Design space construction of multiple dose-strength tablets utilizing bayesian estimation based on one set of design-of-experiments.

Design spaces for multiple dose strengths of tablets were constructed using a Bayesian estimation method with one set of design of experiments (DoE) of only the highest dose-strength tablet. The lubricant blending process for theophylline tablets with dose strengths of 100, 50, and 25 mg is used as a model manufacturing process in order to construct design spaces. The DoE was conducted using various Froude numbers (X(1)) and blending times (X(2)) for theophylline 100-mg tablet. The response surfaces, design space, and their reliability of the compression rate of the powder mixture (Y(1)), tablet hardness (Y(2)), and dissolution rate (Y(3)) of the 100-mg tablet were calculated using multivariate spline interpolation, a bootstrap resampling technique, and self-organizing map clustering. Three experiments under an optimal condition and two experiments under other conditions were performed using 50- and 25-mg tablets, respectively. The response surfaces of the highest-strength tablet were corrected to those of the lower-strength tablets by Bayesian estimation using the manufacturing data of the lower-strength tablets. Experiments under three additional sets of conditions of lower-strength tablets showed that the corrected design space made it possible to predict the quality of lower-strength tablets more precisely than the design space of the highest-strength tablet. This approach is useful for constructing design spaces of tablets with multiple strengths.

Novel method for constructing a large-scale design space in lubrication process by using Bayesian estimation based on the reliability of a scale-up rule.

A reliable large-scale design space was constructed by integrating the reliability of a scale-up rule into the Bayesian estimation without enforcing a large-scale design of experiments (DoE). A small-scale DoE was conducted using various Froude numbers (X(1)) and blending times (X(2)) in the lubricant blending process for theophylline tablets. The response surfaces, design space, and their reliability of the compression rate of the powder mixture (Y(1)), tablet hardness (Y(2)), and dissolution rate (Y(3)) on a small scale were calculated using multivariate spline interpolation, a bootstrap resampling technique, and self-organizing map clustering. A constant Froude number was applied as a scale-up rule. Experiments were conducted at four different small scales with the same Froude number and blending time in order to determine the discrepancies in the response variables between the scales so as to indicate the reliability of the scale-up rule. Three experiments under an optimal condition and two experiments under other conditions were performed on a large scale. The response surfaces on the small scale were corrected to those on the large scale by Bayesian estimation using the large-scale results and the reliability of the scale-up rule. Large-scale experiments performed under three additional sets of conditions showed that the corrected design space was more reliable than the small-scale design space even when there was some discrepancy in the pharmaceutical quality between the manufacturing scales. This approach is useful for setting up a design space in pharmaceutical development when a DoE cannot be performed at a commercial large manufacturing scale.

Latent structure analysis of the process variables and pharmaceutical responses of an orally disintegrating tablet.

A multivariate statistical technique was applied to the design of an orally disintegrating tablet and to clarify the causal correlation among variables of the manufacturing process and pharmaceutical responses. Orally disintegrating tablets (ODTs) composed mainly of mannitol were prepared via the wet-granulation method using crystal transition from the δ to the ß form of mannitol. Process parameters (water amounts (X(1)), kneading time (X(2)), compression force (X(3)), and amounts of magnesium stearate (X(4))) were optimized using a nonlinear response surface method (RSM) incorporating a thin plate spline interpolation (RSM-S). The results of a verification study revealed that the experimental responses, such as tensile strength and disintegration time, coincided well with the predictions. A latent structure analysis of the pharmaceutical formulations of the tablet performed using a Bayesian network led to the clear visualization of a causal connection among variables of the manufacturing process and tablet characteristics. The quantity of ß-mannitol in the granules (Q(ß)) was affected by X(2) and influenced all granule properties. The specific surface area of the granules was affected by X(1) and Q(ß) and had an effect on all tablet characteristics. Moreover, the causal relationships among the variables were clarified by inferring conditional probability distributions. These techniques provide a better understanding of the complicated latent structure among variables of the manufacturing process and tablet characteristics.

Contribution of the physicochemical properties of active pharmaceutical ingredients to tablet properties identified by ensemble artificial neural networks and Kohonen's self-organizing maps.

