The National Health and Medical Research Council recently reported that there were not enough high-quality studies to conclude that associations between health effects and blood lead levels <10 μg/dL were caused by lead. It identified uncontrolled confounding, measurement error and other potential causal factors as common weaknesses. This paper supports those findings with evidence of uncontrolled confounding by parental education, intelligence or household management from several papers. It suggests that inappropriate statistical tests and aggregation of data representing different exposure routes partly explain why confounding has been overlooked. Inadequate correction of confounding has contributed to incorrect conclusions regarding causality at low levels of lead. Linear or log-linear regression models have tended to mask any threshold. While the effects of higher levels of lead exposure are not disputed, overestimation of health effects at low lead exposures has significant implications for policy-makers endeavouring to protect public health through cost-effective regulations.
Keywords: Lead epidemiology; bias; confounding; statistical analysis; threshold.