Andrés Olivares-Morales (1), Leon Aarons (1) and Amin Rostami-Hodjegan (1, 2)
(1) Centre for Applied Pharmacokinetic Research, School of Pharmacy and Pharmaceutical Sciences, University of Manchester, Manchester, UK. (2) Simcyp Limited, Blades Enterprise Centre, Sheffield, UK.
Objectives: The aim of this study was to develop a Receiver Operating Characteristic (ROC) analysis to evaluate the performance of animal oral bioavailability (Foral) data as a predictor of human Foral and to identify the optimum cut off values of Foral for the implementation of a classification model.
Methods: Foral data for both human and preclinical species – mouse, rat, dog and non-human primates (NHP) – for around 180 compounds was collated from literature as described elsewhere [1]. For implementation of the ROC analysis, human Foral was defined as high (≥ 50%, positive) or low (< 50%, negative). The construction of the ROC curve was implemented in Matlab 2012a by varying the animal threshold (tA) for high and low Foral, the resulting specificity and sensitivity for any tA was recorded and plotted. The evaluation of animal models for the prediction of high and low human Foral was determined by the area under the ROC curve (AUC)[2-4]. In addition, the optimal operating points for the animal models were calculated by cost analysis assuming identical cost for false positive (FP) and false negatives (FN)[2, 4].
Results: Employing Foral data for all the species combined, AUC for the animal model was 0.79 whereas the specific values by species were 0.82, 0.73, 0.80 and 0.96 for mouse, rat, dog and NHP, respectively. The optimal operating point for animal Foral was calculated as 46.5%, for all the species combined, whereas for the particular species, values were around 67%, 22%, 58% and 35%, respectively. The results suggest that animal models can be employed for categorical prediction of human Foral. NHP showed the highest AUC value which is consistent with previous results [1, 5, 6]. The results suggest that a value around 50% for animal Foral could predict high and low human Foral with a high sensitivity and moderate specificity. Species specific results suggest a similar approach, consistent with the values reported previously [1, 5-8].
Conclusions: ROC analysis is a powerful tool for the evaluation of the performance of animal Foral as predictor of human Foral. High and low human Foral could de predicted with an acceptable level of confidence from animal Foral. The resulting cut off values could be employed, together with other analysis, as a tool in the decision making process during development within the pharmaceutical industry.
References:
[1] Musther H, Olivares-Morales A, Hatley O, Liu B, Aarons L, Rostami-Hodjegan A. Animal Versus Human Oral Drug Bioavailability: Do They Correlate? European Journal of Pharmaceutical Sciences (In preparation).
[2] Metz CE. Basic principles of ROC analysis. Seminars in nuclear medicine 1978; 8: 283-98.
[3] Brown CD, Davis HT. Receiver operating characteristics curves and related decision measures: A tutorial. Chemometrics and Intelligent Laboratory Systems 2006; 80: 24-38.
[4] Fawcett T. An introduction to ROC analysis. Pattern Recognition Letters 2006; 27: 861-74.
[5] Akabane T, Tabata K, Kadono K, Sakuda S, Terashita S, Teramura T. A Comparison of Pharmacokinetics between Humans and Monkeys. Drug Metabolism and Disposition 2010; 38: 308-16.
[6] Chiou WL, Buehler PW. Comparison of Oral Absorption and Bioavailability of Drugs Between Monkey and Human. Pharmaceutical Research 2002; 19: 868-74.
[7] Caldwell GW, Ritchie DM, Masucci JA, Hageman W, Yan Z. The new pre-preclinical paradigm: compound optimization in early and late phase drug discovery. Curr Top Med Chem 2001; 1: 353-66.
[8] Thomas VH, Bhattachar S, Hitchingham L, Zocharski P, Naath M, Surendran N, Stoner CL, El-Kattan A. The road map to oral bioavailability: an industrial perspective. Expert Opin Drug Metab Toxicol 2006; 2: 591-608.
Reference: PAGE 22 (2013) Abstr 2744 [www.page-meeting.org/?abstract=2744]
Poster: Absorption and Physiology-Based PK