Can animal data predict human outcome? Problems and pitfalls of translational animal research

studies of drugcandidates have always been performed in animals likerodents or dogs, efficacy and adverse effect studies usinganimals as a model for humans were prompted only in thelate 1950s, as a consequence of the thalidomide scandal,where thousands of babies were born with severe extremitydeformations.In the last 10–15 years, preclinical research methodolo-gies have increased crucially regarding measurement ofsensitivity and specificity. Adapted and improved for dedi-cated use in animal research, positron emission tomography(PET) and single photon emission computed tomography(SPECT) are good examples of such methodologies. Itseems to us that increased measurement precision in animalresearch has somewhat raised expectations regarding humanoutcome prediction of preclinical data.In recent years, on the other hand, there has been increas-ing scepticism about the essentiality of animal models formedical progress [1–7]. Statements like “virtually everyachievement of the last century has depended directly orindirectly on the research with animals” are often found inthe literature to emphasize the importance and necessity ofanimal models used in drug development and medical sci-ence. Robert Matthews [6] discussed the validity of thisparticular statement in a critical article in 2008 and con-cluded that this statement is anecdotal and does not gener-ally hold true. He is convinced, though, and surely there isevidence that “animal models can and have provided manycrucial insights that have led to major advances in medicineand surgery”. Hence, he claims that systematic investiga-tions on the use of animal models and on the evidence thatthey possibly can provide are necessary.Gill Langley, in her critical paper in 2009, referred to thesame statement as Matthews a year earlier [7]. Langleyconcluded that relying on animal surrogates of human ill-nesses is a flawed approach in science. Her own investiga-tions, as well as several published systematic reviews of thereliability of animal models, have shown that fewer than50 % of animal studies have predicted human outcomessufficiently. In certain fields of research, e.g. vaccinationdevelopment against acquired immunodeficiency syndrome(AIDS), prediction failure of chimpanzee and macaquemodels is 100 % [7, 8].We are convinced that animal models can be useful toolsin biomedical research, but undoubtedly, it has frequentlybeen observed that effects found in animal models cannot betranslated to the clinic.Translation problemsSena and colleagues reported only three interventions foracute ischaemic stroke treatment with a convincing effect inpatients out of 494 interventions that showed a positiveeffect in animal models [9]. They have tried to identifyreasons for this tremendous discrepancy between preclinicalanimal studies and patient studies in the clinic. The first