study findings. In the Cincinnati prospective study (Dietrich et al. 1990), by 2 years of age there was no relationship overall with either prenatal or postnatal lead exposure and dependent variables. By 4 years, there were differences on the Kaufman battery, but only in children from lower social class. The Danish studies (Lyngbe et al. 1990 a,b, Norby Hansen et al. 1989) are also prospective but did not start with children at birth but in the first grade. T"he study is exemplary for the coverage of its population and the care with which environmental measures were assembled. A total of 1302 children donated teeth for lead analysis. The addresses at which children had lived during their first 3 years of life were ascertained and the length of stay was weighted by the traffic density known for each street. The results showed very clearly that the greater the exposure to traffic (and hence to lead from petrol), the greater the tooth lead levels. This relationship was linear. The investigators also reported that, contrary to previous fears, there was a good relationship between blood and tooth lead estimates of body lead burden, even though their lead levels in Aarhus were much lower than those reported in other European studies.
Review: neurotoxicity of lead 329
When children with relatively high tooth leads (above 18-7 [xg/g, accounting for 8% of the sample) were compared with controls with levels below 5 (xg/g on the Wechsler Scales, it was found that the high exposed group scored, on average, 8 points lower on verbal intelligence. They also performed worse on the Bender Gestalt test and their behaviour was rated less favourably by teachers. These findings remained significant after controlling for confounding factors. Two longitudinal studies In New Zealand decided to measure lead exposure in their large samples well into their studies, Silva et al. (1988) report on 579 11-year-old children. Even when many social covariates were controlled, raised blood lead levels were associated with small but significant increases in behaviour problems, especially problems of inattention and hyperactivity. No effect on intelligence was reported. The average blood lead level was 11-1 |ig/dl, and only two children had blood leads above 30 [ig/dl. Fergusson et al. (1988 a,b,c) regularly studied a cohort of over 1000 children from birth to 9 years. They gathered 996 usable teeth which were classified according to type and position. Their study of the tooth lead itself is of considerable interest. They reported that around 20% of the PbT variance was random error, with 2% being related to the age of shedding, PbT was found to relate significantly to a number of environmental risk factors such as social disadvantage, exposure to old wooden buildings (and hence to old lead paint), and to busy roads (at a time at which New Zealand petrol was heavily leaded). Overall, 10% of the variance of PbT was accounted for by social risk factors. The average PbT level was 6 [Ag/g which is considerably less than the Needleman et al. (1979) Boston value of 14 jxg/g. They note that, in this New' Zealand sample, children from lower social class and of mothers who were educationally less successful were less likely to donate teeth for the study. They tested the children's intelligence, academic attainment and behaviour at both age 8 and 9 years. They report a small but statistically significant relationship between PbT and WISC-IQ and Reading scores. After controlling for social confounds, the relationship between lead exposure and IQ fell below statistical significance, but the relationship with reading attainment remained significant. The correlations between PbT and parent and teacher ratings of behaviour, particularly of items related to attention and activity levels, remained statistically significant even after partialling out the effects of other social factors.
330 W. Yule
Fergusson et al. (1988b) make an important methodological point when they argue that, in their study, both at age 8 and at age 9, there was a correlation of about 0-10 between lead level and 10- This is a low, but stable relationship. Cohen's (1977) power tables suggest that, with a sample of 400 subjects, there is only a 50% chance of reliably detecting this low level of correlation. In samples of 600 to 800, the power increases to from 0-69 to 0-81. To detect this correlation with a chance of 90%, a sample size of over 1000 is needed. Most published studies, many of them otherwise well conducted, have been far too small to detect relationships of this level reliably and so it has been a case of some studies reporting a positive relationship between Pb and 10 and others not. However, small studies still have a major contribution, as is demonstrated by Needleman et al.'s (1990) follow up of 132 of the 270 children they had studied in 1975-1978. This 11-year follow up found that neurobehavioural functioning in young adulthood remained related to tooth lead levels in middle childhood. Those with the highest PbT were much more likely to drop out of high school or to have a reading disability. Needleman argues that this is evidence that lead has caused long-'term deficits in CNS functioning. Needleman & Gatsonis (1990) have undertaken a meta analysis of 24 modern studies of the relationship between low-level lead exposure and children's 10- In 11 of 12 studies where sufficient data were presented, there is a negative relationship between lead exposure and 10. Blood lead studies seemed more sensitive to the effect than tooth lead studies. THE WHO(EURO)/EEC STUDY Meta analyses are problematical when combining many studies which followed different protocols. Yet, to overcome the statistical power problem, it is desirable to aggregate subjects from different studies. One way of doing this is to have investigators adopt a core common protocol, while leaving them free to investigate additional problems of their own interest. This was the strategy adopted by the WHO(EURO)/EEC studies co-ordinated by Winneke at the University of Dusseldorf (see Table 3). In addition to agreeing a common study protocol that would investigate blood lead, neuropsychological performance and social factors, care was taken to incorporate quality
Review: neurotoxicity of lead 331
TABLE 3. The WHO(EURO)/EEC multicentre study Country
City
Principal investigator
Bulgaria Denmark Greece Hungary Italy Romania West Germany Yugoslavia Belgium United Kingdom
Sofia Aarhus Athens Budapest Modena Bucharest Dusseldorf Zagreb Brussels London
Z. Zaprianov 0 . Norby-Hansen A. Hatzakis P. Rudnai G. Vivoli M. Cucu G. Winneke" 0 . Weber R. Cluydts* W. Yule*
*Technical advisors. **Co-ordinating institute.
