Questions and answers
The following section addresses anticipated questions about the spectral library project and the collection of samples. Answers
are provided by the authors, representing DPA and IPEC.
What instruments will be used for screening excipients?
DPA plans to build two libraries, one for NIR and one for Raman spectroscopy. These are both common techniques used in library
searches because their spectral features are highly sensitive to chemical structure. Both have become increasingly common
in portable instrumentation due to their ability to sample through various common sampling media. Some excipients may not
be suitable for both libraries because of some inherent limitations of each technique that must be taken into account when
populating spectral libraries. An NIR spectrum is dependent on moisture content and particle size, for example. It contains
broad and generally weak spectral features (due to sampling of overtone and combination bands). Raman features tend to be
sharper and do not depend on moisture or particle size, but if a sample fluoresces, Raman spectral features can be obscured
and limit the effectiveness of the library method. Some of the limitations for these two techniques can be reduced using common
preprocessing techniques, such as multiplicative scatter correction or first derivative preprocessing. The results of library
searches are strongly dependent on (and often enhanced by) spectral preprocessing.
How will a sample spectrum be matched to the libraries?
The typical library search consists of correlating the full spectrum of a test (unknown) sample to a reference (library) spectrum
using the Hit Quality Index (HQI), shown in the following equation:
where the × indicates the scalar product of the two vectors. As such, library searches are aimed at gauging the full spectrum
correlation of two spectra as opposed to focusing on a particular region or peak. The HQI values in the above equation range
from 0 to 1.000, with 1.000 indicating a perfect match. This match is achieved when the unknown spectrum is identical to the
library spectrum. FDA plans to use this concept when it implements its libraries.
How many materials are expected to be in the library?
DPA has identified 166 excipients that fit in one or more of the following categories: most commonly used, most vulnerable
to adulteration, or needing an update of the US Pharmacopeia monograph. From this list, FDA has narrowed down a priority list
of excipients (see Table I). Once library entries for these priority materials have been obtained, the library will be deployed
to field inspectors along with any other excipients that have been received. The library will continue to be expanded as additional
excipients are received.
Table I: List of priority excipients to be collected for spectral library.
What additional types of data are being considered for this library?
Both an NIR and Raman library will be built. These libraries will contain the spectrum of each excipient. In addition to the
spectrum, each entry will contain the chemical name, date added, and spectrometer details. When the library search algorithm
can distinguish between different manufacturers and/or sites of manufacture of the same excipient, such further identifying
information will be included.