The principle of pooled calibrations and outlier retainment elucidates optimum performance of ion chromatography

A principle with quality assurance of ion chromatography (IC) is presented. Since the majority of scientists and costumers are interested in the determination of the true amount of analyte in real samples, the focus of attention should be directed towards the concept of accuracy rather than focussing on precision. By exploiting the principle of pooled calibrations and retainment of all outliers it was possible to obtain full correspondence between calibration uncertainty and repetition uncertainty, which for the first time evidences statistical control in experiments with ion chromatography. Anions of bromide were analysed and the results were subjected to quality assurance (QA). It was found that the limit of quantification (LOQ) was significantly underestimated by up to a factor of 30 with respect to the determination of concentration of unknowns. The concept of lower-limit of analysis (LLA) and upper-limit of analysis (ULA) were found to provide more acceptable limits with respect to reliable analysis with a limited number of repetitions. An excellent correspondence was found between calibration uncertainty and repetition uncertainty. These findings comply with earlier investigations of method validations where it was found that the principle of pooled calibrations provides a more realistic picture of the analytical performance with the drawback, however, that generally higher levels of uncertainties should be accepted, as compared to contemporary literature values. The implications to the science analytical chemistry in general and to method validations in particular are discussed.

► Outliers of IC should be retained in order to secure correct level of uncertainty.
► Pooled calibrations provide excellent correspondence between uncertainties.
► A new method for the determination of the full-calibration range in IC is proposed.
► The contents determined by IC do not correlate with the sensitivity.
► RSDs of unknowns correspond well to the uncertainty predicted by Horwitz.