Lifetime exposure to environmental lead and children's intelligence at 11-13 years: the Port Pirie cohort studyBMJ 1996; 312 doi: https://doi.org/10.1136/bmj.312.7046.1569 (Published 22 June 1996) Cite this as: BMJ 1996;312:1569
- Shilu Tong, doctoral studenta,
- Peter Baghurst, principal research scientista,
- Anthony McMichael, professor of epidemiologyb,
- Michael Sawyer, directorc,
- Jane Mudge, doctoral studenta
- a Division of Human Nutrition, Commonwealth Scientific Industrial Research Organisation, Adelaide, SA 5000, Australia
- b Department of Epidemiology and Population Sciences, London School of Hygiene and Tropical Medicine, London WC1E 7HT
- c Evaluation Unit, Women's and Children's Hospital, Adelaide, SA 5000, Australia
- Correspondence to: Dr S Tong, Department of Health Services Management and Public Health, University of New England, Armidale, NSW 2351, Australia.
- Accepted 3 April 1996
Objective: To examine the association between environmental exposure to lead and children's intelligence at age 11-13 years, and to assess the implications of exposure in the first seven years of life for later childhood development.
Design: Prospective cohort study.
Subjects: 375 children born in or around the lead smelting town of Port Pirie, Australia, between 1979 and 1982.
Main outcome measure: Children's intelligence quotient (IQ) measured at 11-13 years of age.
Results: IQ was inversely associated with both antenatal and postnatal blood lead concentrations. Verbal, performance, and full scale IQ were inversely related to blood lead concentration with no apparent threshold. Multivariate analyses indicated that after adjustment for a wide range of confounders, the postnatal blood lead concentrations (particularly within the age range 15 months to 7 years) exhibited inverse associations with IQ. Strong associations with IQ were observed for lifetime average blood lead concentrations at various ages. The expected mean full scale IQ declined by 3.0 points (95% confidence interval 0.07 to 5.93) for an increase in lifetime average blood lead concentration from 0.48 to 0.96 µmol/l (10 to 20 µg/dl).
Conclusions: Exposure to environmental lead during the first seven years of life is associated with cognitive deficits that seem to persist into later childhood.
Few longitudinal data have yet been reported on the time course of the effects of exposure to environmental lead
Exposure to environmental lead early in life is associated with cognitive deficits that persist into middle childhood
The duration, intensity, and timing of exposure to lead, as well as other social and familial factors, may influence the nature and degree of reversibility
The formulation of a public health policy for preventing any possible effects of lead exposure should be based on a composite consideration of the child's health and the best use of existing resources
Many studies have reported inverse associations between low level lead exposure and neuropsychological development, particularly cognitive function.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 The accumulation of this evidence has prompted public health authorities in several countries progressively to lower the blood lead concentrations at which environmental intervention and medical evaluation is warranted.18 19 Since both Australian and American data indicate that the high childhood blood lead concentrations of 13-15 years ago are definitely decreasing,19 20 a contemporary question of great interest is whether the effects of early exposure to lead still persist into later life when lead exposure is generally much lower.
The Port Pirie cohort study started in 1979. Within this cohort, the geometric mean blood lead concentration in the children increased from 8.3 µg/dl (0.40 mmol/l) at birth (umbilical cord blood) to 21.2 µg/dl (1.02 mmol/l) at age 2 years, and had decreased to 11.6 µg/dl (0.56 mmol/l) by the age of 7 years.11 In previous studies, we reported that after adjustment for a wide range of confounding factors, postnatal blood lead concentrations in this cohort of children were inversely associated with scores obtained with the Bayley scales of infant development at age 2, the McCarthy scales of children's abilities at age 4, and the Wechsler intelligence scale for children at age 7.9 10 11 This paper reports the findings obtained when the children's intelligence quotients (IQ) were reassessed at age 11-13 years.
The initial sample comprised 723 singleton children born in and around the city of Port Pirie (site of one of the largest lead smelting facilities in the southern hemisphere) during a three year period from 1979 to 1982. These infants represented an estimated 90% of singleton live births in the community during this period.