The aim of this study was to create a tablet database for use in designing tablet formulations. We focused on the contribution of active pharmaceutical ingredients (APIs) to tablet properties such as hardness and disintegration time (DT). Before we investigated the effects of the APIs, we optimized the tablet base formulation (placebo tablet) according to an expanded simplex search. The optimal placebo tablet showed sufficient hardness and rapid disintegration. We then tested 14 kinds of compounds as the model APIs. The APIs were characterized in terms of their physicochemical properties using Kohonen's self-organizing maps. We also prepared model tablets by incorporating the APIs into the optimal placebo tablet, and then examined the tablet properties, including tensile strength and DT. On the basis of the experimental data, an ensemble artificial neural network incorporating general regression analysis was conducted. A reliable model of the correlation between the physicochemical properties of the APIs and the tablet properties was thus constructed. From the correlation model, we clarified the detailed contributions of each physicochemical property to the tablet attributes.

Novel method to construct large-scale design space in lubrication process utilizing Bayesian estimation based on a small-scale design-of-experiment and small sets of large-scale manufacturing data.

A large-scale design space was constructed using a Bayesian estimation method with a small-scale design of experiments (DoE) and small sets of large-scale manufacturing data without enforcing a large-scale DoE. The small-scale DoE was conducted using various Froude numbers (X(1)) and blending times (X(2)) in the lubricant blending process for theophylline tablets. The response surfaces, design space, and their reliability of the compression rate of the powder mixture (Y(1)), tablet hardness (Y(2)), and dissolution rate (Y(3)) on a small scale were calculated using multivariate spline interpolation, a bootstrap resampling technique, and self-organizing map clustering. The constant Froude number was applied as a scale-up rule. Three experiments under an optimal condition and two experiments under other conditions were performed on a large scale. The response surfaces on the small scale were corrected to those on a large scale by Bayesian estimation using the large-scale results. Large-scale experiments under three additional sets of conditions showed that the corrected design space was more reliable than that on the small scale, even if there was some discrepancy in the pharmaceutical quality between the manufacturing scales. This approach is useful for setting up a design space in pharmaceutical development when a DoE cannot be performed at a commercial large manufacturing scale.

Enantiopure cyclic O-substituted phenylphosphonothioic acid: synthesis and chirality-recognition ability.

As a new acidic selector (resolving agent), we synthesized an enantiopure O-alkyl phenylphosphonothioic acid with a seven-membered ring ((R)-5), which was designed on the basis of the results for the enantioseparation of 1-arylethylamine derivatives with acyclic O-ethyl phenylphosphonothioic acid (I). The phosphonothioic acid (R)-5 showed unique chirality-recognition ability in the enantioseparation of 1-naphthylethylamine derivatives, aliphatic secondary amines, and amino alcohols; the ability was complementary to that of I. The X-ray crystallographic analyses of the less- and more-soluble diastereomeric salts showed that hydrogen-bonding networks in the salt crystals are 2(1) -column-type with a single exception which is cluster-type. In the cases of the 2(1) -column-type crystals, stability of the crystals is firstly governed by hydrogen bonds to form a 2(1) -column and secondly determined by intra-columnar T-shaped CH/π interaction(s), intra-columnar hydrogen bond(s), inter-columnar van der Waals interaction and/or inter-columnar T-shaped CH/π interaction(s). In contrast, the cluster-type salt crystal is stabilized by the assistance of inter-cluster T-shaped CH/π and van der Waals interactions. To realize still more numbers of intra- and inter-columnar and -cluster T-shaped CH/π interactions, the seven-membered ring of (R)-5 plays a considerable role.

Factors determining the pattern of a hydrogen-bonding network in the diastereomeric salts of 1-arylethylamines with enantiopure P-chiral acids.

In order to clarify factor(s) determining the pattern of a hydrogen-bonding network in the diastereomeric salts of 1-arylethylamines (1) with enantiopure P-chiral acids, three kinds of enantiopure P-chiral alkylphenylphosphinothioic acids (3) were synthesized, and their chiral recognition abilities for racemic 1 were examined. The characteristics of the diastereomeric salt crystals of 1 with 3 were then studied on the basis of their X-ray crystallographic analyses. Statistical analysis on the molecular conformations observed in the diastereomeric salts with 3 as well as those with P-chiral O-alkyl arylphosphonothioic acids (2), which have a chemical structure similar to that of 3, and molecular orbital calculations for 2 and 3 in a gas phase revealed that the torsion angle between the aromatic plane and the P--O(alkoxy in 2) or P--C(alkyl in 3) plays an important role in determining the pattern of a hydrogen-bonding network in the diastereomeric salts, either a closed globular cluster or an infinite 2(1) column type. The calculations also indicated that there is a hydrogen-bonding-like interaction between the ammonium hydrogen atom of 1-arylethylammonium cations and the P--O(alkoxy) oxygen atom of phosphonothioate anions in the clusters.