Managing sampling plans
Another big challenge to using RTRT is how to approach sampling plans, specifically when and where to take a sample, how much
of a sample to take, and whether compendial or risk-assessment issues need to be considered. Explains Alon Vaisman, applications
manager of pharmaceuticals for Malvern Instruments, "Usually PAT instruments access larger amounts of sample than would be
used for lab analysis. On-line instrumentation would typically utilize automated sampling, in contrast to traditional off-line
techniques that tend to rely on grab sampling at the end of the process. Sample preparation is usually relatively limited
for PAT systems since this is necessary to achieve the measurement rates required for continuous monitoring."
As for where to do sampling on themanufacturing line, Real-Time Analyzers President & CEO Stuart Farquharson suggests that
RTRT is not limited to tablet or pill manufacturing, but also encompases drug synthesis. In this case, the best place to monitor
synthesis is in-situ, or inside the reactor. This allows monitoring and ultimately controlling the reaction rate, reaction end-point, and yield.
"This can be accomplished using a long rod fiber-optic probe. However, efforts must be made to keep the probe head clean.
In-situ measurements allow for true real-time monitoring, which is significantly better than traditional grab sampling, primarily
used to determine when the end point had been reached," he explains.
According to FDA's McNally, "Manufacturers must determine appropriate sampling for their processes. It is important to remember
that compendial tests are standards that any compendial drug must meet if tested. Applicants should consider the claims and
disclaimers made by each compendium they reference.
"The current US Pharmacopeia (USP 33–NF 28 Reissue), for example, notes that their compendial standards are not intended to make inferences about the larger group
of units from which the sample was obtained. The General Notices section also states, 'In all cases, statements about whether
the compendial standard is met apply only to the units tested. Repeats, replicates, statistical rejection of outliers, or
extrapolations of results to larger populations, as well as the necessity and appropriate frequency of batch testing, are
neither specified nor proscribed by the compendia.' So, manufacturers must ensure sampling sizes and plans are statistically
sound and representative of the batch. A quality risk-management approach can be useful to develop an appropriate sampling
Moore adds that the agency is willing to work with applicants to ensure appropriate acceptance criteria for large sample sizes.
Pfizer's Hammond seems to agree that there is work to be done in the area of sampling. "There is the potential to sample very
frequently or to take large numbers of units for a measurement," he explains. Pfizer uses a risk-based approach to determine
sampling frequency, including looking at how the data is used, what calculations are performed, and what statistics are used
for setting a specification.
"There is agreement that conventional statistics do not apply when larger data sets are under consideration. The so-called
'large n' approach is needed. Industry and regulators have been debating the best approach for a while now and some sensible
approaches seem to be evolving from this discussion," adds Hammond.
He references a recent PQRI paper published, with FDA input, that describes three possible "large n" approaches. "In terms
of sampling of blends during analysis using probe systems, careful control of the amount of sample that contributes to a spectrum
is important. Generally the illumination characteristics are engineered to ensure that approximately one unit dose weight
of blend contributes to a spectrum. Analysis of tablet cores is generally via transmission of light, through the tablet, thus ensuring that most of the material in the tablet core contributes to the
Overall, the primary purpose of RTRT, and any PAT or QbD application for that matter, is to understand and control one's product
and processes in a manner that ensures quality final product.
Sampling needs to be sufficient to facilitate real-time control, generate knowledge to allow the control of the next manufacturing
step, and measure critical material attributes. With end-of-line testing, according to one expert, the purpose of sampling
is to assess what the quality is, and the testing of samples has two roles: one enables control, and another validates the
control system and strategy. Within RTRR, therefore, sampling's role is to assure that the controls are appropriate.