Methods

Sample

The study sample was derived from the Avon Longitudinal Study of Parents and Children (ALSPAC), a population-based study investigating environmental and genetic influences on the health, behaviour and development of children. All births in the former Avon Health Authority area with an expected date of delivery between 1st April 1991 and 31st December 1992 were eligible for ALSPAC, resulting in a total cohort of 14 062 surviving live births. The social and demographic characteristics of this cohort were similar to those found in UK national census surveys.16 Ethics approval for the study was obtained from the ALSPAC Law and Ethics Committee and the Local Research Ethics Committees. Further details about ALSPAC can be found on its web site (http://www.bris.ac.uk/alspac).

From the population cohort, a 10% randomly selected sample of parents whose babies were born within the last 6 months of the survey were invited to bring their children to a research clinic (Children in Focus, CIF) at 4, 8 and 12 months of age and 6-monthly intervals thereafter, where a number of clinical, physiological and psychological assessments were carried out. For this study, data were obtained from infants attending the CIF clinic at age 30 months. Parental consent was sought for venous blood samples, and was given for 81% of the 1135 children in the CIF group. A venous blood sample was collected in lead-free tubes from 71% (n = 653) of attendees at the clinic; however, 69 samples were insufficient, thus leaving 582 samples for lead analysis.

Lead levels

The blood lead level was measured at Southampton General Hospital, England, by atomic absorption spectrometry using micro sampling flame atomisation.17 The procedures for quality assurance in the Southampton laboratory were based on those used successfully for earlier studies for the Department of the Environment.18 For quality control, about 20% of the samples were reanalysed within the same laboratory, by electrothermal atomisation and atomic absorption spectrometry. The mean difference on these repeat analyses was 0.4 μg/dl, and 97% of the results agreed to within ±2 μg/dl. Approximately 10% of the samples were sent to an external laboratory (Glasgow Royal Infirmary, Scotland) for confirmatory analysis by a different ETA-AAS method: the mean difference between the Southampton and Glasgow laboratories was ±0.2 μg/dl and only one result differed by >2.5 μg/dl. The frequency distribution of the blood levels of lead is shown in fig 1. The majority of the blood lead levels (94%) were below 10 μg/dl.

Outcomes

Follow-up data were collected by questionnaires completed by parents and teachers, and from tests undertaken at special research clinics when the children were 7 and 8 years old. The continuous measures used are listed in appendix 1a, and categorical measures in appendix 1b (see online supplements). Child behaviour was assessed using the Strengths and Difficulties Questionnaire (SDQ)19 at age 7 years, the Development And Well-being Assessment (DAWBA)20 at age 8 years, and the Anti-social Behaviour Interview21 at age 8 years. Attention was measured using the Test of Everyday Attention for Children (TEACh)22 at age 8 years. The educational performance of the children was captured from the results of the national Standard Assessment Tests (SATs)23 at age 7 years (Key Stage 1). These tests are applied at school to all children in mainstream education in the UK, and at this age level 2 scores are average, with a range from level 1 to 4.

Confounders

A number of variables, both continuous and categorical, were included in the analyses as potential confounders of the relationship between lead concentrations and developmental outcome (see online appendices) These were: gender, the child’s IQ (measured by the Wechsler Intelligence Scale For Children (WISC-III) at age 8 years), maternal educational qualification (which was categorised as none/Certificate of Secondary Education, vocational, Ordinary level General Certificate of Education, Advanced level General Certificate of Education or degree), home ownership, maternal smoking, a home facilities score at 6 months, paternal socio-economic status at the time of pregnancy (the current or last main job of the partner), Family Adversity Index (a cumulative index of adversity including housing quality, financial difficulties, partner relationships, maternal mental health, education, criminality, excess alcohol/drugs) and parenting attitudes at 6 months. These variables were gathered prospectively from self-report questionnaires completed by the study mothers at regular intervals throughout the study period. Initial analyses were performed including and excluding IQ as a confounder; no statistical differences were apparent and so IQ is not included in the final results.

Statistical analyses

The study had 80% power to detect a difference of 0.35 SD between two groups of 125 (we had four unequal exposure groups), or an 80% power to detect an odds ratio (OR) of 2.07 (or 0.48) between two groups of 125 each.

The role of blood lead level was investigated as a continuous covariate and as a categorical variable. When used as a covariate, the data were transformed to a log scale to take account of the non-linear relationship. These analyses provided a more powerful test of the overall relationship with outcomes. When analysed as a categorical variable, no specific dose–response relationship was assumed. Each category’s effect was estimated to give the best explanation of the outcome. This was useful in investigating potential thresholds of blood lead levels below which no discernible deterioration in performance was observed. Categories of 0 to <2 (referred to as 0–2), 2 to <5 (2–5) and 5 to <10 (5–10) μg/dl were chosen to create groups below the current level of concern.

Outcomes were analysed using regression analysis to allow easier interpretation of effect sizes. Categorical outcomes were recoded as integers with equal increments so as to reflect their ordinal sequence. To check the robustness of these analyses, they were repeated using ordinal regression.24

Results

Case ascertainment

Of the 582 cases where lead concentrations were assessed, 488 cases (84%) had complete data on confounders and at least one valid observation on outcomes. In practice, missing data for the outcomes meant that 337–425 cases (58–73%) were used in the analyses (fig 2).

