Blend Uniformity and Unit Dose Sampling

Publisher Website
Google Scholar
Abstract
Process validation for solid dosage forms must include a component that demonstrates content uniformity of the final blend. As a result of a recent court ruling this aspect of validation has received increased attention from both industry and government. In particular, the court ruled that the appropriate sample size for content uniformity testing of the final blend in validation batches is three times the run weight of the finished product. Furthermore, recent FDA communications suggest that the uniformity of the final blend should be held to a higher standard than that of the tablet in order to provide reasonable assurance that the finished product will exhibit acceptable uniformity. The purpose of this article is to communicate some problems that our firm encountered during the validation of a lower strength of a currently marketed tablet. The validated process for the marketed product has a long history of providing tablets that exhibit acceptable content uniformity at the higher strength. Extensive data from three validation batches manufactured at each of two production sites demonstrate consistently and significantly lower drug content in the lubricated granulation than in the finished product. The variability of the granulation data as characterized by the standard deviation is, for the most part, acceptable and comparable to that of the tablets. Unfortunately, these batches fail validation when subjected to the acceptance criteria suggested in recent FDA communications. We attribute the lower assay values of the granulation to sampling bias that occurs when small volumes of powder are extracted from a large V-blender with a sampling thief. In order to circumvent this problem and avoid rejecting acceptable processes we recommend an alternative acceptance criteria in which the mean assay of the powder samples is not relevant, This criteria relies solely on the standard deviation of the blend.

This publication has 4 references indexed in Scilit: