Background: The Children’s Health Exposure Analysis Resource (CHEAR) program allows researchers to expand their research goals by offering the assessment of environmental exposures in their previously collected biospecimens. Samples are analyzed in one of CHEAR’s network of six laboratory hubs with the ability to assess a wide array of environmental chemicals. The ability to assess inter-study variability is important for researchers who want to combine datasets across studies and laboratories. Objective: Herein we establish a process of evaluating inter-study variability for a given analytic method. Methods: Common quality control (QC) pools at two concentration levels (A and B) in urine were created within CHEAR for insertion into each batch of samples tested at a rate of three samples of each pool per 100 study samples. We assessed these QC pool results for seven phthalates analyzed for five CHEAR studies by three different lab hubs utilizing multivariate control charts to identify out-of-control runs or sets of samples associated with a given QC sample. We then tested the conditions that would lead to an out-of-control run by simulating outliers in an otherwise “in-control” set of 12 trace elements in blood QC samples (NIST SRM 955c). Results: When phthalates were assessed within study, we identified a single out-of-control run for two of the five studies. Combining QC results across lab hubs, all of the runs from these two studies were now in-control, while multiple runs from two other studies were pushed out-of-control. In our simulation study we found that 3–6 analytes with outlier values (5xSD) within a run would push that run out of control in 65–83% of simulations, respectively. Significance: We show how acceptable bounds of variability can be established for a given analytic method by evaluating QC materials across studies using multivariate control charts.
|Number of pages||10|
|Journal||Journal of Exposure Science and Environmental Epidemiology|
|State||Published - Mar 2021|
- Analytical Methods
- Empirical/Statistical Models