assurance measures wherever possible. New data were gathered by eight groups of investigators from different countries, with two groups providing additional technical advice. Overall, 1879 school-aged children aged 6 to 11 years were studied. The blood lead levels ranged from below 5 to greater than 60 [xg/dl. A mean is not meaningful as not all studies were population-based. Some deliberately studied highly exposed groups. In passing, it should be noted that measures of PbB were found to be highly stable over a 3-year period (r > 0-8) (Winneke et al. 1990). A study such as this poses many challenges, such as how to translate rating scales into different languages, what to do about measures of intelligence when these have not been standardized in translation in different countries, and how to measure socio-economic status in countries of different political organization. Addressing these questions helped sharpen the understanding of the basic processes the researchers wished to measure. Given Fergusson et al.'s (1988 a,b,c) views, the sample size was large enough to detect small effects. The model of statistical analysis required the measurement of a standard set of covariates — age, sex, father's occupation and mother's education — to be entered into multiple regression analyses. Even with the widely differing versions of the Wechsler Intelligence Scales that were used, there was a borderline significant relationship between PbB and 10. Stronger relationships were found with some of the more neuropsychological/
332 W. Yule
neurophysiological measures such as the Wiener Determinations Gerat that measure complex reaction times. Such measures rely less on language and translation. Indeed, the Bender Gestalt form copying task, scored very strictly, remained significantly related to PbB even after controlling for social factors. Winneke et al. (1990) conclude that the contribution of PbB to the observed variance never exceeded 0-8%, but was detectable. Higher lead levels were consistently associated with poorer performance. A risk assessment was undertaken on the data and it was concluded that 'by increasing blood-levels from a low 5 [Ag/dl to an elevated level of 20 [ig/dl, the proportion of children in the intermediate risk category (1 SD) increases by 2-4% (WISC), 3-9% (GFT-N) or 4-7% (GFT-I)'. Moreover, the authors could not detect a threshold below which no adverse health effects were found.
CONCLUSIONS This brief overview cannot do justice to the many findings from increasingly well-conducted investigations of the effects of lead exposure on children's health and development. The studies highlighted above were chosen to illustrate both findings and methodological issues, the latter being applicable to the study of many other toxins and hazards. When the author was first introduced to this field on the British Department of Health Working Party, the chairman. Professor Lawther, wisely spent an early session in which different disciplines explained the nature of the measures they were using and the relevance of these for the study of the effects of lead. In interdisciplinary research, this is vital. Other disciplines' findings always seem so much more certain than one's own. It is only by such discussions that one can appreciate that there is a need to demonstrate the quality of all measures and to provide estimates of their reliability and validity. The author now has a far better understanding of the problems both in taking blood and in estimating its lead content and hopes his colleagues have a better understanding of the problems inherent in measuring children's cognitive ability and behaviour. The studies of the past decade have amply demonstrated increasing sophistication in measurement of both the independent and dependent variables.
Review: neurotoxicity of lead 333
Finally, it is necessary to determine how far the studies have addressed the methodological problems which were highlighted at the beginning of the paper. Selection bias There are far fewer studies of small, highly selected samples, but within large cross-sectional and prospective studies, the problem of biased attrition remains. It has been demonstrated that children who donate teeth differ from those who do not, and the ones who do not come from the less socially advantaged groups. Assessment of body burden of lead Better quality control has lead to better cross-study comparability. Blood lead has been found to be much more stable than once thought, and may even be more sensitive to outcome measures than tooth lead. Other indices of body lead burden are scarcely used in serious studies. Psychological functioning There is better agreement on which measures to use, but still the choice of instruments is guided more by previous results than by any theory. The move toward developing more fundamental measures of psychological functioning such as attention and learning shows promise, but still has a long way to go. Confounding variables It proved possible to measure confounding variables adequately across different countries. There is better agreement on how to handle such confounds statistically, and increasing sophistication in statistical modelling of results when the data sets are large enough. Sample size Bigger is not always better, but samples of over 1000 are needed in order to be able to detect the level of effects that are now thought to occur in the low exposure ranges that are typical of the 1990s. Over the past 20 years, there has been a very interesting interaction between scientists and policy makers and pressure groups. On the whole, scientists are not trained to deal with the pressures they are put under when a policy issue on which they are researching hits the public agenda. At such times it is vital to retain high scientific standards. Simplistic answers are rarely found in the complexities of real life.
334 W. Yule
Scientists had to recognize that the concerns expressed by environmental pressure groups were genuinely held and that there was a scientific questi'on to investigate. In the case of the effects of lead, the policy decision to remove lead from petrol and to reduce its distribution from other sources was taken fairly quickly — almost too quickly for the full range of scientific questions to be answered. From the evidence that the author has reviewed, he believes that lead exposure does directly cause adverse heahh effects. At the levels Of lead exposure most children experience today, these effects will not be serious for most children but, to society as a whole, there will be an unnecesary increase of children with special educational needs. Since lead has no known beneficial effects on the human body, the case to remove it from the environment is persuasive. But exactly how lead exerts its effect on the CNS is still not known. With lead disappearing from the political agenda, at least for the time being, it will be more difficult to obtain funding to finish the studies that were showing most promise. It seems accepted now that lead has many effects within the body and CNS. At high levels it may cause physical, structural damage; at low levels, it interferes with neuro transmitters. A good argument can be made that, by studying the effects of lead on neurotransmitters, we may get a better understanding of the causes of attention disorders and hyperactivity. This review has also touched on a number of remaining paradoxes. Early on, it was argued that lead exposure would have its strongest effect on the brain when it was developing most rapidly, that is at and around birth. However, the prospective studies are finding that relationships between lead exposure and cognitive functioning disappear or become inconsistent by 4 to 5 years. Against this, Needleman et al.'s (1990) 11-year follow-up study finds impressive relationships between earlier exposure and socially meaningful pathology in early adulthood. Just over 10 years ago, it was advised that it was unethical to take blood samples from children who were not at high risk of having some problem. This edict in part made the use of estimates from teeth more attractive. In turn, it was argued that teeth were better indicators of life-long exposure. The data now show that blood lead levels are much more stable than was once thought and that they may be more strongly correlated with contemporaneous measures of functioning. Scientists are still left with the dilemma that different indices are only partly correlated.
