Sampling Bias in Blending Validation and a Different Approach to Homogeneity Assessment

Abstract

Sampling of batches studied for validation is reported. A thief particularly suited for granules, rather than cohesive powders, was used in the study. It is shown, as has been demonstrated in the past, that traditional 1x to 3x thief sampling of a blend is biased, and that the bias decreases as the sample size increases. It is shown that taking 50 samples of tablets after blending and testing this subpopulation for normality is a discriminating manner of testing for homogeneity. As a criterion, it is better than sampling at mixer or drum stage would be even if an unbiased sampling device were available.

Characterisation and surface-profiling techniques for composite particles produced by dry powder coating in pharmaceutical drug delivery

“…Although the best sampling technique is to test the final dosage form for content uniformity, in process testing provides the formulator with the option of identifying any intervention if needed [36,37]. Sampling locations and the number of samples to be collected are the two main variables in sampling procedure.…”

“…Effective sampling requires collection of adequate number of representative samples from the entire mix taking into consideration areas of poor mixing (identified through prior evaluation of blending device) whilst obtaining enough samples to create accurate and reproducible results without the need for over-sampling. Although, greater number of samples usually led to more accurate results, the objective of any sampling scheme is to gather reliable results with the fewest number of samples [36,[38][39][40].…”

Adequacy and verifiability of pharmaceutical mixtures and dose units by variographic analysis (Theory of Sampling) – A call for a regulatory paradigm shift

“…The situation is characterized by two draft guidances, one of which one has been withdrawn, and the second never approved(U.S. Human Services, 2002, 2015). This can only be characterized as a colossal negative outcome after trying so long to formulate a guidance, or to build a consensus for a problem that has attracted intense attention (Berman and Planchard, 1995;Boehm, 2001;Chang et al, 1996;El-Hagrasy et al, 2006;Garcia et al, 1998;Hwang et al, 1998;Kraemer et al, 1999;Muzzio et al, 2003;Muzzio et al, 1997;Muzzio et al, 1999;Prescott and Garcia, 2001;Romañach, 2015;Wu et al, 2000).The debate has encompassed working groups composed of regulators, statisticians, industrial scientists and academicians (Bergum et al, 2015;Bergum et al, 2014;Boehm et al, 2003;Boehm et al, 2004;Garcia et al, 2015).…”