The main thrust of this blog as well as our books and articles has been that animal models are not predictive modalities for human response to drugs and disease. Despite being touted as such by the vested interest groups, many analyses have exposed this long-held concept as a myth. In addition to empirical evidence, complexity science and evolutionary biology explain why animal models fail as predictive models. Thus, science has provided both theory and evidence. The case is closed scientifically speaking. We have not been alone in pointing this out. The following are just a few examples of what the scientific community has been saying about animal models as predictive modalities.
Markou et al stated in 2009:
Despite great advances in basic neuroscience knowledge, the improved understanding of brain functioning has not yet led to the introduction of truly novel pharmacological approaches to the treatment of central nervous system disorders. This situation has been partly attributed to the difficulty of predicting efficacy in patients based on results from preclinical studies. . . . Few would dispute the need to move away from the concept of modeling CNS diseases in their entirety using animals. However, the current emphasis on specific dimensions of psychopathology that can be objectively assessed in both clinical populations and animal models has not yet provided concrete examples of successful preclinical-clinical translation in CNS drug discovery.
Fernandes and Vali write in Predicting Pathway Effects in Drug Discovery & Development, 2012:
Currently, the identification of efficacy and safety risks for a lead compound primarily uses cell line and in vivo studies. Unfortunately, these experimental systems are black boxes that offer limited visibility into selected phenotypes and biomarkers and very little insight into the effects of a compound on important physiological pathways. Due to this lack of transparency into pathway effects, it is difficult to generate insights into system-level changes in the physiological network. This is often a reason for potential oversight of toxicity issues and incorrect assessment of efficacy. In the era of molecularly targeted drugs that affect specific targets and pathways, developers must have insights into the off-pathway effects of drug candidates. Use of predictive methodologies that emulate human physiology to test the impact of the drug candidate prior to moving the drug into clinical testing is crucial to improve the drug development success rate. By predicting clinical outcomes early on, the success rate of drug development can dramatically be improved.
Sarter and Tricklebank in 2012:
Concerns about the ability of preclinical psychiatric drug discovery to find novel candidate drugs and accurately predict their clinical efficacy have been a key factor in the recent reduction in investment by industry in this field. . . . Given the subsequent failure of many such compounds in clinical trials, it is clear that the predictive validity of existing preclinical screens for drugs with novel mechanisms is often poor. . . . A single animal model is of course unlikely to accurately reflect all of the clinical features of a complex disease such as schizophrenia. The traditional approach is to select one or more of the standard animal models for schizophrenia and simply determine whether a candidate compound improves the behavioural/cognitive functions in these animals. This approach does not depend on knowledge of the circuitry underlying these dysfunctions; rather, it ignores the implications of a substantial systems neuroscience gap. It may produce a list of beneficial effects, but provide little insight into the mediating mechanisms and little opportunity to alter the preclinical strategy as a result of false negative or positive predictions. If a candidate compound failed to produce therapeutic effects in patients, we would not know how to translate this finding back in terms of what animal models to use. What is at fault in this situation: the model, the target or the compound?
Griebel and Holsboer in 2012:
The search for novel drugs for treating psychiatric disorders is driven by the growing medical need to improve on the effectiveness and side-effect profile of currently available therapies. Given the wealth of preclinical data supporting the role of neuropeptides in modulating behaviour, pharmaceutical companies have been attempting to target neuropeptide receptors for over two decades. However, clinical studies with synthetic neuropeptide ligands have been unable to confirm the promise predicted by studies in animal models. . . . Unfortunately, nearly 15 years after the publication of the antidepressant effect of aprepitant in a clinical trial, no drug acting at a neuropeptide system has made it to the market for the treatment of a psychiatric disease (although aprepitant was later approved for alleviating chemotherapy-induced nausea). Many Phase II/III trials of drugs targeting neuropeptide receptors have failed (Table 1). Furthermore, among the 40 drugs that are currently being tested in Phase II/III trials for schizophrenia, MDD or anxiety disorders, very few are compounds that target neuropeptide receptors. . . . Among the most commonly noted reasons for the failure to successfully develop neuropeptide receptor ligands for MDD, anxiety disorders or schizophrenia is the poor predictivity of the animal models that have been used to screen these molecules. It is beyond the scope of this Review to address the issue of animal models of psychiatric diseases, which has been reviewed thoroughly in a recent article56, but it is noteworthy that in some instances what was considered to be compelling preclinical data did not translate into convincing clinical efficacy findings, thereby questioning the profiling strategies that were used for those drugs.
