Statistical Analysis of a Method to Predict Drug–Polymer Miscibility

• Published in: Journal of pharmaceutical sciences Vol. 105; no. 1; pp. 362 - 367
• Main Authors: Knopp, Matthias Manne, Olesen, Niels Erik, Huang, Yanbin, Holm, René, Rades, Thomas
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.01.2016
• Subjects: Algorithms; amorphous; amorphous solid dispersion; Calorimetry, Differential Scanning; Chemistry, Pharmaceutical; crystallization; Felodipine - chemistry; kinetics; Linear Models; Models, Theoretical; Pharmaceutical Preparations - chemistry; physical stability; polymers; Polymers - chemistry; Predictive Value of Tests; Solubility; Temperature; thermal analysis; Thermodynamics; amorphous; thermodynamics; crystallization; polymers; physical stability; kinetics; amorphous solid dispersion; thermal analysis
• ISSN: 0022-3549; 1520-6017; 1520-6017
• DOI: 10.1002/jps.24704

In this study, a method proposed to predict drug–polymer miscibility from differential scanning calorimetry measurements was subjected to statistical analysis. The method is relatively fast and inexpensive and has gained popularity as a result of the increasing interest in the formulation of drugs as amorphous solid dispersions. However, it does not include a standard statistical assessment of the experimental uncertainty by means of a confidence interval. In addition, it applies a routine mathematical operation known as “transformation to linearity,” which previously has been shown to be subject to a substantial bias. The statistical analysis performed in this present study revealed that the mathematical procedure associated with the method is not only biased, but also too uncertain to predict drug–polymer miscibility at room temperature. Consequently, the statistical inference based on the mathematical procedure is problematic and may foster uncritical and misguiding interpretations. From a statistical perspective, the drug–polymer miscibility prediction should instead be examined by deriving an objective function, which results in the unbiased, minimum variance properties of the least-square estimator as provided in this study.