Limitations of Low-level 2378-TCDD and 2378-TCDF Testing as Revealed by Statistical Analysis of Composite Calibration Curves, 1993 Environmental Conference Proceedings

Twenty-six instrument calibration curves used for the quantification of 2378-TCDD and 2378-TCDF in pulp and paper industry samples were collected from seven laboratories. Ml investigation of these data has revealed a statistically significantly higher response in the low concentration region, defined by x < 1="" pg="" analyte="" (1="" pg="l0^-12" gram)="" when="" compared="" to="" the="" region,="" x="">= 1 pg analyte. For 2378-TCDD the magnitude of the difference, 9%, is largely due to the presence of 2378-TCDD as a contaminant in the isotopically labelled standards. Isolated cases of integrations inaccurate peak and/or carryover are also believed to contribute to this effect.

These two regions of the calibration curve were also found to have significantly statistically different variances; the variance of the x < 1="" pg="" population="" is="" almost="" twice="" that="" of="" the="" x="">= 1 pg population.

The factors which cause the elevated response in the low-concentration region may also cause positive bias in low-level samples. Procedural changes, data review guidelines, and changes in the quality assurance specifications are recommended specifically for samples which require accurate low-level analysis, such as compliance testing.