A series of tests were carried out in 10L and 60L PharmaConnect® granulators (GEA Process Engineering, Inc.). Two formulations were used. A placebo formulation consisted of microcrystalline cellulose, anhydrous lactose, croscarmellose sodium, and different concentrations (1%, 3%, and 5% w/w) of HPC. 40% w/w water was added. A brivanib formulation consisted of brivanib alaninate, HPC, croscarmellose sodium and microcrystalline. 48% w/w, 58% w/w and 67% w/w water was added. Both formulations were pre-mixed for 30s and then water was added over the period of 3 minutes, followed by 20-25 min wet mass mixing. In all granulators the impeller tip speed was kept constant at 4.8 m/s.
DFF sensors (model P-4000-40) were used for the tests. The sensor was vertically inserted into the granulator bowl through an existing port in the lid and positioned inside the moving powder mass. The tip of the DFF sensor probe was positioned 2.54 cm above the blade at approximately half of the blade length from the center, at 90◦angle to the blade rotation. The second probe was inserted through the side port 3.3 cm and 5.9 cm above the blades for 10L and 60L granulators, respectively. LOI optical interrogator was used to record sensors response at 500 Samples/s rate.

Placebo granulation

A typical DFF sensor response is shown in figure below (blue line). The data were recorded with the high acquisition rate, 500 Samples/s, which allowed for detecting fine features of the powder flow, acting on the sensor. The sensor signal may be interpreted as a continuous periodic force, which can be approximated with a harmonic function (orange curve) overlapped with sudden narrow peaks of variable magnitude. The wet mass, therefore, is modeled as a two-component mixture of particles that interact with the probe as: (1) a continuous liquid-like flow which exerts a periodic force on the pin (force is the highest when the granulator’s blade is passing near the sensor, and lowest when the sensor is between two blades); and (2) rare large wet mass agglomerates or “lumps” > 1 mm in diameter that manifest themselves as narrow peaks.

In the presented figure, the sensor response was baseline corrected: level of the force, observed in the middle point between the blades was adjusted to zero.

The baseline corrected sensor response for three different concentration of HPC is shown below.

Evolution of magnitude of the peaks due to consolidated granule impacts averaged over 100 consecutive blade occurrences is shown below, followed by the time course histograms of peak amplitude distributions (for subsets of 100 peaks)

Figures above demonstrate clear difference between placebo formulations with 1%, 3%, and 5% (w/w) of HPC. For each formulation the peak magnitude distribution is widest during water addition, while maximum peaks are observed soon after the water addition was stopped, with gradual decrease afterwards.

Brivanib Alaninate Granulation

The sensor response was similar to that obtained in placebo granulation: a continuous periodic force overlapped with narrow peaks of variable magnitude. The baseline corrected sensor output for three different concentration of water in 10L granulator and for 58% w/w water concentration in 60L granulator is shown below.

The data indicate consistency of pattern over time for both side sensor and top sensor, even though the amplitude of peaks was higher for the side sensor—attributable to higher blade local linear velocity. Expectedly,the signal magnitude increased with increasing water concentration at the 10-L granulator scale. Similar to the tests with placebo, the DFF signal response over a period of time showed a pattern of increase in wet mass consistency during the phase of water addition and immediately after the end of water addition, and a decline during extended wet massing. In addition, the second maxima is present which is especially pronounced for the tests with 58% w/w water concentration. This second maximum was not present in tests with placebo. Time between the end of water addition and the second maximum is different for different granulators: ~160s for 10L granulator and ~360s for 60L. The time domain differences vanished when the x-axis is plotted as number of blade rotations under the sensor: see, for example, plots with peak magnitudes averaged over 100 blade rotations for the side sensor.

The time delay to the second maximum can be utilized to derive parameters that can assist scale-up of granulations.

You can find more information in our paper published in the Journal of Pharmaceutical Sciences.