EPA/OPP Teleconferences on Utility of Quantitative Bias Analysis of the CCCEH Study
December 12, 2017 and February 1, 2018
                                       
Participants
EPA/OPP: Jeff Dawson, David Miller, Dana Vogel, Hayley Hughes, Nicole Zinn, James Nguyen, Yu-Ting Guilaran, Kevin Costello
Emory University: Dr. Timothy Lash

Summary
EPA held calls with Dr. Timothy Lash on 12/14/17 and 2/1/18 to discuss the utility and feasibility of using quantitative bias analysis (QBA) to analyze confounders and interpret other aspects of the Columbia Center for Children's Environmental Health (CCCEH) study. Further statistical analyses outside of QBA were discussed. The EPA committed to preparing a list of questions for CCCEH and again requesting the raw data in order to conduct these analyses.      

Key Discussion Topics

 EPA asked whether QBA could be done to confirm analysis of confounders. Dr. Lash said that a QBA could be done with an analysis between lead and measured chlorpyrifos exposures, and information on how lead effects neurological development.  However, Dr. Lash confirmed his opinion that lead not a great confounder in that it had been remediated and is not strongly associated with the exposure. 

 Dr. Lash raised a concern of selection bias (due to loss-to-follow up), suggesting EPA follow up with CCCEH to determine if the study participants were informed their exposure levels and test results of their children at 3 years old. He noted QBA could be useful in interpreting potential loss to follow up/retention of cohort participants and whether their decision to remain in the cohort or remove themselves was dependent upon results of cognitive testing in their children. 

 Dr. Lash indicated statistical analyses outside of QBA related to treatments of limit of detection data and LOQs are not warranted because CCCEH PIs imputed the non-detects. Dr. Lash thought was the best approach to handling the data.

 EPA and Dr. Lash discussed whether additional statistical analyses were warranted (i.e. whether Working Memory Index data and CPY exposure data should be transformed (or not), the impact of rounding).  Dr. Lash indicated they could only be performed if the raw data from the study was obtained. 
       

