Skip to main content

Publication bias in animal studies

Macleod and colleagues recently published a paper in PLoS Biology. They analyzed 16 review articles published in the scientific literature. The articles covered 525 scientific studies analyzing treatments for strokes in animal models. They found that only 2% of the 525 papers reported on a treatment that had failed in the animal model. This calls into question the reliance on animal models to predict human response. They suggest that because of the bias for journals to only publish positive responses there is an overestimation of treatment efficacy.

            The Macleod paper is a good scientific study and I said this when I was interviewed for a news report in Nature News. The publication has received good coverage in the media, as it should. However, what I said in Nature is true: "What you really want in drug trials is an animal model that can predict human responses, and that just violates the rules of evolutionary biology."

            Publication bias (publishing only positive results) in the scientific literature is well known. Macleod’s article gave more substance to the assumption that this is true of studies in animals as well as nonanimal studies. But I still say the reporting surrounding the article misses the even bigger point that animal models simply cannot predict human response. Granted, negative results from animal studies do not get published as much as positive ones and granted that pharmaceutical companies release positive data from their in house animal studies to support their marketing efforts. (Pharmaceutical companies also hold back negative studies from the FDA.) But there are studies where the effects of drugs on humans have been retrospectively compared to the results from animals and these studies clearly reveal that the animal studies did not predict the human response. (See (Shanks and Greek 2009) for these studies and others supporting my position.)

            I am not alone in this position. Leslie: [According to Mark Davis, an immunologist at Stanford] “Researchers' reliance on mice deserves some of the blame for this ignorance . . . But mice, says Davis, make a "lousy model" for the human immune system. The human and mouse lineages diverged some 65 million years ago, and the rodent's immune system has adapted to safeguard a small, short-lived animal that scurries around with its nose in the dirt . . . "Hundreds of clinical trials have been based on curing mice, but almost none led to clinical treatments." (Leslie 2010)

            Matthews: ‘In other words, the data provide no statistically credible evidence that these animal models contribute any predictive value, either separately or in combination.” (Matthews 2008)

            Hurko and Ryan of Wyeth 2005: “Experience demonstrated that behavior of a compound in a patient cannot be predicted perfectly from interactions with isolated human molecules or cells in vitro, nor from animal models.” (Hurko and Ryan 2005)

            Friese et al: “Mice are not the natural disease host, and conventional models have proved to be poor predictors of efficacy and safety in recent trials aiming to translate drug and biologic treatments to humans.” (Friese, Jensen, Willcox and Fugger 2006)

            Duyk: “The villain in this story [the failure of new drugs to make it to the market] is the inherent lack of predictability of our available models for complex biological processes and the inability of our current life science paradigm to provide an effective road map for improvement.” (Duyk 2003)

            January 12, 2006, U.S. Secretary of Health and Human Services Mike Leavitt: “Currently, nine out of ten experimental drugs fail in clinical studies because we cannot accurately predict how they will behave in people based on laboratory and animal studies.” (FDA 2006)

            Such example can be easily multiplied.


Duyk, G. 2003. Attrition and translation. Science 302 (5645):603-5.

FDA. 2010. FDA Issues Advice to Make Earliest Stages Of Clinical Drug Development More Efficient. FDA, June 18, 2009 2006 [cited March 7 2010]. Available from

Friese, M. A., L. T. Jensen, N. Willcox, and L. Fugger. 2006. Humanized mouse models for organ-specific autoimmune diseases. Curr Opin Immunol 18 (6):704-9.

Hurko, O., and J. L. Ryan. 2005. Translational research in central nervous system drug discovery. NeuroRx 2 (4):671-82.

Leslie, M. 2010. Biomedical research. Immunology uncaged. Science 327 (5973):1573.

Matthews, R. A. 2008. Medical progress depends on animal models - doesn't it? J R Soc Med 101 (2):95-8.

Shanks, N, and R Greek. 2009. Animal Models in Light of Evolution: Brown Walker.


Popular Video