Blood samples for measurement of lead concentrations were taken from pregnant women at specific stages of pregnancy and from each child at delivery (from the umbilical cord) and at ages 6, 15, and 24 months, and annually thereafter up to age 7 years. Measurements of developmental status of the children were made at ages 2, 4, and 7 years; full details are provided in the reports of the results.9 10 11
The base population for the present study comprised the 494 children who were assessed at age 7 years (fig 1). Of the 22 children who were considered ineligible, 21 were excluded because they had missed more than two blood lead measurements during their first seven years, and one was excluded on the basis of having suffered a head injury.
Of the 472 eligible children, 55 (11.7%) could not be contacted (34 had moved and 21 could not be reached despite intensive efforts to contact them at home); 37 (7.8%) refused to participate; and a further five (1.0%) were contacted but were living too far away for the assessor to be sent. The number of children finally assessed was 375 (79.4%). The median age of the children on the day of testing was 12.0 (SD 0.7) years (25th and 75th centiles 11.4 and 12.5 years).
The 375 children who were evaluated at age 11-13 years did not differ significantly from the 119 children lost to follow up on most characteristics, including demographic, socioenvironmental, and biomedical factors. However, parents of the children remaining in the cohort at age 11-13 years had slightly higher occupational prestige than those lost to follow up (table 1).
The possibility of bias in the estimated lead-IQ relation due to a selective loss to follow up was evaluated by comparing the estimated Pearson correlation coefficient of lifetime average blood lead concentration at age 7 with IQ (also at age 7) in the children lost to follow up in those still in the cohort at age 11-13. The estimates were almost identical: -0.27 for those lost to follow up and -0.25 for the children remaining in the cohort.
The revised version of the Wechsler intelligence scale for children was used to assess the cognitive abilities of each child at age 11-13 years.21 This scale is a test of general intelligence developed for use with children aged 6-16 years. In addition to providing a rating of a child's overall intelligence, the test includes a verbal IQ, which provides a rating of verbal comprehension, and a performance IQ, which provides a rating of a child's perceptual organisation. There are 12 subtests within the scale. Scores for the information, similarities, arithmetic, vocabulary, and comprehension subtests contribute to the verbal IQ; performance IQ is assessed from the scores obtained with the picture completion, picture arrangement, block design, object assembly, and coding subtests. Two further subtests, digit span and mazes, are supplementary and provide additional qualitative and quantitative information about a child's intellectual abilities.21
To reduce measurement bias, quality control procedures were undertaken throughout the study. Firstly, the subjects were evaluated by the same trained examiner, who had not participated in earlier phases of the cohort study and was unaware of any aspect of the children's developmental histories and exposure to lead. Secondly, all assessments were conducted according to a standardised protocol. Finally, children with different exposure status in early life were assessed in an intermixed order.
Mean IQ scores showed a slight downward drift (not statistically significant) toward the end of the study, which suggests that intraexaminer variability over time was unlikely to have been a problem in this study.
MEASUREMENTS OF LEAD EXPOSURE
Venous blood samples were obtained at age 11-13 years, and estimations of blood lead concentrations were performed in the department of chemical pathology at the Adelaide Women's and Children's Hospital by electrothermal atomisation atomic absorption spectrometry.22 The blood lead assay was subject to internal and external quality control procedures, with consistently satisfactory results.10 11 A certified, commercially prepared product was used to monitor intrabatch accuracy and to ensure uniformity between batches (the coefficient of variation for measurements was </=5.7%). External quality control, entailing assays of regularly supplied samples, was ensured by participation in the international programmes run by the health department of Pennsylvania, USA, and the Wolfson Research Laboratory (Birmingham, UK). Estimates were standardised to a packed cell volume of 50% for cord blood and 35% for all other samples.
MEASUREMENT OF COVARIATES
An important methodological consideration is the selection and measurement of covariates that influence childhood intelligence and might be related to lead exposure, and which might therefore confound any true relation between lead exposure and IQ. The covariates considered in this study included parents' occupational prestige, which was assessed using the Daniel scale of prestige of occupations in Australia (lower scores equate to more prestigious jobs, which are generally associated with higher socioeconomic status)23; the child's family functioning, which was measured using the general functioning scale of the family assessment device24; mother's psychopathological status, which was evaluated with the general health questionnaire25; the caregiving environment, which was assessed with the home observation for measurement of the environment inventory26; and maternal intelligence, which was estimated using the Wechsler adult intelligence scale.27 Other factors that were measured included the child's iron status, sex, age at testing, and school grade; marital status of the parents; parental smoking habits and parents' years of secondary education; family size; life events; length of mother's residence in Port Pirie; assessment site; maternal age at delivery; birth weight; birth rank; feeding method during infancy (breast, formula, or both); duration of breastfeeding, whether any medication was used in the two weeks before testing; and prolonged absences from school for any single school term during the past five years.
