Nested designs in ruggedness testing

Abstract

Nested designs were performed in order to execute a ruggedness test according to the United States Pharmacopeia definition for ruggedness, in which mainly non-procedure related factors are examined. Several nested designs have been executed on a high performance liquid chromatography assay to determine tetracycline and related substances in bulk samples of tetracycline. Factors such as different laboratories, analysts, instruments, columns, days and batches were examined. The interpretation methods described in the literature were found to cause problems. In these methods the variances of the examined factors are estimated from the calculated mean square values and from the equation for the expected mean squares. Very frequently, negative variance estimates were obtained. Their absolute values were found to be dependent on the influence of the factor examined below it in the design, on the examined response. Therefore an alternative interpretation method for nested designs, based on pooled variances, was proposed and found to be appropriate to use for ruggedness testing purposes. Both approaches, the one from the literature and the one proposed here, were tested on simulated data coming from a nested design with four factors and on the experimentally measured data.