Review: neurotoxicity of lead 335
Some of the very early studies of clinically poisoned children used positive effects of chelation as an argument that lead (at least at high doses) had effects on children's behaviour. Recent evidence from animal studies (Cory-Schlecta et al. 1987) suggests that Ca disodium EDTA mobilizes lead but this then gets redeposited in vital organs such as the brain and liver. Such findings make one wonder how the early chelation studies showed such dramatic effects and whether, on longer term follow-up, the effects would persist. At the very least, the findings remind one that there is still a long way to go in understanding the effects of lowlevel lead exposure on children's development, but a great deal has been learned in the past 10 years. REFERENCES Bellinger D., Leviton A., Waternaux C , Needleman H. & Rabinowitz M. (1987) Longitudinal analyses of prenatal and postnatal lead exposure and early cognitive development. New England Journal of Medicine 316. 1037-1043 Bellinger D., Leviton A. & Sloman J. (1990) Antecedents and correlates of improved cognitive performance in children exposed in utero to low levels of lead. Environmental Health Perspectives 89, 5-11 Byers R.K. & Lord E.E. (1943) Late effects of lead poisoning on mental development. American Journal of Diseases of Childhood 66. 194-471 Cohen J. (1977) Statistical Power Analysis for ihe Social Sciences. Academic Press, New York Cory-Schlecta D.A., Weiss B. & Cox C. (1987) Mobilization and redistribution of lead over the course of calcium disodium ethylenediaihine tetraacetate chelation therapy. Journal of Pharmacology and Experimental Therapeutics 243, 804-813 Cox A., Rutter M., Yule B. & Quinton D. (1977) Bias resulting from missing information: some epidemiological flndings. British Journal of Preventive and Social Medicine 31, 131-136 Department of Health and Social Security (DHSS) (1980) Lead and Health. The Report of a Working Party on Lead in the Environment (Lawther Report). HMSO. London Dietrich K.N., Succop P.A., Bornschein R.L., Krafft K.M., Berger C , Hammond P.B. et al. (1990) Lead exposure and neurobehavioral development in late infancy. Environmental Health Perspectives 89, 13-19 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988a) A longitudinal study of , dentine lead levels, intelligence, school performance and behaviour: Part I. Dentine lead levels and exposure to environmental risk factors. Journal of Child Psychology and Psychiatry 29, 781-792 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988b) A longitudinal study of dentine lead levels, intelligence, school performance and behaviour: Part II. Dentine lead and cognitive ability. Journal of Child Psychology and Psychiatry 29, 793-810 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988c) A longitudinal study of dentine lead levels, intelligence, school performance and behaviour: Part III. Dentine lead levels and attention/activity. Journal of Child Psychology and Psychiatry 29, 811-824 Lansdown R. & Yule W. (eds) (1986) The Lead Debate: The Environment, Toxicology and Child Health. Croom Helm. BeckenlTam, Kent
336 W. Yule
Lansdown R., Yule W., Urbanowicz M.A. & Millar I.B. (1983) Blood lead, intelligence, attainment and behaviour in school children: overview of a pilot study; In Lead Versus Health: Sources and Effects of Low Level Lead Exposure, eds M. Rutter & R. Russel Jones. J. Wiley, Chichester Lyngbe T., Hansen O.N. & Grandjean P. (1990a) Predictors of tooth-lead level with special reference to traffic: a study of lead-exposure in children. International Archives of Occupational and Environmental Health 62, 417-422 Lyngbe T., Jorgensen P.J., Grandjean P. & Hansen O.N. (1990b) Validity and interpretation of blood lead levels: a study of Danish school children. Scandinavian Journal of Clinical Laboratory Investigations SO, 441-449 McMichael A.J., Baghurst P.A., Wigg, N., Vimpani G.V., Robertson E.F. & Roberts R.J. (1988) Port Pirie cohort study: environmental exposure to lead and children's abilities at the age of four years. New England Journal of Medicine 319, 468-475 Needieman H.L. & Gatsonis C.A. (1990) Low-level lead exposure and the IQ of children: a meta-analysis of modern studies. Journal of the American Medical Association 263, 673-678 Needieman H.L., Gunnoe C , Leviton A.. Reed M., Peresie H., Maher C. et al. (1979) Deficits in psychological and classroom performance of children with elevated dentine lead levels. New England Journal of Medicine 300, 689-695 Needieman H.L., Schell, A., Bellinger D., Leviton A. & Allred E.N. (1990) The long-term effects of exposure to low doses of lead in childhood: an 11-year follow-up report. New England Journal of Medicine 322, 83-88 Norby Hansen O., Trillingsgaard A., Beese I., Lyngbe T. & Grandjean P. (1989) A neuropsychoiogical study of children with elevated dentine lead level: assessment of the effect of lead In different socio-economic groups. Neurotoxicology and Teratology 11, 205-213 Royal Commission on Environmental Pollution (1983) Ninth Report: Lead in the Environment. HMSO, London Rutter M. (1980) Raised lead levels and impaired cognitive/behavioural functioning: a review of the evidence. Developmental Medicine and Child Neurology 22 (suppl.), 42 Rutter, M. & Russell Jones R. (eds) (1983) Lead Versus Health: Sources and Effects of Low Level Lead Exposures, i. Wiley. Chichester Silva P.A., Hughes P., Williams S. & Faed J.M. (1988) Blood lead, intelligence, reading attainment, and behaviour in eleven year old children in Dunedin, New Zealand. Journal of Child Psychology and Psychiatry 29, 43-52 Smith M., Delves T., Lansdown R., Clayton B. & Graham P. (1983) The effect of lead exposure on urban children. Institute of Child Health/University of Southampton study. Developmental Medicine and Child Neurology 25 (suppl.), 47 Smith M.A., Grant L.D. & Sors A.I. (eds) (1989) Lead Exposure and Child Development: An International Assessment. Kluwer, London Vahter M. (ed.) (1982) Assessment of Human Exposure to Lead and Cadmium Through Biological Monitoring. National Swedish Institute of Environmental Medicine and Department of Environmental Hygiene, Karolinska Institute, Stockholm Winneke G., Hrdina K.-G. & Brockhaus A. (1982) Neuropsychoiogical studies in children with elevated tooth-lead concentrations I. Pilot study. International Archives of Occcupational and Environmental Health 51. 169-183 Winneke G., Kramer U., Brockhaus A., Ewers U., Kijanek G.. Lechner H. et al. (1983) Neuropsychoiogical studies in children with elevated tooth-lead concentrations II. Extended study. International Archives of Occupational and Environmental Health 51,231-252 Winneke G., Brockhaus A., Ewers U., Kramer U. & Neuf M. (1990) Results from the European multicenter study on lead neurotoxicity in children: implications for risk assessment. Neurotoxicology and Teratology 12, 553-559 Yule W. & Rutter M. (1985) Effects of lead on children's behaviour and cognitive performance:
Review: neurotoxicity of lead 337
a critical review. In Dietary and Environmental Lead: Human Health Effects, ed. K.R. Mahaffey. Elsevier, Amsterdam Yule W., Lansdown R., Millar I.B. & Urbanowicz M.A. (1981) The relationship between blood lead concentrations, intelligence and attainment in a school population: a pilot study. Developmental Medicine and Child Neurology 25. 567-576
Review: neurotoxicity of lead* WILLIAM YULE Department of Psychology, University of London Institute of Psychiatry, De Crespigny Park, London SES 8AF, England
Accepted for publication 22 April 1992
Summary 'Lead and its compounds are potentially toxic; the element has no known physiological function; it is widely distributed in nature and as a result of man's activities' {Lawther Report: DHSS 1980). Concern grew in the late 1970s that even low levels of exposure to lead caused adverse changes in children's adjustment and academic attainment. This paper traces the rapid developments of the past decade in investigating the relationship between body lead burden and relevant outcome measures. Research strategies moved from smallscale, clinical-descriptive studies to large-scale epidemiological studies; from cross-sectional studies to longitudinal ones. Agreement had to be reached on how to measure body lead burden. Decisions had to be made on which outcome measures to use. Measures of global intelligence were complemented by measures of academic attainment, emotional adjustment and hyperactivity, as well as by experimental measures of more basic psychophysiological functioning. Ways had to be found of dealing experimentally and statistically with socioeconomic factors which acted as confounds. The investigations were marked by tremendous collaboration between investigators, between national agencies and across countries. The methodological issues are relevant to the investigation of other toxins and they define a field of developmental behavioural toxicology. This case study can act as a model for investigating the effects of other environmental toxins and hazards.
