Kikuchi Shingo, Takayama Kozo
Department of Pharmaceutics, Hoshi University, 2-4-41 Ebara, Shinagawa-ku, Tokyo, Japan.
Int J Pharm. 2009 Jun 5;374(1-2):5-11. doi: 10.1016/j.ijpharm.2009.02.016. Epub 2009 Mar 6.
A novel method of evaluating the reliability of the optimal formulations of pharmaceutical products was developed based on statistical techniques. Hydrogel ointments, PEGylated emulsions, and solid dispersions were chosen as the model data for the pharmaceutical products, and the formulations of these models were optimized using a nonlinear response surface method incorporating multivariate spline interpolation. A bootstrap resampling method combined with a Kohonen's self-organizing map, was used to estimate the confidence intervals of the optimal formulations. To understand the factors significantly affecting the optimal formulations, a leave-one-factor-out (LOFO) method and a random number technique were introduced as sensitivity analyses. Our results suggest that the random number technique is a better approach than the LOFO method. To determine the design space and control space based on a scientific rationale that can satisfy a number of specifications of the pharmaceutical responses, a novel approach that takes advantage of the random number technique was investigated. The control space was successfully defined as a super-cubic area inscribed in a super-spherical area in the design space of the factors.
基于统计技术开发了一种评估药品最佳配方可靠性的新方法。选择水凝胶软膏、聚乙二醇化乳液和固体分散体作为药品的模型数据，并使用结合多元样条插值的非线性响应面法对这些模型的配方进行优化。采用结合了Kohonen自组织映射的自助重采样方法来估计最佳配方的置信区间。为了解显著影响最佳配方的因素，引入了逐一因素排除（LOFO）法和随机数技术作为敏感性分析。我们的结果表明，随机数技术是比LOFO法更好的方法。为了基于能够满足多种药物响应规格的科学原理确定设计空间和控制空间，研究了一种利用随机数技术的新方法。在因素的设计空间中，控制空间成功地被定义为内切于超球面区域的超立方区域。
