LC phases improve, but not all assays do: Metformin bioanalysis revisited

Summary

Two assay methods for the antidiabetic metformin, one developed and validated in 1990 and one developed and validated in 1996, are compared. The first method, using an octadecyl phase and an ion pairing agent in the eluent, could not be reproduced some five years later, but another method, using a phenyl phase and no ion pairing agent, could be successfully applied. This paper shows that the retention mechanism of the positively charged analyte is not due to ion-pair formation, as originally assumed, but to interaction with free silanol groups in the LC phase. It is suggested that the number of free silanol groups in octadecyl phases was strongly reduced between 1990 and 1996, whereas for phenyl phases this was not the case. For the second method, validation results on linearity, specificity, accuracy, precision, recovery and stability as well as application of the method to samples from a clinical trial are shown. The validated calibration range is from 20.0 to 2000 ng.mL⁻¹, with accuracy (bias) and precision (coefficient of variation) being below 15% at all levels. Using automated solid-phase extraction for sample preparation, a sample throughput of typically 100 per day can be achieved.