18 Near-Infrared Spectroscopy in the Pharmaceutical Industry
405
the sensor’s window, resulting in the need for data collection at large scale which
can be expensive and time consuming.
Continuous blending
In situations where the blend is manufactured continuously, approaches have been
developed to monitor the homogeneity as a function of time. However, a few additional considerations need to be taken into account when discussing continuous
blending. First, the scale of scrutiny (or the amount of sample analyzed at a given
time) becomes more important. In batch blend monitoring, the same powder will
be analyzed rotation after rotation until the entirety of the powder becomes homogeneous and stable. In continuous blending, the powder may never be re-analyzed
and so the sampling must be representative of the mass of powder that patients will
take. Thus, spectra representing at least one-unit dosage form should be collected so
a relevant determination of homogeneity can be achieved. A discussion of scale of
scrutiny and spectral collection geometry on model performance showed the interdependence between process design and sampling representativity [51]. Probe fouling
can also be a significant issue in continuous blending. While in batch, it would be as
simple as stopping the bin, wiping the window, and restarting the batch if coating on
the window is detected, it is not possible to do that often in continuous blending as
manufacturing time dictates the quantity produced. When sufficient shear is present,
and the window self-cleans, little to no control may be necessary. This may be the
case in a tablet press feed-frame for instance. But where the shear is low or the
material sticky, it may be necessary to have an active control that will wipe or clean
the probe at set frequencies. Commercial solutions are available to perform these
frequent probe cleanings.
The monitoring of blend homogeneity in a continuous process has been reported at
the exit of a blender or in the feed-frame of a tablet press. Similarly to batch blending,
continuous blend monitoring can utilize qualitative or quantitative approaches,
depending on the intended purpose of the method. A qualitative approach based
on the F-test was proposed by Fonteyne et al. using the same principles of the block
F-test discussed above for the monitoring of blends [52]. A number of articles discuss
the prediction of blend uniformity for continuous systems. A triangle interface was
used by Vargas et al. to monitor the homogeneity of powder at the discharge of a
blender [53]. Quantitative models were developed for the active ingredient after a
spectral quality evaluation ensured that the powder bed was representative and did not
contain air pockets potentially formed as the blend travels from the blender to downstream unit operations. Other articles report the monitoring of powder uniformity in
feed-tube to a tablet press and other powder interfaces [54].
A significant amount of work has been performed on the analysis of powder
in the feed-frame. Initially presented by Liu and Blackwood [55], an example of
method development and validation was published by De Leersnyder [11]. Authors
investigated the relationship between press parameters (turret speed, paddle speed
405
the sensor’s window, resulting in the need for data collection at large scale which
can be expensive and time consuming.
Continuous blending
In situations where the blend is manufactured continuously, approaches have been
developed to monitor the homogeneity as a function of time. However, a few additional considerations need to be taken into account when discussing continuous
blending. First, the scale of scrutiny (or the amount of sample analyzed at a given
time) becomes more important. In batch blend monitoring, the same powder will
be analyzed rotation after rotation until the entirety of the powder becomes homogeneous and stable. In continuous blending, the powder may never be re-analyzed
and so the sampling must be representative of the mass of powder that patients will
take. Thus, spectra representing at least one-unit dosage form should be collected so
a relevant determination of homogeneity can be achieved. A discussion of scale of
scrutiny and spectral collection geometry on model performance showed the interdependence between process design and sampling representativity [51]. Probe fouling
can also be a significant issue in continuous blending. While in batch, it would be as
simple as stopping the bin, wiping the window, and restarting the batch if coating on
the window is detected, it is not possible to do that often in continuous blending as
manufacturing time dictates the quantity produced. When sufficient shear is present,
and the window self-cleans, little to no control may be necessary. This may be the
case in a tablet press feed-frame for instance. But where the shear is low or the
material sticky, it may be necessary to have an active control that will wipe or clean
the probe at set frequencies. Commercial solutions are available to perform these
frequent probe cleanings.
The monitoring of blend homogeneity in a continuous process has been reported at
the exit of a blender or in the feed-frame of a tablet press. Similarly to batch blending,
continuous blend monitoring can utilize qualitative or quantitative approaches,
depending on the intended purpose of the method. A qualitative approach based
on the F-test was proposed by Fonteyne et al. using the same principles of the block
F-test discussed above for the monitoring of blends [52]. A number of articles discuss
the prediction of blend uniformity for continuous systems. A triangle interface was
used by Vargas et al. to monitor the homogeneity of powder at the discharge of a
blender [53]. Quantitative models were developed for the active ingredient after a
spectral quality evaluation ensured that the powder bed was representative and did not
contain air pockets potentially formed as the blend travels from the blender to downstream unit operations. Other articles report the monitoring of powder uniformity in
feed-tube to a tablet press and other powder interfaces [54].
A significant amount of work has been performed on the analysis of powder
in the feed-frame. Initially presented by Liu and Blackwood [55], an example of
method development and validation was published by De Leersnyder [11]. Authors
investigated the relationship between press parameters (turret speed, paddle speed