(274a) How Raw Material Attributes and Process Parameters Affect the Global Residence Time Distribution Behavior in Continuous Direct Compression

Title: How Raw Material
Attributes and Process Parameters Affect the Global Residence Time Distribution
Behavior in Continuous Direct Compression

Author: Samantha Hurley

The pursuit of drug product
continuous manufacturing will revolutionize the pharmaceutical industry by improving
product flow in our supply chains, such as leveraging fast cycle times and
flexible batch sizes, and by reducing lead times and inventory levels. A
successful batch to continuous manufacturing conversion requires a very robust
control scheme. One commonly described control scheme involves the use of NIR
spectroscopy for blend composition monitoring. This study evaluates the use of
an alternate method, a Residence Time Distribution (RTD) based process model
for control of the continuous direct compression (CDC) process without direct
measurement of blend uniformity. In some cases, this method is preferred over
traditional spectroscopic techniques as the primary control across a wide range
of throughputs, blender speeds, and material properties.  

The CDC system studied consists
of two continuous convection blenders in series followed by a tablet press.
Each of the blenders and the tablet press are modeled as a Plug Flow Reactor
(PFR) in series with two Continuously Stirred Tank Reactors (CSTRs). Since each
PFR and CSTR is characterized by a single time constant, t_Lag and τ
respectively, each blending unit can be described by threeparameters. Tracer
experiments were conducted by spiking API into either the inlet of blender 1 or
blender 2. The effluent API concentration was measured either by blend NIR spectroscopy
at the exit of blender 2 or tablet HPLC at the exit of the tablet press. The
RTD model parameters were optimized by using a least squares algorithm between
the blend NIR/tablet HPLC and the RTD model, previously described. The model
parameters were developed across various processing conditions and material
properties (e.g., bulk densities and flow indices).

The results from the tracer
experiments performed show that a ratio
of τ:t_Lag is constant within each
blending unit across all the various conditions. This universal behavior thus
allows the development of a simple empirical relationship which describes RTD
model parameters across the entirety of the design space.  The results indicate
that the relationship between the blender 1 and blender 2 model parameters
versus throughput (TP) and blender speed (BS) can be described as

where A, B, C, and D are constants. The results also indicate the
relationship between the tablet press model parameters versus throughput can be
described as

where A, B, and C are constants. Knowledge of this universal
behavior allows interpolation of model parameters anywhere within the design
space, enables simpler model validation, and significantly simplifies process
model development for future products.