Low-level environmental lead exposure in childhood and adult intellectual function: a follow-up study
Ian Wilson, Queensland Government (formerly)
25 May 2012
The paper by Mazumdar et al. [1] does little for the credibility of prospective health studies. It lacks accuracy and produces questionable conclusions. The report is based on 43 of the 249 individuals who were originally selected from 9489 births at a Boston hospital between 1979 and 1981. The authors indicate that their sample is too small to construct a model to test the significance of the potential confounding factors. As a result the conclusions are based on simple linear regressions between IQ and blood lead levels. Therefore, the statement that 'All analyses included prespecified covariates ¿' is not accurate. This is confirmed by a later statement that their models 'do not control for the effects of confounders.' Other minor errors include a statement that the maximal blood lead level was at age two while Table 1 shows that it occurred at 12 months and inconsistent references to one set of measurements as occurring at either 4 years or 57 months.
Earlier studies of this cohort showed that on adjustment for covariates, the apparent negative correlation coefficients between IQ and blood lead decreased in magnitude and most became non-significant [2]. Mazumdar et al. had access to a vast array of data on potential confounders and conducted linear regressions on these separately. They indicated that inclusion of maternal IQ in the model reduced the regression coefficient from ¿1.89 to ¿1.11 and changed its significance, but they did not indicate whether the relationship still remained significant. They were unable to test for further reduction of the significance of their key relationship by adding other potentially confounding factors such as HOME scores or maternal education to a multiple regression. The omission of confounding factors contributed to the following weaknesses in their findings:
¿ The apparent negative correlation between childhood lead exposure and adult IQ may be reduced in significance or eliminated by confounding factors.
¿ Even if that relationship is significant, their data do not show any significant effect on IQ at blood levels less than 10 ug/dL as they state that only five of the 43 participants had not recorded blood lead levels above this level.
¿ The apparent relationship between adult IQ and school age blood lead levels may not remain significant when corrected for covariates [2], hence there may not be evidence of greater susceptibility to lead exposure during school age years.
Their attempt to explain the relationship between maternal IQ and child IQ as a result of the mothers with higher IQ actively limiting lead exposure of their children or living where there is less likelihood of exposure may well be partially correct. However, that is not a reason to leave maternal IQ out of the analysis. In fact it should be used as a control variable potentially affecting both IQ and blood lead and the analysis should investigate the relationship between the residuals after controlling for maternal IQ. Their conclusion that a larger study should evaluate the effect of maternal IQ is a sensible way forward but it should include potentially confounding factors and control for any interrelationships between the factors.
1. Mazumdar M, Bellinger DC, Gregas M, Abanilla K, Bacic J, Needleman HL: Low-level environmental lead exposure in childhood and adult intellectual function: a follow-up study. Environ Health 2011, 10:24.
2. Bellinger DC, Stiles KM, Needleman HL: Low-Level Lead Exposure, Intelligence and Academic Achievement: A Long-term Follow-up Study. Pediatrics 1992, 96(6):855-861.
Competing interests
None declared
Response to comment by Dr. Wilson
Maitreyi Mazumdar, Boston Children's Hospital
25 May 2012
We agree with Dr. Wilson that the small sample size of our study limits the conclusions that can be drawn regarding associations between childhood blood lead concentrations and intellectual function in adulthood. In fact, we decided not to present results of models with multiple covariates because we believed our sample did not have sufficient power to justify the use of multivariate regression. We did, however, want to present data regarding potential confounders in this study that would be relevant to future studies with a larger number of participants. Thus, of the covariates we thought might be important, we sought to determine which ones affected the association between blood lead concentration and Full Scale IQ. We present that analysis in detail in Table 3. Maternal IQ was not left out of the analysis, but instead was highlighted as a potential confounder. To our knowledge, this study reports the longest period of follow-up of participants from the many influential studies of lead and child development initiated in the 1970s and 1980s. From a scientific perspective, it provides evidence that the effects of lead are persistent, albeit with the limitations of small sample size.