INTRODUCTION It had long been recognized that lead at high levels was very toxic and •Paper presented at the Third Meeting of the International Neurotoxicology Association. 1-5 July 1991, Salsomaggiore Terme. Italy.
321
322 W. Yule
often had serious effects on children's health and development. However, until the publication of Byers & Lord (1943) it was not fully realized that there could be persistent neurotoxic impairments following milder episodes of lead poisoning that fell short of overt encephalopathy. Untiljhe 1970s, there was a general assumption that increased lead levels were of little clinical importance if there were no recognizable symptoms of poisoning (in terms of anaemia, peripheral neuropathy, renal damage or neurological dysfunction), and if blood levels were below about 50 or 60 [Ag/dl. These assumptions were challenged both by scientists and by environmental pressure groups from the late 1970s onwards. The studies undertaken have been fully reviewed elsewhere (Rutter 1980, Rutter & Russell Jones 1983, Yule & Rutter 1985, Lansdown & Yule 1986, Smith et al. 1989). In this paper, the main findings up until the early 1980s will be summarized and then attention drawn to the methodological advances over the past decade. STUDIES PRIOR TO 1980 Prior to 1980, published studies had mainly been of small, highly selected groups of children with, by today's standards, very high body lead burdens. By 1980, it was concluded (Rutter 1980, Lawther Report: DHSS 1980) that blood lead levels persistently raised above 60 (Ag/dl were associated with an average reduction of some 3 to 4 10 points. It was thought likely, but less certainly, that there might also be adverse sequelae in the 40-60 [Ag/dl range. There were even suggestions that there may be possible psychological risks with lead levels below 40 (Ag/dl, but the evidence was scanty and inconclusive. These early studies established that there was a case to investigate. It seemed biologically plausible that low levels of lead exposure might have subtle effects on children's development and functioning, but the case was far from being established. The studies, and their critics, also drew attention to a number of important methodological problems that had to be addressed (see Table 1). The main problems are outlined below. Selection bias in sampling Small studies of highly selected subjects may be useful in helping formulate hypotheses, but these need to be tested out in large samples
Review: neurotoxicity of lead 323
TABLE 1. Methodological issues 1. Selection bias in sampling 2. Assessment of body lead burden blood versus tooth . quality control 3. Assessment of psychological functioning global IQ theory-related problem of multiple tests 4. Confounding variables 5. Sample size 6. Low-level lead exposure
drawn from total populations. Even with large numbers, it is important to trace a very high proportion as it is well established that those who do not co-operate with epidemiological investigations are often more pathological than those who do (Cox et al. 1977), In particular. Smith et al. (1983) established that those children who donate teeth for a tooth lead study are more likely to be middle-class girls, introducing systematic bias before the study is done. Assessment of body burden of lead There was considerable debate about the most appropriate index to take of body lead burden, A single estimate from one blood sample was seen as unsatisfactory as it would reflect only recent exposure and be unable to identify children who had a past history of an acute exposure to high lead levels. The quality of blood lead measurement left a lot to be desired, before standard measures with quality assurance built in were developed (Vahter 1982), When it became more feasible to study levels of lead in shed deciduous teeth, this was hailed as a breakthrough, partly on the grounds that bones acted as reservoirs that integrated lead exposure over time and so would provide better estimates of long-term exposure. However, these could only be used with children up to the age of shedding the teeth (7-9 years) and were not available for younger or older children. Assessment ofpsychological impairment Early studies used poorly standardized tests of neuropsychological functioning. It was clear that there was a need to include better
324 W. Yule
measures of cognitive performance and of behaviour. There were no well-worked-out theories of which brain functions were most affected by the action of lead on the CNS, so bigger and bigger batteries were being used. This brought the added statistical problem that a large number of univariate statistical comparisons between high and low exposure groups were bound to produce some 'significant' results by chance alone. Better statistical handling of the data involving multivariate statistics and much more replication was called for. Confounding variables It was becoming well established that there is a substantial overlap between social disadvantage and high level lead exposure (Rutter & Russell Jones 1983). For reasons unconnected with lead exposure, social disadvantage has an adverse influence on children's psychological development. Put crudely, some people argued that there was a relationship between lead and intelligence, say, that was solely due to inadequate mothers living in socially disadvantaged situations both failing to stimulate their children and allowing them to get contaminated in dirty houses or streets. Thus, if the effects of lead are to be properly assessed, it is essential to take account of the non-lead effects of the social disadvantage with which lead tends to be associated. Sample size As it came to be accepted that the effects of lead were likely to be small and because there was a need to control adequately for confounding variables, it was clear that large groups were needed to be able reliably to observe statistically significant results. Low level lead exposure With the focus shifting to effects of lead exposure in the general population and with the publication of data in Europe and North America suggesting that, with the exception of a few polluted areas, few children had blood lead levels above 35 ng/dl, research in the 1980s had to concentrate on children with levels below this.
STUDIES PUBLISHED 1980-1985 Studies published between 1980 and 1985 have been reviewed in Yule & Rutter (1985). The landmark study was that of Needleman et al.