Comparisons of the characteristics of the 488 cases with the rest of the ALSPAC cohort are shown in table 1. As expected in a longitudinal study after 8 years of follow-up, the cases ascertained were from families where the mother was better educated and smoked less and was more likely to be a homeowner, and where there was a better home environment with fewer adversities.

Appendices 1a and 1b show the correlations with log lead concentration, demonstrating the unadjusted associations. Only SATs and TEACh scores showed any indication of an association using this simple linear model. For confounders, the correlations may be interpreted as a measure of the extent of confounding. In five out of the eight variables, there were strong associations with log lead, emphasising the importance of adjustment.

Lead as a continuous exposure

After adjustment and using a log dose–response model for lead concentration, reading, writing and spelling scores (SATs) and antisocial activities showed significant associations with lead levels (table 2). The effect sizes appeared small, with a 100% change (doubling) in blood lead level being associated with a 0.3 point (95% CI −0.5 to −0.1) decline in SATs grade.

Lead as a categorical exposure

Table 2 shows the relationship between categorical lead concentrations for the main outcomes.

There is evidence of a threshold up to 5 μg/dl, with estimates in this range not being significantly different and with the estimates for the five outcomes with significant results being split 4/1 as beneficial/detrimental. Above 5 μg/dl, SATs outcomes deteriorated, but the two behavioural outcomes (hyperactivity and anti-social behaviour) did not show marked deterioration until above 10 μg/dl.

The effect of doubling exposure from 5 to 10 μg/dl caused a decline in SATS scores for writing of 0.2 points (95% CI −0.03 to −0.8). It was also associated with an increase in hyperactivity scores reported by the teacher of 0.3 points (95% CI −0.9 to 0.63). To illustrate the dose–response effects considered in this study, a graph of writing versus lead concentration is shown in fig 3.

The parent was asked in the 38-month questionnaire if their child showed pica behaviour. The presence of this behaviour appeared to have a detrimental effect on SATs results, although none of the results were statistically significant. To investigate the mediating effect of lead on pica behaviour (ie, pica behaviour may cause higher lead concentrations), analyses were repeated without log lead concentration in the model. These results suggest that any effect of pica behaviour is not mediated through higher lead concentrations. Indeed, concentrations tended to be lower in those exhibiting this behaviour (1.14 vs 1.24 μg/dl, p = 0.6).

Discussion

Our results confirm that exposure to lead early in childhood, even at low levels, is harmful to subsequent behaviour and school performance. Although the effect sizes were small, the associations demonstrated are important at a population level. Our study also provides evidence for a threshold effect, and suggests that the level of concern should be reduced from 10 μg/dl to 5 μg/dl.

The strengths of this study are that the sample was drawn from a well-documented representative cohort, with information on a large number of potential confounders, and the outcomes used well validated scales. The analysis of lead in two different laboratories was also well validated. The main limitation is the missing data, as in many other longitudinal studies, resulting in some selection bias as the families who came to the follow-up research clinic and gave blood samples were better educated and better off financially. Given the recognised associations of lead toxicity with socio-economic deprivation and poor housing,8 this selection bias has probably resulted in an underestimate of the impact of lead on educational and behavioural outcomes. Also, we only had one measure of lead available when the children were 30 months old, which may not be the critical age for exposure and limits the capacity of the study to assess the impact of lead on development.

Our results are consistent with other studies25 26 showing effects on behaviour and cognition of blood lead levels under 10 μg/dl. Although the effects of lead appear small compared to the impact of parenting and social factors on educational attainment, they are detectable many years after exposure. The clinical importance of these findings is that exposure to lead may interact with other environmental factors associated with educational disadvantage to have a cumulative long term impact. Volumetric analyses of whole brain MRI data have shown significant decreases in frontal lobe volume in adulthood associated with childhood blood lead concentrations.27

Lead has been removed from paint and petrol by law in the UK, but it is still widespread in the environment. Young children who are especially susceptible to lead ingestion through oro-motor exploratory stimulation include those with cognitive impairments, autistic spectrum disorders and those who display pica behaviour. Once diagnosed as having lead poisoning, affected children should be monitored regularly and the local environmental health department (Health Protection Agency) involved to find the source and offer appropriate lead abatement advice and strategies: removal of old paint, replacement of old pipes, wet cleansing of floors and surfaces, washing of soft toys and use of a vacuum cleaner with filters are all recommended. Affected children should have regular developmental assessments and be offered appropriate behavioural and educational support. However, the mainstay of management of lead toxicity is primary prevention, through the education of carers about possible sources of lead, not only from old pipes, paint and house dust but also from painted toys. Increased surveillance data through more public health reporting, especially at levels below the current level of concern, will give a better estimate of the true extent of the problem. Further legislative measures should be taken to limit the lead content of products to which children are exposed.

Conclusion

Lead poisoning is a continuing hazard and should be considered in children presenting with behavioural or educational difficulties.28 Early childhood exposure to lead affects later educational attainment and behaviour even at low blood levels (5–10 μg/dl), and the level of concern should be lowered to 5 μg/dl.