In the first part of the current study, the influence of water quantity and extrusion speed was investigated for a highly
soluble drug product at a fixed concentration in the formula. A design of experiments underlined significant differences between
the three extrusion systems (1). This article will continue to compare radial, dome, and axial single-screw extruders in terms
of process and pellet quality. This article specifically will investigate the influence of formulation and spheronization
variables on process and pellet properties by using a response surface design of experiments, a powerful statistic tool allowing
a rational study of the experimental parameters and enhanced process comprehension.
The influence of formulation and spheronization conditions on product properties has been widely described in the literature.
Numerous authors have shown the impact of excipient and drug substance properties (e.g., solubility, particle size) or content
on the extrusion process and/or pellet quality for one kind of extrusion system (2–16). Other authors have underlined the
effect of spheronization time or speed on pellet properties for one system of extrusion (7, 9, 15–26).
These studies show that formulation and spheronization variables allow improved comparison of extrusion systems by providing
complementary information about their efficiency under different conditions. Other authors have compared the extrusion of
different formula under constant spheronization conditions, the effect of different spheronization conditions on pellet quality
for different extrusion systems for the same formulation, and other combinations of conditions to evaluate the extrusion process
for axial and radial extruders (14, 27–36).
The research here fully compares the dome extruder with other systems using a design-of-experiments approach to enable full
analysis and good process understanding. Three single-screw extrusion systems (radial, dome, and axial) are compared for their
capacity to produce good quality pellets of various drug concentrations and solubility, under different spheronization speeds
and times. The study analyzes the results against each other, introducing the notions of robustness and flexibility.
Materials and methods
Pellets were prepared from a binary mixture of a drug substance (DS) and microcrystalline cellulose (MCC). Two drug substances,
DS1 and DS2 supplied by Pierre Fabre Research Institute, were tested. DS1 corresponded to the antidepressant drug product
studied in Part I of the study, and DS2 corresponded to the monohydrate theophylline. The drugs were chosen for their different
solubility in water (1250 g/L for DS1 and 8 g/L for DS2). MCC (Avicel PH101), supplied by FMC Biopolymer, is insoluble in
water. Three ratios of DS to MCC were tested for both drug substances: 20:80, 36:64, and 52:48 (% w/w). Purified water was
used as liquid binder. The optimal water quantity used for each of the six formulations was determined by preliminary experiments
and was found to be dependant on drug solubility and concentration. The optimal water level decreased with water solubility
and increased concentration of the drug. This effect has been observed before for drugs and excipients (2, 5, 13, 29).
Experimental design and pellet preparation.
Pellets were prepared according to the manufacturing conditions described in Part I of the study (1). A response surface design
of experiments was built with Design Expert software, version 188.8.131.52 (Stat-Ease). The mathematical model targeted for each
response studied was a quadratic model with first-order interactions. Five factors were studied: drug solubility in water
(g/L), drug concentration (%), extrusion system, spheronization speed (rpm), and spheronization time (min). To analyze the
results, the drug solubility and the extruder system were included as qualitative factors, whereas the others were considered
as continuous factors. DS1 and DS2 were both tested at concentrations of 20, 36, and 52%. Spheronization speed was tested
at 800, 1000, and 1200 rpm. Spheronization time was tested at 2, 3, and 4 min. These intervals were determined by preliminary
trials, beyond these limits, it was difficult to obtain acceptable pellets. All other experimental conditions were constant.
The extrusion speed was 40 rpm.
The design of experiments was built as a set of six Box–Behnken designs, each corresponding to one combination of two qualitative
factors (i.e., drug solubility and extrusion system). For each of these six designs, three replicates of the central point
(level 0, i.e., 36% of drug substance, 1000 rpm of spheronization speed, and 3 min of spheronization time) were run (see Figure
1). The whole experimental design included a total of 96 (6 × 16) experiments. Figure 2 summarizes factors and responses selected
for the global design of experiments.
Figure 1: Factor levels of the experimental design for each qualitative combination. (ALL FIGURES ARE COURTESY OF THE AUTHORS)
Responses specific to this part of the study, shown in italics in Figure 2, are described below. All other responses are described
in Part I of the study (1). Pellet dispersion could not be analyzed because the design of experiments was inadequate for the
model for this response.
Figure 2: Factors and responses of experimental design.
Roughness analysis of the pellet yield fraction was assessed by measuring solidity factor using a Morphologi G2 (Malvern Instruments).
Analysis was carried out on around 300 pellets from the usable yield fraction. Solidity factor (S) was calculated according
to the formula: S = A÷(A+B) in which, A is pellet area and A+B is the area enclosed by the convex hull (A+B). High solidity
is desirable because it corresponds to low roughness; rough pellets may generate fines or have poor flow characteristics.
Surface roughness of the pellets is also an important characteristic when considering eventual coating or compression into
Pycnometric density of pellets, D
pycno (g.cm3), was determined using a helium pycnometer (Accupyc 1330, Micromeritics Instrument) Samples were degassed under 6.5 Pa vacuum
(VacPrep 061, Micromeritics Instrument) for two days at about 25 °C. Measurements were performed using a 10 cm3 cell, and repeated until the value stabilized. The mean pycnometric density was calculated from the final three stabilized