Tillman, from 1999: “An emerging bottleneck in psychopharmacological drug discovery is the relative paucity of preclinical behavioral models predictive of clinical efficacy and the need to carry out early clinical trials to demonstrate therapeutic utility.” In 2008, Mark Davis , said that mice were “lousy models” for studying human disease. Davis reasoned that this was the case because of evolutionary divergence of mice and humans, among other reasons. He suggested humans be studied instead.
Roep et al in 2012:
The inability to predict the likelihood of success of therapies in humans using animal models may be related to the distinct differences in the immune systems of mice and humans. Any immune phenotype or immunological process or mechanism may have species-specific features that preclude direct extrapolation or comparison between mice and humans. Marked differences between the immune systems of mice and humans include key discrepancies in both innate and adaptive immunity. In a published list of more than 80 of these differences, some examples include the differential expression of regulatory T (Treg) cell markers (notably, forkhead box P3 (Foxp3)), variations in the balance of leukocyte subsets, defects in antigen-presenting cells, dysregulation of thymic selection, differences in the role played by cells that produce interleukin-17 (IL-17) in disease, inflammation and immune regulation and complement deficiencies [7,8]. This large number of disparities between mice and humans has a considerable impact on the differences in the immune processes that drive the development of autoimmune disorders in the two species. The time has come to challenge the notion that immunologic processes in mice are similar to those in humans. Any given animal model will not be representative of the entire breadth of molecular and clinical heterogeneity present in human populations but, rather, the animal model more likely corresponds to specific aspects of the complex disease manifestations in humans observed in the clinic. This hypothesis has implications for the truism entertained by many institutional review boards, the US Food and Drug Administration and the European Medicines Agency that any immune intervention therapy should first be validated using preclinical animal models. The key issue here should be how well the animal model reproduces the specific immunologic characteristics of human disease. To address this, we must first understand the unique immunologic alterations present in each human autoimmune disease. This approach will allow for the use of relevant models to test safety and assess clinical efficacy so that drugs that perform well at the preclinical stage will be more likely to be effective in humans.
I disagree with that last sentence as mice and humans are complex systems, thus extrapolating between modules or subsystems is going to be problematic. But Roep et al do understand the problem of animal modeling in general.
MacDonald and Robertson state: “The approaches that have been taken to assess human risk of adverse effects from chemical exposure have changed very little over the last several decades. . . . The global regulatory requirements for registration of new human pharmaceutical chemicals—the data requirements have changed very little since their establishment three and sometimes four decades ago despite the dramatic advances in the sciences that are used for this activity.” 
The FDA has acknowledged the need to make toxicology science-based  with FDA Commissioner Margaret Hamburg echoing MacDonald and Robertson stating: “Most of the toxicology tools used for regulatory assessment rely on high-dose animal studies and default extrapolation procedures, and have remained relatively unchanged for decades, despite the scientific revolutions of the past half-century.”
Turka and Bluestone: “Although it is relatively easy to induce transplantation tolerance in rodents, few approaches have shown success in people, partly due to the robust alloimmune response.” Elias Zerhouni, former director of NIH and current head of R&D at Sanofi was quoted in the June 25, 2012 issue of Forbes as saying: “R&D in pharma has been isolating itself for 20 years, thinking that animal models would be enough and highly predictive, and I think I want to just bring back the discipline of outstanding translational science, which means understand the disease in humans before I even touch a patient.”
Scannell et al: “The past 60 years have seen huge advances in many of the scientific, technological and managerial factors that should tend to raise the efficiency of commercial drug research and development (R&D). Yet the number of new drugs approved per billion US dollars spent on R&D has halved roughly every 9 years since 1950, falling around 80‑fold in inflation-adjusted terms.” 
Raven discusses rodent models of sepsis and quotes two experts:
“The mouse models really don't reflect the human condition,” says Shaw Warren, an infectious disease specialist at the Massachusetts General Hospital in Boston. “Clearly, current animal models seem to be incapable of predicting results in human trials of new agents,” says Mitchell Fink, a surgeon at the University of California–Los Angeles.
