Assessing Calibration-Related Measurement Bias Near the Limit of Quantitation

Oral Presentation

Prepared by T. Strock1, S. Reimer2, D. Gregg3, W. Whipple1
1 - US Environmental Protection Agency, 536 S. Clark St., 10th Flr (ML-10C), Chicago, IL, 60605, United States
2 - US EPA Region X Manchester Laboratory, 7411 Beach Drive East (LAB), Port Orchard, WA, 98366, United States
3 - US EPA Region VI Laboratory, Houston Branch, 10625 Fallstone Road (6MD), Houston, TX, 77099, United States


Contact Information: strock.troy@epa.gov; 312-353-8362


ABSTRACT

For many applications, limiting measurement bias near the limit of quantitation is a desirable feature of analytical chemistry data generated with EPA reference methods. However, existing guidance regarding the influence of calibration options on measurement bias near the intercept of the calibration function is limited. For calibration models not forced through the origin, response per unit concentration changes in a non-intuitive way near the origin. For example, depending on the curve, a response of zero can produce a concentration near that of the low point of the calibration, or a peak that meets all qualitative identification criteria can produce a concentration ≤ 0.

When the intercept of the calibration function is near the low calibration point, measurement bias can create problems for interpreting sample data, including evaluating the influence of method blanks on low level sample results or recovery of low level standard additions. Incorporating evaluation criteria for the intercept of the calibration function can help laboratories identify and avoid underlying measurement bias issues that are not always evident based on calculated concentration. This evaluation may be particularly important for applications with regulatory or risk-based decision limits near the laboratory’s limit of quantitation. Examples will be presented for organic analytical methods from multiple EPA programs, including GC/MS and LC/MS. Emphasis will be placed on the intended use of the data as the key consideration for whether careful evaluation of measurement bias near the limit of quantitation is important, and recommendations will include proposed modifications to instrument manufacturers’ data processing software to include evaluation of the intercept of the calibration function as an additional parameter to consider during data evaluation.