Residence time distribution as a traceability method for lot changes in a pharmaceutical continuous manufacturing system.

Residence time distribution (RTD) models were developed to track raw material lots and investigate batch transitions in a continuous manufacturing system. Two raw materials with similar physical properties (granular metformin and lactose) were identified via Principal Component Analysis (PCA) from a library of bulk material properties and used to simulate the switching of lots during production. In-line near-infrared (NIR) spectra were collected with the powder flowing through a chute in a continuous manufacturing system to monitor metformin and lactose concentration in step-change experiments with Partial Least Squares (PLS) models. RTD provided an understanding of raw material propagation through the continuous manufacturing system. Transition times between raw material changes were identified using the results of two multivariate approaches PLS and PCA. The methodology was implemented to discriminate the transition zone in a raw material change, contributing to design control strategies for acceptance and diverting mechanisms.

An innovative sampling interface for monitoring flowing pharmaceutical powder mixtures.

A chute was designed following the principles of the Theory of Sampling to minimize the variations in powder flow and provide all particles in the flowing blends with the same opportunity of being selected as a sample. The design also reduces the thickness of the chute to allow the analysis of a higher portion of the flowing blends by a near infrared spectrometer. The blends that flowed through the chute had Carr's index values that fluctuated between 23 and 25 percent, indicating passable flowability. A powder fowling evaluation demonstrated that there was no powder accumulation at the inspection window of the chute. The mass flow rate profiles indicated that the system achieves mass steady-state in approximately 30 s and a throughput of 30 kg/h which makes it suitable for continuous manufacturing operations. An in-line NIR calibration model was developed to quantify caffeine concentrations between 1.51 and 4.52 % w/w. The spectra obtained from each experiment had minimal baseline variation. The developed NIR method was robust to throughput changes up to approximately ±7 %. The test blends in the caffeine concentration range between 2.02 % w/w and 4.02 % w/w met the dose uniformity requirements of the Ph.Eur. 9.0, chapter 2.9.47. Variographic analysis was done to estimate the analytical and sampling errors which yielded values below 0.01 (%w/w)2. The obtained results showed that this chute could also be used in a continuous manufacturing line or other applications with flowing powders.

In-line monitoring of low drug concentration of flowing powders in a new sampler device.

A novel sampler device for flowing powders was tested to quantify drug concentrations as low as 0.76% w/w in pharmaceutical powder blends. The sampler device was developed based on the powder flow behavior within a tablet press feed frame, following the principles laid down in the Theory of Sampling. Two Near-Infrared (NIR) spectroscopic calibration models were developed with powder blends that varied from 0.52 to 2.52% w/w and 1.51-4.52% w/w. The calibration models were able to determine caffeine concentration in test set blends with root mean square error of predictions and bias below 0.1% w/w. Samples were collected from the sampler device and analyzed by ultraviolet-visible (UV-Vis) to determine the caffeine concentration. A high agreement between the in-line NIR predictions and the sampled UV-Vis results was found. The paddle wheel speed in the sampler can be varied up to ±10% without affecting NIR predictions; however, the models did not respond adequately to a 25% increase in this speed. Variographic analysis showed that the sampler device may quantify low drug concentrations with nugget effects below 0.0050 (%w/w)2. This study demonstrate that the sampler device may handle throughputs up to 45 kg/h, without significantly affecting the physical properties of powder blends.

A sampling system for flowing powders based on the theory of sampling.

An innovative chute and stream sampler system for flowing powders has been developed and tested. The system is designed for representative sampling based on the principles of the Theory of Sampling (TOS). The sampling system was used in combination with near infrared (NIR) spectroscopy to determine the drug concentration of flowing powders. The system is comprised of three parts: a chute, a stream sampler and a sample collection port. The NIR spectra were obtained at the chute, before entering the sampler, and as the powder flowed through the stream sampler. Samples were also collected from the sample collection port to be analyzed using an ultraviolet-visible (UV-Vis) reference method to determine drug content. A total of eight pharmaceutical powder blends, ranging in concentration from 10.5(%w/w) to 19.5(%w/w) of caffeine, were used to test the sampling system. Materials were characterized before blends were made to provide information on flow properties. The throughput of the system was between 30 and 35 kg/h based on the flow properties of the blend. Drug concentration was effectively determined at the chute and stream sampler. The NIR calibration models showed low root mean squared errors of prediction, 0.65(%w/w) and 0.51(%w/w), for the chute and stream sampler respectively. The NIR calibration models also showed low bias values -0.36(%w/w) at the chute and 0.057(%w/w) at the stream sampler. Significant agreement was obtained between the results from the nondestructive NIR versus the destructive UV-Vis method. Variographic analysis was performed to estimate the analytical and sampling errors when determining the drug concentration at the chute and stream sampler respectively. The variographic analysis showed low analytical errors, 0.103(%w/w)2 and 0.181(%w/w)2 at the chute and stream sampler respectively. The analysis also showed that the minimum practical error (MPE) was around 0.2(%w/w)2 at both chute and stream sampler.