Jocelyn Rice writes in 2012:
Dozens of interventions, some as simple as the spice turmeric and others as sophisticated as gene therapy, have shown promise in animal models of MS. . . . EAE has been troublingly unreliable for screening human MS treatments. For every drug that has translated well into humans, there have been many more that did not work. Finding that a drug can cure EAE gives almost no indication of whether it will help treat MS. In fact, the EAE success stories are so random as to be “almost coincidental”, says Richard Ransohoff, director of the Neuroinflammation Research Center at the Cleveland Clinic in Ohio. In fact, some interventions that proved therapeutic in animals with EAE — such as interferon γ and tumour necrosis factor (TNF)-α blockade — actually make human MS worse. Even the success stories sometimes have a dark side. The drug natalizumab, for example, which can reduce relapses and slow the progression of disability in human MS, was temporarily withdrawn from the market when some patients developed a potentially fatal brain infection not seen in the animal model. “If you wanted to have a robust, predictable platform for drug screening, you'd consider the animal-model situation to be very disappointing,” says Ransohoff. “We would like to have a model that would tell us it's worth pushing a drug, or that we can drop the drug without a second thought,” says immunologist Roland Liblau of the Rangueil University Hospital in Toulouse, France, president of the French Society of Immunology. “But it's not so easy in real life.” 
Cook et al:
Over many years now there has been a poor correlation between preclinical therapeutic findings and the eventual efficacy of these [anti-cancer] compounds in clinical trials [17, 18]. . . . The development of antineoplastics is a large investment by the private and public sectors, however, the limited availability of predictive preclinical systems obscures our ability to select the therapeutics that might succeed or fail during clinical investigation.
The above can be easily multiplied.
The lack of predictive ability is not confined to one disease, one aspect of drug development, or one species. It is a fact of evolved complex systems. When vivisection activists play word games to cast doubt on the fact that animal models are not predictive modalities for drug and disease response, consider the source.
1. Markou A, Chiamulera C, Geyer MA, Tricklebank M, Steckler T: Removing obstacles in neuroscience drug discovery: the future path for animal models.Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology 2009, 34:74-89.
2. Predicting Pathway Effects [http://www.dddmag.com/articles/2012/06/predicting-pathway-effects?et_cid=2695933&et_rid=45518461&linkid=http%3a%2f%2fwww.dddmag.com%2farticles%2f2012%2f06%2fpredicting-pathway-effects]
3. Sarter M, Tricklebank M: Revitalizing psychiatric drug discovery.Nat Rev Drug Discov 2012, 11:423-424.
4. Griebel G, Holsboer F: Neuropeptide receptor ligands as drugs for psychiatric diseases: the end of the beginning?Nat Rev Drug Discov 2012, 11:462-478.
5. Tallman JF: Neuropsychopharmacology at the new millennium: new industry directions.Neuropsychopharmacology 1999, 20:99-105.
6. Davis MM: A prescription for human immunology.Immunity 2008, 29:835-838.
7. Roep BO, Atkinson M, von Herrath M: Satisfaction (not) guaranteed: re-evaluating the use of animal models of type 1 diabetes.Nat Rev Immunol 2004, 4:989-997.
8. Mestas J, Hughes CC: Of mice and not men: differences between mouse and human immunology.J Immunol 2004, 172:2731-2738.
9. Roep BO, Buckner J, Sawcer S, Toes R, Zipp F: The problems and promises of research into human immunology and autoimmune disease.Nat Med 2012, 18:48-53.
10. MacDonald JS, Robertson RT: Toxicity testing in the 21st century: a view from the pharmaceutical industry.Toxicological sciences : an official journal of the Society of Toxicology 2009, 110:40-46.
11. Advancing Regulatory Science at FDA: A Strategic Plan [http://www.fda.gov/ScienceResearch/SpecialTopics/RegulatoryScience/ucm267719.htm]
12. Hamburg MA: Advancing Regulatory Science.Science 2011, 331:987.
13. Community Corner: A step closer to effective transplant tolerance?Nat Med 2012, 18:664-665.
14. Scannell JW, Blanckley A, Boldon H, Warrington B: Diagnosing the decline in pharmaceutical R&D efficiency.Nat Rev Drug Discov 2012, 11:191-200.
15. Raven K: Rodent models of sepsis found shockingly lacking.Nat Med 2012, 18:998-998.
16. Rice J: Animal models: Not close enough.Nature 2012, 484:S9-S9.
17. Johnson JI, Decker S, Zaharevitz D, Rubinstein LV, Venditti JM, Schepartz S, Kalyandrug S, Christian M, Arbuck S, Hollingshead M, Sausville EA: Relationships between drug activity in NCI preclinical in vitro and in vivo models and early clinical trials.Br J Cancer 2001, 84:1424-1431.
18. Suggitt M, Bibby MC: 50 years of preclinical anticancer drug screening: empirical to target-driven approaches.Clinical cancer research : an official journal of the American Association for Cancer Research 2005, 11:971-981.
19. Cook N, Jodrell DI, Tuveson DA: Predictive in vivo animal models and translation to clinical trials.Drug Discovery Today 2012, 17:253-260.