Review: neurotoxicity of lead 325
(1979) from Boston. Of a total population of 3329 eligible children, 70% donated teeth for analysis: 58 children were recognized as having high tooth lead levels and 100 with low tooth lead levels. These were given 4 hours of detailed neuropsychological testing. For most children who gave teeth, teachers completed an U-item questionnaire on the children's behaviour. The average IQ of the low lead group was 106-6 while that of the high lead group was found to be 102-1, after correcting for social confounds. This difference was statistically significant. At this point it may be useful to comment on the significance of such a difference between the groups,. It is known that statistical significance and clinical significance are not always the same. It was also known that IQ tests, like all measurement, are subject to random error, and that they are sufficiently unreliable that any one person's IQ may fluctuate by more than 3 or 4 points over time. Wrongly, this sort of argument has led people to dismiss the significance of such findings as those of Needleman, The error lies in confusing the significance for an individual with the significance for the total population. If the group findings can truly be generalized to the total population, it is saying that the effects of exposure to lead is to shift the distribution of IQ scores down to the left by 3 to 4 points. This has little impact on the vast majority of people, but it has the effect of increasing the proportion of individuals who score below an IQ of 70 — that is, the children who often require special educational provision — by a factor of three or four. That would be a very significant social and educational impact, Needleman et al.'s (1979) study showed impressive dose response relationships between tooth lead and teacher's ratings in over 2000 children, although in this large group they could not control for social confounds, A surprisingly similar pattern was reported by Lansdown et al. (1983) and Yule et al. (1981) in their first London study. Again, social factors were not controlled, but the findings were extended both by using other better-validated scales that showed that lead was particularly associated with activity levels, and in finding that effects on intelligence were more pronounced in children coming from disadvantaged backgrounds. This period also saw the publication of Winneke etal.'s (1982, 1983) impressive tooth lead studies in Germany and Smith et a/,'s (1983) tooth lead study in London, All the studies in this period were crosssectional and were beginning to show increased sophistication, both in
326 W. Yule
the dependent and independent variables measured and in the statistical procedures used to control for social factors. By 1985, Yule & Rutter were able to conclude that the neuropsychological effects of lead within the range seen today in ordinary environments without special high risks are quite small . . , lead accounts for a tiny proportion of the variance in cognitive performance (probably no more than 2 or 3%) , . . even that minor effect is sufficient to be of importance from the viewpoint of both the individual and the policy maker.
They were less certain that this was a true causal effect but thought on balance that it was. The causal effect might largely apply within socially disadvantaged sections of the population, as is found with perinatal complications and malnutrition. It was more difficult to reach clear conclusions on the relationship between lead exposure and behaviour as findings were far less consistent across studies. Further, uncertainty remains on the possibility of a threshold below which lead effects are minimal or nonexistent . . , it is possible (although not demonstrated) that lead effects are negligible with blood levels below 20 )ig/dl.
Relevant to this important issue of a possible threshold, the (British) Royal Commission on Environmental Pollution (1983) observed: At present the average blood lead concentration of the UK population is about one quarter of that at which features of frank lead poisoning may occasionally occur. We are not aware of any other toxin which is so widely distributed in human and animal populations and which is also universally present at levels that exceed even one tenth of that at which clinical signs and symptoms may occur.
They concluded: TTie main tasks that remain for the future concern the biological mechanisms that underlie the neuropsychological effects; the determination of the neuropsychological processes affected by lead (and in particular whether these involve attention and behaviour); and, most of all, the elucidation of possible interaction effects by which lead exposure and social disadvantage may potentiate each other's ill-effects.
STUDIES PUBLISHED 1985-1991 The commission of the European Communities and the US Environmental Protection Agency jointly sponsored an international meeting in Edinburgh in September 1986 to undertake an up-to-date assessment of the area, paying particular attention to findings from longitudinal studies (Smith et al. 1989), In this paper, there is only
Review: neurotoxicity of lead 327
space to refer to some of the main findings from this impressive body of work (Table 2), In the Port Pirie smelter study of 831 pregnant women, McMichael et al. (1988) reported that women with elevated blood lead levels have a significantly increased risk of having a preterm delivery. In turn, the preterm infants will have an increased risk of developmental problems. Even in this smelter town, the average blood lead level in the women was only 9-9-10-4 [Ag/dl, When the children reached 4 years of age, they were tested on the McCarthy Scale of intelligence. After controlling for social confounds, a difference of 7-2 IQ points was found between those with very high lead levels (1-50 pimol/I) and those with average concentrations (0-50 M.mol/1) [1-0 fxmol/1 = 21 [xg/dl]. The investigators had used an integrated postnatal measure of blood lead concentration, and suggested that the effects may be cumulative. They noted a change in blood lead levels between the first and second 2 years of life and found the later estimates related less strongly than the earlier ones, Bellinger et al. (1987, 1990) also report a longitudinal cohort of children whose mothers were recruited during pregnancy. They also found that prenatal lead levels relate more strongly to children's later functioning than the children's later, postnatal lead levels. Two hundred and forty-nine infants were assigned to one of three prenatal exposure groups based on umbilical cord level of lead. The low exposure group (< 3 \ig/d\) scored nearly 5 points higher on the Bayley Mental Developmental Index at 2 years of age than did the high exposure group (> 9 jxg/dl). Again, infant outcome scores were not related to their postnatal blood lead levels.
TABLE 2. Prospective studies Principal investigator
Study
Bornschein Vimpani Bellinger Moore Graziano Needleman Silva Fergusson
Cincinnati birth cohort Port Pirie smelter "Boston (cord blood lead) Glasgow Yugoslavia (smelter/control) Boston 11-year follow-up Dunedin longitudinal Christchurch longitudinal
700 mothers 595 children 249 children 151 children 1000 children 132 children 579 children 1035 children
328 W. Yule
By 5 years of age, the children had compensated for the early exposure. They were reassessed on the McCarthy Scales and venous blood samples were analysed. Data on social factors such as quality of interactions were also gathered. Children from higher socio-economic backgrounds, with better scores on the HOME inventory (a measure of good parent-child interaction and stimulation), with brighter mothers, and with lower current blood lead levels, fared better. Gender was also an important risk factor in that girls fared better than boys. Thus, although the study shows some evidence of catch-up, the degree of progress does not apply equally to all children who experienced high exposure to lead in uterus. Even these two sets of studies are sufficient to illustrate that prospective studies clarify some issues and complicate others. It is certainly not the case that better, serial estimates of body lead burden facilitate the search for effects on dependent variables. Of course, here one runs into the reality of trying to measure anything in rapidly developing children. The Bayley test in infancy is different in many ways from the McCarthy test used at 3-5 years, and these differ again from the Wechsler Intelligence Scales appropriate for older children. The tests are age-related and while they all purport to measure underlying cognitive ability, it is difficult to disentangle differences across tests from real differences in study findings. In the Cincinnati prospective study (Dietrich et al. 1990), by 2 years of age there was no relationship overall with either prenatal or postnatal lead exposure and dependent variables. By 4 years, there were differences on the Kaufman battery, but only in children from lower social class. The Danish studies (Lyngbe et al. 1990 a,b, Norby Hansen et al. 1989) are also prospective but did not start with children at birth but in the first grade. T"he study is exemplary for the coverage of its population and the care with which environmental measures were assembled. A total of 1302 children donated teeth for lead analysis. The addresses at which children had lived during their first 3 years of life were ascertained and the length of stay was weighted by the traffic density known for each street. The results showed very clearly that the greater the exposure to traffic (and hence to lead from petrol), the greater the tooth lead levels. This relationship was linear. The investigators also reported that, contrary to previous fears, there was a good relationship between blood and tooth lead estimates of body lead burden, even though their lead levels in Aarhus were much lower than those reported in other European studies.