Environmental Health
Low-level environmental lead exposure in childhood and adult intellectual function: a follow-up study
25 May 2012
The paper by Mazumdar et al. [1] does little for the credibility of prospective health studies. It lacks accuracy and produces questionable conclusions. The report is based on 43 of the 249 individuals who were originally selected from 9489 births at a Boston hospital between 1979 and 1981. The authors indicate that their sample is too small to construct a model to test the significance of the potential confounding factors. As a result the conclusions are based on simple linear regressions between IQ and blood lead levels. Therefore, the statement that 'All analyses included prespecified covariates ¿' is not accurate. This is confirmed by a later statement that their models 'do not control for the effects of confounders.' Other minor errors include a statement that the maximal blood lead level was at age two while Table 1 shows that it occurred at 12 months and inconsistent references to one set of measurements as occurring at either 4 years or 57 months.
Earlier studies of this cohort showed that on adjustment for covariates, the apparent negative correlation coefficients between IQ and blood lead decreased in magnitude and most became non-significant [2]. Mazumdar et al. had access to a vast array of data on potential confounders and conducted linear regressions on these separately. They indicated that inclusion of maternal IQ in the model reduced the regression coefficient from ¿1.89 to ¿1.11 and changed its significance, but they did not indicate whether the relationship still remained significant. They were unable to test for further reduction of the significance of their key relationship by adding other potentially confounding factors such as HOME scores or maternal education to a multiple regression. The omission of confounding factors contributed to the following weaknesses in their findings:
¿ The apparent negative correlation between childhood lead exposure and adult IQ may be reduced in significance or eliminated by confounding factors.
¿ Even if that relationship is significant, their data do not show any significant effect on IQ at blood levels less than 10 ug/dL as they state that only five of the 43 participants had not recorded blood lead levels above this level.
¿ The apparent relationship between adult IQ and school age blood lead levels may not remain significant when corrected for covariates [2], hence there may not be evidence of greater susceptibility to lead exposure during school age years.
Their attempt to explain the relationship between maternal IQ and child IQ as a result of the mothers with higher IQ actively limiting lead exposure of their children or living where there is less likelihood of exposure may well be partially correct. However, that is not a reason to leave maternal IQ out of the analysis. In fact it should be used as a control variable potentially affecting both IQ and blood lead and the analysis should investigate the relationship between the residuals after controlling for maternal IQ. Their conclusion that a larger study should evaluate the effect of maternal IQ is a sensible way forward but it should include potentially confounding factors and control for any interrelationships between the factors.
1. Mazumdar M, Bellinger DC, Gregas M, Abanilla K, Bacic J, Needleman HL: Low-level environmental lead exposure in childhood and adult intellectual function: a follow-up study. Environ Health 2011, 10:24.
2. Bellinger DC, Stiles KM, Needleman HL: Low-Level Lead Exposure, Intelligence and Academic Achievement: A Long-term Follow-up Study. Pediatrics 1992, 96(6):855-861.
Competing interests
None declared
Response to comment by Dr. Wilson
25 May 2012
We agree with Dr. Wilson that the small sample size of our study limits the conclusions that can be drawn regarding associations between childhood blood lead concentrations and intellectual function in adulthood. In fact, we decided not to present results of models with multiple covariates because we believed our sample did not have sufficient power to justify the use of multivariate regression. We did, however, want to present data regarding potential confounders in this study that would be relevant to future studies with a larger number of participants. Thus, of the covariates we thought might be important, we sought to determine which ones affected the association between blood lead concentration and Full Scale IQ. We present that analysis in detail in Table 3. Maternal IQ was not left out of the analysis, but instead was highlighted as a potential confounder. To our knowledge, this study reports the longest period of follow-up of participants from the many influential studies of lead and child development initiated in the 1970s and 1980s. From a scientific perspective, it provides evidence that the effects of lead are persistent, albeit with the limitations of small sample size.
Competing interests
None declared