Meta-Analysis Comparing National Children's Study Values to Comparable Literature Values

• Activity Number: 554
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): John Rogers*
• Companies: Westat
• Keywords: Meta analysis ; National Children's Study ; mixed model

Abstract

Biospecimen handling methods proposed for large multi-site epidemiological studies like the National Children's Study (NCS) should be validated to ensure they provide measurements similar to other published results. We developed regression models to compare blood concentrations of HbA1c, fasting glucose and fasting insulin from pregnant women in the NCS Vanguard Study with concentrations in similar subpopulations in peer-reviewed publications. The data were compared based on means and standard deviations. The models developed account for measurement error, differences between subpopulations reported in the same publication, and differences between publications. The measurements are roughly log-normally distributed; thus the distribution of reported means and standard deviations are skewed. This paper uses simulation to compare various statistical models considered for the analysis and presents the reasons for selecting the final model. We show that the blood concentrations of HbA1c, fasting glucose and fasting insulin from pregnant women in the NCS Vanguard Study are similar to reported values from other studies.