All reported mean blood lead values are geometric because of their approximate log normal distribution. For calculation of lifetime average blood lead concentrations, a plot of blood lead concentrations against age was constructed for each child. A lifetime average blood lead concentration up to a particular age was estimated by dividing the area under this curve by the specified age.
For describing results in tabular form, blood lead was categorised into thirds at each time of blood sampling. In multivariable analyses, the effects of confounding factors were evaluated with multiple regression models. The selection of potential confounders was based on both a priori and empirical considerations.28 29 Firstly, important antecedents or correlates of IQ were considered as potential confounders (for example, scores on the home observation for measurement of the environment scale and maternal IQ). Secondly, the variables that were associated, independently, with both blood lead and developmental outcomes within this data set were considered as potential confounders. Thirdly, a change in estimate criterion was used to evaluate each potential confounder: if the adjusted estimated regression coefficient of the blood lead term differed from the unadjusted estimate by more than 10%, that covariate was included in the final regression model. Finally, the combined effect of potential confounders was assessed by the change in magnitude of the regression coefficients of the lead term after all the potential confounding factors had been adjusted for. The covariates used in the multiple regression models comprised sex, age, school grade, parents' occupational prestige, home observation for measurement of the environment scores, maternal IQ, family functioning scores, parental smoking habits, marital status, parents' level of education, maternal age at delivery, birth weight, birth rank, feeding method during infancy, duration of breastfeeding, family size, life events, and prolonged absences from school for any single school term during the past five years.
Table 2 shows the variation of geometric mean blood lead concentrations with age. A sharp increase in blood lead concentration occurred during the first two years of life and was followed by a gradual decline, with the mean concentration at 11-13 years being slightly lower than the level recorded at birth.
The mean scores for the 12 subscales of the Wechsler intelligence scale for children ranged from 9.2 to 11.3, and the mean scores for verbal, performance, and full scale IQ were 97.6 (95% confidence interval 96.6 to 98.7), 103.0 (101.6 to 104.4), and 100 (98.8 to 101.2), respectively (table 3).
AGE SPECIFIC BLOOD LEAD CONCENTRATION AND CHILDRENS' IQ
There was a consistent inverse relation between blood lead concentration and scores for all the IQ scales, unadjusted for covariates (table 4). IQ was significantly associated with blood lead concentrations at all ages except at birth. Decreases in mean IQ scores between the top and bottom thirds of exposure varied from 1.5 points for performance IQ with cord blood lead concentration to 9.1 points for full scale IQ with lifetime average blood lead concentration at age 11-13 years.
The magnitude of the deficit in IQ with increased blood lead was similar for the verbal and performance scales. There was an inverse and statistically significant association between children's IQ and blood lead concentrations measured at most ages, although the relatively larger IQ deficits were associated with blood lead concentrations measured at earlier ages and with lifetime average blood lead concentration. The proportion of the variance of full scale IQ that could be attributed to blood lead concentrations at different ages, without consideration of the potential confounders, varied from 0.8% (cord sample) to 10.1% (lifetime average value).
Figure 2 (top) shows the unadjusted relation between children's mean IQ and lifetime average blood lead concentration. Generally, for each IQ scale, there was an inverse gradient across the whole range of blood lead concentration without an apparent threshold. The trend was similar for each IQ scale. Figure 2 (bottom) shows the variation of mean residual IQ (observed minus expected values) after IQ was regressed on the potential confounders used in this study. A dose related decrease in IQ with increasing blood lead concentration was still evident, although the range in residual IQ was substantially less than that seen with unadjusted IQ data.