Review: neurotoxicity of lead 329
When children with relatively high tooth leads (above 18-7 [xg/g, accounting for 8% of the sample) were compared with controls with levels below 5 (xg/g on the Wechsler Scales, it was found that the high exposed group scored, on average, 8 points lower on verbal intelligence. They also performed worse on the Bender Gestalt test and their behaviour was rated less favourably by teachers. These findings remained significant after controlling for confounding factors. Two longitudinal studies In New Zealand decided to measure lead exposure in their large samples well into their studies, Silva et al. (1988) report on 579 11-year-old children. Even when many social covariates were controlled, raised blood lead levels were associated with small but significant increases in behaviour problems, especially problems of inattention and hyperactivity. No effect on intelligence was reported. The average blood lead level was 11-1 |ig/dl, and only two children had blood leads above 30 [ig/dl. Fergusson et al. (1988 a,b,c) regularly studied a cohort of over 1000 children from birth to 9 years. They gathered 996 usable teeth which were classified according to type and position. Their study of the tooth lead itself is of considerable interest. They reported that around 20% of the PbT variance was random error, with 2% being related to the age of shedding, PbT was found to relate significantly to a number of environmental risk factors such as social disadvantage, exposure to old wooden buildings (and hence to old lead paint), and to busy roads (at a time at which New Zealand petrol was heavily leaded). Overall, 10% of the variance of PbT was accounted for by social risk factors. The average PbT level was 6 [Ag/g which is considerably less than the Needleman et al. (1979) Boston value of 14 jxg/g. They note that, in this New' Zealand sample, children from lower social class and of mothers who were educationally less successful were less likely to donate teeth for the study. They tested the children's intelligence, academic attainment and behaviour at both age 8 and 9 years. They report a small but statistically significant relationship between PbT and WISC-IQ and Reading scores. After controlling for social confounds, the relationship between lead exposure and IQ fell below statistical significance, but the relationship with reading attainment remained significant. The correlations between PbT and parent and teacher ratings of behaviour, particularly of items related to attention and activity levels, remained statistically significant even after partialling out the effects of other social factors.
330 W. Yule
Fergusson et al. (1988b) make an important methodological point when they argue that, in their study, both at age 8 and at age 9, there was a correlation of about 0-10 between lead level and 10- This is a low, but stable relationship. Cohen's (1977) power tables suggest that, with a sample of 400 subjects, there is only a 50% chance of reliably detecting this low level of correlation. In samples of 600 to 800, the power increases to from 0-69 to 0-81. To detect this correlation with a chance of 90%, a sample size of over 1000 is needed. Most published studies, many of them otherwise well conducted, have been far too small to detect relationships of this level reliably and so it has been a case of some studies reporting a positive relationship between Pb and 10 and others not. However, small studies still have a major contribution, as is demonstrated by Needleman et al.'s (1990) follow up of 132 of the 270 children they had studied in 1975-1978. This 11-year follow up found that neurobehavioural functioning in young adulthood remained related to tooth lead levels in middle childhood. Those with the highest PbT were much more likely to drop out of high school or to have a reading disability. Needleman argues that this is evidence that lead has caused long-'term deficits in CNS functioning. Needleman & Gatsonis (1990) have undertaken a meta analysis of 24 modern studies of the relationship between low-level lead exposure and children's 10- In 11 of 12 studies where sufficient data were presented, there is a negative relationship between lead exposure and 10. Blood lead studies seemed more sensitive to the effect than tooth lead studies. THE WHO(EURO)/EEC STUDY Meta analyses are problematical when combining many studies which followed different protocols. Yet, to overcome the statistical power problem, it is desirable to aggregate subjects from different studies. One way of doing this is to have investigators adopt a core common protocol, while leaving them free to investigate additional problems of their own interest. This was the strategy adopted by the WHO(EURO)/EEC studies co-ordinated by Winneke at the University of Dusseldorf (see Table 3). In addition to agreeing a common study protocol that would investigate blood lead, neuropsychological performance and social factors, care was taken to incorporate quality
Review: neurotoxicity of lead 331
TABLE 3. The WHO(EURO)/EEC multicentre study Country
City
Principal investigator
Bulgaria Denmark Greece Hungary Italy Romania West Germany Yugoslavia Belgium United Kingdom
Sofia Aarhus Athens Budapest Modena Bucharest Dusseldorf Zagreb Brussels London
Z. Zaprianov 0 . Norby-Hansen A. Hatzakis P. Rudnai G. Vivoli M. Cucu G. Winneke" 0 . Weber R. Cluydts* W. Yule*
*Technical advisors. **Co-ordinating institute.
