operators are not analytical chemists.
Weighing kilo-sized portions is far simpler
than weighing gram-sized portions. The
chance of sub or super-potent final products is enhanced by the far smaller mass
of API needed to be weight and carefully
transferred to the mixing apparatus.
The precise amount needed for the
product is added with mechanical accuracy in a CM process, eliminating most
errors in high or low doses. The fact that
the API is in an enclosed vessel also
protects operators, aside from the initial loading of the bin, from exposure to
the drug. Two potential problems, solved
with one device.
Content uniformity can be a more difficult target to“hit” with miniscule dosage
levels, as well. Physically, the number of
particles of excipient so greatly outnumber
the particles of API that any minor heterogeneity is amplified. The “good enough”
approach for larger doses (10-50% w/w)
is not good enough for a case where the
number of particles of drug may almost be
counted in the tablet/capsule. Just a small
amount of API, either high or low, could
lead to super- or sub-potent doses. Again,
traditional methods of testing—stopping
the blender, using a sample thief, sending
samples to QC, waiting, then deciding to
pass or fail—may not be sufficient for low
dose products. More samples may need to
be taken and analyzed to assure homogeneity, although, as a physical chemist, I
know powders cannot truly be homogeneous, just well-blended.
This is a case where the current in-process (PAT) “standby,” near-infrared
spectroscopy (NRS), may not suffice.
The acknowledged lower level of NIRS
is roughly 1% API. That would leave LIF
(Light-Induced Fluorescence) as the technology of choice. Briefly, LIF (where the
“L” was originally for LASER) is a very
simple concept: a bright beam of light, a
laser-diode or SLED (super-luminescent
light-emitting diode), is shone on the
powder mix, causing the complex organics (the API, in this case) to fluoresce. As
the blending proceeds, the level of fluorescence is measured and when the levels
are constant, the mix is declared complete.
Since fluorescence is a very sensitive
technique, it has been routinely used in
lieu of NIR for years when the API level
approached 1% or below. While many of
the units are non-specific, very little of
the blend, aside from the API, fluoresce.
By contrast, NIR and Raman can see all
organics and some multivariate equation
is needed to see only the API during the
blending. This makes the calculation of
the end-point of the blend rather simple:
without fancy, complicated and potential-
ly time-consuming to validate math, the
signal merely needs to level out, within
predetermined limits, and the blending is
complete. Of course, as with any step in a
GMP-compliant process, this will need to
be validated with compendial technology
Cleaning the production machinery is
never fun, simple, or obvious. The operator in charge of cleaning needs to be sure
that all the water-soluble and water-insol-uble components are removed from the
surfaces of the equipment. The process
usually involves multiple steps, involving
solvents, detergents, and purified water.
Then, someone needs to swab the surface
with appropriate solvents, attempting to
determine that there is little to no material remaining. The spec for cleanliness is
some level of dose, perhaps less than one
unit dose, over a set area (often 10 cm2),
determined buy some analytical method.
While merely time-consuming for larger
doses, the aforementioned very small dose
levels make the process quite difficult.
Two alternatives may be employed to
make the low dose cleaning faster and
more effective. First, for a batch process—
either classic GMP or PAT controlled—
the manual swabbing of the equipment
could be followed by use of an Ion Mobility (IM) unit. This device, a later version of the ones found at airports, works
by vaporizing the materials removed
from the surface of the equipment, giving
them an electrical charge, and running
them through a column to a detector. The
principle is much like a time-of-flight
mass spectrometer, but does not require
a vacuum or fancy software.
The components of the wash solution acquire a single charge and travel
along the tube, arriving by increasing
molecular weight. The recorded signal resembles a chromatograph, where
the arrival time gives the molecular
mass and the signal strength gives the
amount present. The entire analysis
takes roughly 20 milliseconds.
In addition to being faster and possibly
more accurate than the standard “sample
and sent to the lab” approach, the operator
knows, within seconds, whether he/she
needs to perform further cleaning or that
the unit can be placed back into service
immediately. This step, alone, increases
the availability of process equipment, cutting back on the need for more hardware
The second way to assure cleanliness
is achieved in a continuous manufacturing set-up. After the product has been
completed, a common excipient—salt,
lactose, or microcrystalline cellulose—is
run through the system until clean as per
validation. The subsequent product blend
excipient blend may then be used to clean
out the cleaning powder, prior to initiating
the actual product blend.
This benefit also obviates the need for
the operator who does the cleaning to take
precautions, such as mask, gloves, and
even full covering for highly toxic drugs.
In addition, this does away with the dismantling of the CM unit and the washing
step, again, allowing the unit to be used in
short order, cutting the needed inventory
of process lines.
As for the last difficulty, morphology,
there is one possible solution for measuring in real time: time-gated Raman.
Overall both Raman and NIR are very
good for determining polymorphic forms
of API and excipients, such as sugars.
However, we have already established
that the level of API makes NIR problematic and, at lower levels, normal Raman’s fluorescence background makes it
a poor choice for analysis. In a previous
column, I showcased a product where the
LASER that strikes the sample is pulsed
nanosecond bursts and the emitted light
is samples in picosecond time frames,
allowing the Raman spectrum to be enhanced. This would be a good tool to try
for determining the polymorphic state
of the API, especially is it needed to be
amorphous for solubility’s sake.
So, it appears that a number of answers
to problems in analysis or procedure not
only make handling and analysis possible,
but they speed up and help with the quality of the product, itself. And, for an added
bonus, they free up the process equipment
more rapidly for subsequent batches. Win-win-win, I would say. CP