REGRESSION ANALYSES OF IQ SCALE AND SUBSCALE SCORES
In simple regression analyses, all measures of blood lead except those for the cord sample were significantly inversely associated with IQ. The magnitude of regression coefficients of blood lead concentration for verbal IQ was similar to that for performance IQ (table 5).
In multiple regression analyses, the inverse associations between blood lead concentration and IQ were attenuated after the effects of potential confounders were adjusted for. In particular, the associations of children's IQ with maternal and cord blood lead concentrations became insignificant. The covariates contributing most to this attenuating effect were those identified as being most closely related to both blood lead and children's IQ—that is, socioeconomic status, scores on the home observation for measurement of the environment scales, and maternal intelligence. However, the inverse associations between various measures of blood lead concentrations over the age range 15 months to 7 years and IQ (mainly verbal and full scale IQ) remained statistically significant or marginally significant after potential confounders had been taken into account. The stronger associations were found between lifetime average blood lead concentrations at various ages and IQ. It is estimated that the mean score for full scale IQ declined by 3.0 points (0.07 to 5.93) for a doubling in lifetime average blood lead concentration at age 11-13 years within the range studied.
Multiple regression analyses of the Wechsler subscale scores show that most of the mean subscale scores were inversely associated with the lifetime average blood lead concentration at age 11-13 years (table 6). However, the strength of these associations varied considerably. The associations of lifetime average blood lead concentration with the information, arithmetic, block design, and maze subscales were stronger than those for any other subscales.
PERSISTENCE OF ASSOCIATION BETWEEN BLOOD LEAD CONCENTRATION AND CHILDREN'S DEVELOPMENT
Data on the histories of lead exposure and cognitive development in the children by thirds of lifetime average blood lead concentration up to age 2 years (the age when the children's developmental status was assessed for the first time in this cohort study) indicate that the adjusted differences in developmental scores between the top and bottom thirds of exposure were 4.0 points on the Bayley mental developmental index at age 2; 4.8 points on the McCarthy general cognitive index at age 4; and 4.9 and 4.5 IQ points at 7 years and 11-13 years, respectively, after potential confounders had been taken into account (fig 3).
In the latest stage of this prospective study we re-evaluated a group of children aged 11-13 years with lifelong histories of lead exposure and found that the inverse associations between blood lead and cognitive development at ages 2, 4, and 7 years9 10 11 persisted into later childhood. The estimated deficit in full scale IQ at age 11-13 years was 3.0 points for a shift in lifetime average blood lead concentration from 0.48 to 0.96 µmol/l (10 to 20 µg/dl).
This study provides evidence that an association between early exposure to environmental lead and cognitive development persists into later childhood, even though blood lead concentrations in these children had declined substantially since their third year of life. Nevertheless, both the initial effects of environmental lead exposure and the persistence of those effects may depend not only on the magnitude of exposure but also on the timing and chronicity of exposure and other social and familial factors.
Only a few studies have looked for a persistent effect of exposure to low levels of lead on cognitive development. A cohort study in Boston found a consistent association between cord blood lead and cognitive development at ages 6, 12, 18, and 24 months, but not at later ages.3 4 5 At ages 57 months and 10 years, only the blood lead concentration measured at 2 years of age was found to be associated with the child's development. The exposure related differences in performance at age 10 years were reported to be approximately twice the size of those observed at 57 months, but the relative magnitude of the association at age 10 years may have been distorted because of differences in characteristics between the continuing participants and those lost to follow up.5 Two other studies found that raised concentrations of lead in dentine were associated with deficits in neuropsychological functioning that persisted into later childhood or young adulthood.2 30 However, since dentine lead concentration was the only measure of exposure in these two studies, the chronology of change in lead exposure status is unknown, making it difficult to judge whether the persistent effects resulted from early exposure or accumulated exposure throughout a lifetime.
There are several possible explanations for the association observed in this study. Firstly, a rather special potential problem for studies of lead exposure and cognitive development is reverse causation—that is, a lead-IQ association which arises because children who have a lower IQ are allegedly more likely to exhibit behaviours (such as thumb sucking or poorer personal hygiene) that might result in increased lead intake.31 32 To explore this possibility we examined the temporal relations between lead exposure and children's IQ.