assurance measures wherever possible. New data were gathered by eight groups of investigators from different countries, with two groups providing additional technical advice. Overall, 1879 school-aged children aged 6 to 11 years were studied. The blood lead levels ranged from below 5 to greater than 60 [xg/dl. A mean is not meaningful as not all studies were population-based. Some deliberately studied highly exposed groups. In passing, it should be noted that measures of PbB were found to be highly stable over a 3-year period (r > 0-8) (Winneke et al. 1990). A study such as this poses many challenges, such as how to translate rating scales into different languages, what to do about measures of intelligence when these have not been standardized in translation in different countries, and how to measure socio-economic status in countries of different political organization. Addressing these questions helped sharpen the understanding of the basic processes the researchers wished to measure. Given Fergusson et al.'s (1988 a,b,c) views, the sample size was large enough to detect small effects. The model of statistical analysis required the measurement of a standard set of covariates — age, sex, father's occupation and mother's education — to be entered into multiple regression analyses. Even with the widely differing versions of the Wechsler Intelligence Scales that were used, there was a borderline significant relationship between PbB and 10. Stronger relationships were found with some of the more neuropsychological/
332 W. Yule
neurophysiological measures such as the Wiener Determinations Gerat that measure complex reaction times. Such measures rely less on language and translation. Indeed, the Bender Gestalt form copying task, scored very strictly, remained significantly related to PbB even after controlling for social factors. Winneke et al. (1990) conclude that the contribution of PbB to the observed variance never exceeded 0-8%, but was detectable. Higher lead levels were consistently associated with poorer performance. A risk assessment was undertaken on the data and it was concluded that 'by increasing blood-levels from a low 5 [Ag/dl to an elevated level of 20 [ig/dl, the proportion of children in the intermediate risk category (1 SD) increases by 2-4% (WISC), 3-9% (GFT-N) or 4-7% (GFT-I)'. Moreover, the authors could not detect a threshold below which no adverse health effects were found.
CONCLUSIONS This brief overview cannot do justice to the many findings from increasingly well-conducted investigations of the effects of lead exposure on children's health and development. The studies highlighted above were chosen to illustrate both findings and methodological issues, the latter being applicable to the study of many other toxins and hazards. When the author was first introduced to this field on the British Department of Health Working Party, the chairman. Professor Lawther, wisely spent an early session in which different disciplines explained the nature of the measures they were using and the relevance of these for the study of the effects of lead. In interdisciplinary research, this is vital. Other disciplines' findings always seem so much more certain than one's own. It is only by such discussions that one can appreciate that there is a need to demonstrate the quality of all measures and to provide estimates of their reliability and validity. The author now has a far better understanding of the problems both in taking blood and in estimating its lead content and hopes his colleagues have a better understanding of the problems inherent in measuring children's cognitive ability and behaviour. The studies of the past decade have amply demonstrated increasing sophistication in measurement of both the independent and dependent variables.
Review: neurotoxicity of lead 333
Finally, it is necessary to determine how far the studies have addressed the methodological problems which were highlighted at the beginning of the paper. Selection bias There are far fewer studies of small, highly selected samples, but within large cross-sectional and prospective studies, the problem of biased attrition remains. It has been demonstrated that children who donate teeth differ from those who do not, and the ones who do not come from the less socially advantaged groups. Assessment of body burden of lead Better quality control has lead to better cross-study comparability. Blood lead has been found to be much more stable than once thought, and may even be more sensitive to outcome measures than tooth lead. Other indices of body lead burden are scarcely used in serious studies. Psychological functioning There is better agreement on which measures to use, but still the choice of instruments is guided more by previous results than by any theory. The move toward developing more fundamental measures of psychological functioning such as attention and learning shows promise, but still has a long way to go. Confounding variables It proved possible to measure confounding variables adequately across different countries. There is better agreement on how to handle such confounds statistically, and increasing sophistication in statistical modelling of results when the data sets are large enough. Sample size Bigger is not always better, but samples of over 1000 are needed in order to be able to detect the level of effects that are now thought to occur in the low exposure ranges that are typical of the 1990s. Over the past 20 years, there has been a very interesting interaction between scientists and policy makers and pressure groups. On the whole, scientists are not trained to deal with the pressures they are put under when a policy issue on which they are researching hits the public agenda. At such times it is vital to retain high scientific standards. Simplistic answers are rarely found in the complexities of real life.
334 W. Yule
Scientists had to recognize that the concerns expressed by environmental pressure groups were genuinely held and that there was a scientific questi'on to investigate. In the case of the effects of lead, the policy decision to remove lead from petrol and to reduce its distribution from other sources was taken fairly quickly — almost too quickly for the full range of scientific questions to be answered. From the evidence that the author has reviewed, he believes that lead exposure does directly cause adverse heahh effects. At the levels Of lead exposure most children experience today, these effects will not be serious for most children but, to society as a whole, there will be an unnecesary increase of children with special educational needs. Since lead has no known beneficial effects on the human body, the case to remove it from the environment is persuasive. But exactly how lead exerts its effect on the CNS is still not known. With lead disappearing from the political agenda, at least for the time being, it will be more difficult to obtain funding to finish the studies that were showing most promise. It seems accepted now that lead has many effects within the body and CNS. At high levels it may cause physical, structural damage; at low levels, it interferes with neuro transmitters. A good argument can be made that, by studying the effects of lead on neurotransmitters, we may get a better understanding of the causes of attention disorders and hyperactivity. This review has also touched on a number of remaining paradoxes. Early on, it was argued that lead exposure would have its strongest effect on the brain when it was developing most rapidly, that is at and around birth. However, the prospective studies are finding that relationships between lead exposure and cognitive functioning disappear or become inconsistent by 4 to 5 years. Against this, Needleman et al.'s (1990) 11-year follow-up study finds impressive relationships between earlier exposure and socially meaningful pathology in early adulthood. Just over 10 years ago, it was advised that it was unethical to take blood samples from children who were not at high risk of having some problem. This edict in part made the use of estimates from teeth more attractive. In turn, it was argued that teeth were better indicators of life-long exposure. The data now show that blood lead levels are much more stable than was once thought and that they may be more strongly correlated with contemporaneous measures of functioning. Scientists are still left with the dilemma that different indices are only partly correlated.
