Premium
Estimating the predictive validity of diabetic animal models in rosiglitazone studies
Author(s) -
Varga O. E.,
Zsíros N.,
Olsson I. A. S.
Publication year - 2015
Publication title -
obesity reviews
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.845
H-Index - 162
eISSN - 1467-789X
pISSN - 1467-7881
DOI - 10.1111/obr.12278
Subject(s) - rosiglitazone , animal studies , diabetes mellitus , animal model , medicine , predictive validity , consistency (knowledge bases) , human studies , clinical trial , econometrics , computer science , endocrinology , mathematics , clinical psychology , artificial intelligence
Summary For therapeutic studies, predictive validity of animal models – arguably the most important feature of animal models in terms of human relevance – can be calculated retrospectively by obtaining data on treatment efficacy from human and animal trials. Using rosiglitazone as a case study, we aim to determine the predictive validity of animal models of diabetes, by analysing which models perform most similarly to humans during rosiglitazone treatment in terms of changes in standard diabetes diagnosis parameters (glycosylated haemoglobin [ HbA1c ] and fasting glucose levels). A further objective of this paper was to explore the impact of four covariates on the predictive capacity: (i) diabetes induction method; (ii) drug administration route; (iii) sex of animals and (iv) diet during the experiments. Despite the variable consistency of animal species‐based models with the human reference for glucose and HbA1c treatment effects, our results show that glucose and HbA1c treatment effects in rats agreed better with the expected values based on human data than in other species. Induction method was also found to be a substantial factor affecting animal model performance. The study concluded that regular reassessment of animal models can help to identify human relevance of each model and adapt research design for actual research goals.