Dentine lead and intelligence prior to school entry: a statistical sensitivity analysis

The relationship between circumpulpal dentine lead and IQ at age 4 years 10 months was examined in 164 urban children. Negative correlations were observed between dentine lead and IQ scores, but were reduced in magnitude after adjustment for social covariates. Analyses using errors-in-variables regression models indicated that the degree of this reduction depended on difficult-to-verify assumptions regarding the magnitudes of (i) measurement error in the lead variable and the covariates, and (ii) possible changes in the caretaking environment resulting from conjectured effect of lead on the child's cognitive and behavioral development. Sensitivity analyses were conducted in which estimates of the lead effect were repeatedly computed and compared for a range of possible values for factors (i) and (ii). It was found that the statistical significance of the lead effect on Full Scale IQ depended on the relative magnitudes of these factors, and that failure to incorporate measurement error in the analysis would have led to gross overestimation of the precision of the findings.