Review: neurotoxicity of lead 335
Some of the very early studies of clinically poisoned children used positive effects of chelation as an argument that lead (at least at high doses) had effects on children's behaviour. Recent evidence from animal studies (Cory-Schlecta et al. 1987) suggests that Ca disodium EDTA mobilizes lead but this then gets redeposited in vital organs such as the brain and liver. Such findings make one wonder how the early chelation studies showed such dramatic effects and whether, on longer term follow-up, the effects would persist. At the very least, the findings remind one that there is still a long way to go in understanding the effects of lowlevel lead exposure on children's development, but a great deal has been learned in the past 10 years. REFERENCES Bellinger D., Leviton A., Waternaux C , Needleman H. & Rabinowitz M. (1987) Longitudinal analyses of prenatal and postnatal lead exposure and early cognitive development. New England Journal of Medicine 316. 1037-1043 Bellinger D., Leviton A. & Sloman J. (1990) Antecedents and correlates of improved cognitive performance in children exposed in utero to low levels of lead. Environmental Health Perspectives 89, 5-11 Byers R.K. & Lord E.E. (1943) Late effects of lead poisoning on mental development. American Journal of Diseases of Childhood 66. 194-471 Cohen J. (1977) Statistical Power Analysis for ihe Social Sciences. Academic Press, New York Cory-Schlecta D.A., Weiss B. & Cox C. (1987) Mobilization and redistribution of lead over the course of calcium disodium ethylenediaihine tetraacetate chelation therapy. Journal of Pharmacology and Experimental Therapeutics 243, 804-813 Cox A., Rutter M., Yule B. & Quinton D. (1977) Bias resulting from missing information: some epidemiological flndings. British Journal of Preventive and Social Medicine 31, 131-136 Department of Health and Social Security (DHSS) (1980) Lead and Health. The Report of a Working Party on Lead in the Environment (Lawther Report). HMSO. London Dietrich K.N., Succop P.A., Bornschein R.L., Krafft K.M., Berger C , Hammond P.B. et al. (1990) Lead exposure and neurobehavioral development in late infancy. Environmental Health Perspectives 89, 13-19 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988a) A longitudinal study of , dentine lead levels, intelligence, school performance and behaviour: Part I. Dentine lead levels and exposure to environmental risk factors. Journal of Child Psychology and Psychiatry 29, 781-792 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988b) A longitudinal study of dentine lead levels, intelligence, school performance and behaviour: Part II. Dentine lead and cognitive ability. Journal of Child Psychology and Psychiatry 29, 793-810 Fergusson D.M., Fergusson J.E., Horwood L.J. & Kinzett N.G. (1988c) A longitudinal study of dentine lead levels, intelligence, school performance and behaviour: Part III. Dentine lead levels and attention/activity. Journal of Child Psychology and Psychiatry 29, 811-824 Lansdown R. & Yule W. (eds) (1986) The Lead Debate: The Environment, Toxicology and Child Health. Croom Helm. BeckenlTam, Kent
336 W. Yule
Lansdown R., Yule W., Urbanowicz M.A. & Millar I.B. (1983) Blood lead, intelligence, attainment and behaviour in school children: overview of a pilot study; In Lead Versus Health: Sources and Effects of Low Level Lead Exposure, eds M. Rutter & R. Russel Jones. J. Wiley, Chichester Lyngbe T., Hansen O.N. & Grandjean P. (1990a) Predictors of tooth-lead level with special reference to traffic: a study of lead-exposure in children. International Archives of Occupational and Environmental Health 62, 417-422 Lyngbe T., Jorgensen P.J., Grandjean P. & Hansen O.N. (1990b) Validity and interpretation of blood lead levels: a study of Danish school children. Scandinavian Journal of Clinical Laboratory Investigations SO, 441-449 McMichael A.J., Baghurst P.A., Wigg, N., Vimpani G.V., Robertson E.F. & Roberts R.J. (1988) Port Pirie cohort study: environmental exposure to lead and children's abilities at the age of four years. New England Journal of Medicine 319, 468-475 Needieman H.L. & Gatsonis C.A. (1990) Low-level lead exposure and the IQ of children: a meta-analysis of modern studies. Journal of the American Medical Association 263, 673-678 Needieman H.L., Gunnoe C , Leviton A.. Reed M., Peresie H., Maher C. et al. (1979) Deficits in psychological and classroom performance of children with elevated dentine lead levels. New England Journal of Medicine 300, 689-695 Needieman H.L., Schell, A., Bellinger D., Leviton A. & Allred E.N. (1990) The long-term effects of exposure to low doses of lead in childhood: an 11-year follow-up report. New England Journal of Medicine 322, 83-88 Norby Hansen O., Trillingsgaard A., Beese I., Lyngbe T. & Grandjean P. (1989) A neuropsychoiogical study of children with elevated dentine lead level: assessment of the effect of lead In different socio-economic groups. Neurotoxicology and Teratology 11, 205-213 Royal Commission on Environmental Pollution (1983) Ninth Report: Lead in the Environment. HMSO, London Rutter M. (1980) Raised lead levels and impaired cognitive/behavioural functioning: a review of the evidence. Developmental Medicine and Child Neurology 22 (suppl.), 42 Rutter, M. & Russell Jones R. (eds) (1983) Lead Versus Health: Sources and Effects of Low Level Lead Exposures, i. Wiley. Chichester Silva P.A., Hughes P., Williams S. & Faed J.M. (1988) Blood lead, intelligence, reading attainment, and behaviour in eleven year old children in Dunedin, New Zealand. Journal of Child Psychology and Psychiatry 29, 43-52 Smith M., Delves T., Lansdown R., Clayton B. & Graham P. (1983) The effect of lead exposure on urban children. Institute of Child Health/University of Southampton study. Developmental Medicine and Child Neurology 25 (suppl.), 47 Smith M.A., Grant L.D. & Sors A.I. (eds) (1989) Lead Exposure and Child Development: An International Assessment. Kluwer, London Vahter M. (ed.) (1982) Assessment of Human Exposure to Lead and Cadmium Through Biological Monitoring. National Swedish Institute of Environmental Medicine and Department of Environmental Hygiene, Karolinska Institute, Stockholm Winneke G., Hrdina K.-G. & Brockhaus A. (1982) Neuropsychoiogical studies in children with elevated tooth-lead concentrations I. Pilot study. International Archives of Occcupational and Environmental Health 51. 169-183 Winneke G., Kramer U., Brockhaus A., Ewers U., Kijanek G.. Lechner H. et al. (1983) Neuropsychoiogical studies in children with elevated tooth-lead concentrations II. Extended study. International Archives of Occupational and Environmental Health 51,231-252 Winneke G., Brockhaus A., Ewers U., Kramer U. & Neuf M. (1990) Results from the European multicenter study on lead neurotoxicity in children: implications for risk assessment. Neurotoxicology and Teratology 12, 553-559 Yule W. & Rutter M. (1985) Effects of lead on children's behaviour and cognitive performance:
Review: neurotoxicity of lead 337
a critical review. In Dietary and Environmental Lead: Human Health Effects, ed. K.R. Mahaffey. Elsevier, Amsterdam Yule W., Lansdown R., Millar I.B. & Urbanowicz M.A. (1981) The relationship between blood lead concentrations, intelligence and attainment in a school population: a pilot study. Developmental Medicine and Child Neurology 25. 567-576
