Why animal studies are often poor predictors of human reactions to exposure.

摘要: The concept that animal research, particularly relating to pharmaceuticals and environmental agents, may be a poor predictor of human experience is not new. A thousand years ago, Ibn Sina commented on the need study humans rather than animals,1 Alexander Pope's dictum ‘The proper mankind man’ well known has been widely cited.2 Pharmacologists, in particular, have long recognized difficulties inherent extrapolating drug data from animals man.3,4 Given large number studies conducted, it would expected some experiments do predict reactions. For example, penicillin was observed protect both mice staphylococcal infections,5 isotretinoin (‘Acutane’) causes birth defects rabbits monkeys as (although or rats).6 By contrast, corticosteroids are teratogenic but humans,7 thalidomide teratogen many species humans.8 Recent phase 1 monoclonal antibody TGN 1412 resulted life-threatening morbidity all six healthy volunteers, reflecting inadequate prediction even non-human primates response.9 One reason why often translate into replications trials10,11 cancer chemoprevention12–14 poorly designed, conducted analysed. Another possible contribution failure replicate results research reviews summaries evidence methodologically inadequate.15 In one survey, only 1/10,000 Medline records were tagged being meta-analyses compared with 1/1000 studies.16 recent reports, quality synthesis documented by comprehensive search which found 25 systematic research.17 Other similarly 3015 5718 any type research. These deficiencies important because provides rationale for hypotheses studied epidemiologists clinical researchers. The paper Perel his colleagues19 added James Lind Library made an methodological understanding authors relevant areas where confident estimates intervention effects (benefit harm) demonstrated randomized trials. interventions were: head injury; antifibrinolytics reduce bleeding; tissue plasminogen activator death disability after stroke; tirilazad ischaemic antenatal lung preterm newborns; bisphosphonates increase bone mineral density. three trials substantially discordant; others essentially similar. however, major limitations widespread publication bias identified. Systematic review most advanced field stroke research,20,21 area almost no new therapies developed despite decades FK506, 29 found, had blinded investigators two observers during outcome assessment. None met 10 criteria applied reviewers (one criteria; highest score 7). Meta-analysis strong trend weakest show strongest protective effects, either weak effects.22 The few literature done also pointed other difficulty humans,23 concern increasingly fields development evaluation.24,25 Mathews26 recently challenged those who claimed ‘virtually every medical achievement last century depended directly indirectly animals’ provide justifying their assertion. Some key problems summarized Pound her colleagues17: Disparate strains, variety metabolic pathways metabolites, leading variation efficacy toxicity; Different models inducing illness injury, varying similarity condition; Variations dosing schedules regimens uncertain relevance condition; Variability study, methods randomization, choice comparison therapy (none, placebo, vehicle); Small experimental groups statistical power; simple analyses account confounding; follow intention-to-treat principles; Nuances laboratory technique influence results, blinding investigators, neither nor reported; Selection measures, surrogates precursors disease, condition; Variable duration follow-up, correspond disease latency humans. As referred this commentary shown, will become more valid predictors reactions exposures treatments if there substantial improvement scientific evolves. Systematic they used inform design trials, respect appropriate dose, timing crucial aspects regimen, further improve predictability