Blood lead measures most strongly related to IQ had all been measured before IQ was assessed (table 5), and none of the developmental scores at earlier ages were significantly associated with the current blood lead measure at age 11-13 years. These analyses—supported by animal research33 34 35—strengthen the notion that lead exposure influences cognitive development, and not the reverse.
Secondly, it might be argued that the association is an artefact of incomplete or inaccurate assessment of confounding factors.36 Though this concern can rarely be dispelled in any observational epidemiological study, persistence of the effects of early exposure to environmental lead has also been found in animal studies, where these confounders are not present.33 34 35
Thirdly, bias due to loss to follow up is unlikely to have an important role in this study because the characteristics of the children who were reassessed at age 11-13 years were similar (with respect to earlier measurements) to those lost to follow up, and the relations between blood lead concentrations and cognitive status at earlier ages were almost identical for these two groups.
THRESHOLD, INTERACTION, AND SENSITIVE INDICES
In assessing public health risk from exposure to environmental lead, there has been strong interest in the existence, or otherwise, of a level of exposure below which a toxicological effect does not occur. Within the range of blood lead concentration encountered in this study, there was no clear evidence that such a threshold exists.
An important issue in risk assessment of lead exposure is whether the apparent effects of lead exposure are greater in some subgroups, such as children of one sex or a particular socioeconomic background. Stratified analyses indicate that the association between lifetime average blood lead concentration and full scale IQ was stronger for girls (partial regression coefficient —7.4; 95% confidence interval -13.1 to-1.7) than for boys (-2.6; -8.0 to 2.9), which is consistent with our findings at ages 4 and 7 years.37 The results of further analyses also indicated that the regression coefficients of blood lead were more highly negative for children living in Port Pirie than those living elsewhere, and that children in families with lower socioeconomic status, from home environments of poorer quality, and with mothers of lower intelligence, seemed to be more affected by exposure to lead than those from more advantaged backgrounds—although these differences were not significant after adjustment for confounding factors.
Another important issue that is still being explored is which indices of outcome are the most sensitive to lead. Although global measures of cognitive function, such as full scale IQ, are generally regarded as reliable indices of the effects of lead, specific abilities seem to be more sensitive to the effects of low level exposure to lead than are global measures of cognition. In this study, four subtests of the Wechsler scale were found to be more strongly associated with lifetime average blood lead concentration than the others. Visual-motor coordination, attention, concentration, and memory are considered to be important contributors to these subtests.38 39
MECHANISMS FOR THE EFFECTS OF LEAD
The mechanisms by which lead might exert a toxic effect on cognitive development remain unclear. It has been shown that lead alters release processes for neuro-transmitters such as dopamine, norepinephrine, and acetylcholine by interfering with calcium metabolism or synaptic functioning, or both,40 41 42 and that lead exerts effects on the activities of such enzymes as kinase C, calmodulin, tyrosine hydroxylase, and choline acetyl-transferase, and also on brain energy metabolism.43 44 45 However, the direct relevance of these effects to the higher level processes of cognitive functioning assessed during developmental testing remains to be determined.
Within the context of other studies of the putative effects of lead on cognition it is pertinent to note that the estimates of “effect” produced by the Port Pirie cohort study are a little higher in magnitude than those reported by most, but not all, of the other studies.36 The problem of deciding whether these effect estimates are too high because of imperfections in the measurement of other determinants of IQ, or too low because of excessive statistical adjustment for factors which are causally related to exposure, is unlikely to be resolved in the near future. Objective assessment of the currently available epidemiological evidence suggests that any effect of subacute exposure to lead on cognition is likely to be modest. However, since any such effect may be persistent, and since human activities have already greatly increased the overall penetration of lead into the environment, it is entirely appropriate for public health authorities to continue to strive for a reduction in the dispersive uses of this heavy metal.
We are indebted to Mrs Maureen Wauchope for blood sampling and interviews, to Mr Charles Greeneklee for assessing packed cell volume of blood samples, to Ms Elaine Whitham for blood lead and iron analyses, to Mr Jim Lyster for supervising developmental assessments, and to the families who participated in this study.
Funding This research was supported by a series of grants from the National Health and Medical Research Council, the Channel 7 Children's Research Foundation, and the University of Adelaide.
Conflict of interest None.