I frequently point out that since humans vary in their responses to drugs and disease, it should come as no surprise that humans and animals also vary. Another example of intra-species variation recently came to light. Scientists used fMRI to study the human brain and discovered that areas of the brain involved in control and attention exhibited more differences, in terms of the connections they make, than areas involved with sensory input.  This intra-species variation may explain behavioral and cognitive differences among humans.
Compare the above with the following from /t Hart et al, which is typical for the animal model community:
Nonhuman primates (NHPs) are important models in preclinical research enabling understanding of pathogenic mechanisms in human disease that readily translate into therapy development. Safety and efficacy testing of biologics, such as antibodies, soluble receptors and cytokines, often precludes the use of lower species, for example when drugs fail to bind targets. These issues explain increasing interest in NHPs as preclinical disease models, despite costs and ethical limitations.  (Emphasis added.)
By stating that research on NHPs readily translates to humans in terms of disease mechanisms, /t Hart et al are claiming the models are predictive for humans. But note they provide no evidence for this claim. If forced to explain this statement, the authors would probably fall back on anecdotes where similarity was seen between species. This does help use determine positive or negative predictive value. Anecdotes can be found for the use of dowsing rods to find water and randomly drilling wells for oil. Yet, very few successful business people routinely use random drilling to find oil or dowsing rods to find water. Why? Because there is no predictive value in the use of either.
Moreover, when they state that testing biologics on lower species is not valid, they ignore the fact that the reasons such animals are not valid predictors also apply to their model. (Note: science does not use the term lower species anymore as it has creationist origins and denies the complexity of all life forms.) This is common among animal modelers. Their species of choice is great but everyone else’s is not valid. So fund them and not the other guys. Finally, I know of no increasing interest in using NHPs. The evidence confirms that genetically modified mice are the animal models generating the most interest as well as the most use.
/t Hart et al continue:
The common marmoset (Callithrix jacchus) is a small-bodied New World monkey (Fig. 1a) widely used in studies ranging from toxicology, neurological and autoimmune disease, and reproductive biology to stem cell biology and transgenics. Marmoset colonies are outbred reflecting the genetic heterogeneity of the human population, although differences exist compared with humans. 
Widely used is not synonymous with successful. As I have repeatedly stated, if I were to highlight one area for proclaiming the success of animal models, it would not be toxicology. That is the one area where direct human to animal model comparisons have been repeatedly made and animal models of all species have failed. For example, drugs known to damage the human fetus are found to be safe in 70% of cases when tried on monkeys. [ p312-13] As for neurological research, Markou, Chiamulera, Geyer, Tricklebank (of Eli Lilly), and Steckler (of Johnson and Johnson) state:
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. 
Such statements could be easily multiplied. Finally, the fact that populations of two different species are heterogenous does not mean they are different in the same relevant ways. Thus, their statement is meaningless in terms of the validity of using NHPs in research designed to predict human response to drugs and disease.
/t Hart et al continue:
As an example [of genetic heterogeneity], the major histocompatibility complex genomic region, which encodes central regulatory molecules of immune responses, is much more polymorphic in humans and Old World monkeys (e.g. cynomolgus or rhesus monkey) than in common marmosets. Laboratory-housed colonies are often kept under conventional conditions that expose monkeys to naturally occurring pathogens that shape their immune system. As a consequence, marmoset disease models are appropriately complex and their use requires in-depth knowledge of marmoset biology and optimal laboratory management. There is already a vast body of information available about the marmoset as a model of human disease (e.g. [2–4]) and on related biotechnical aspects [5–7]. To be of use for the biomedical research community, this information needs to be readily accessible. . . .  (Emphasis added.)
Once again, the fact that disease models are complex is the problem not the solution to the problem. I am beginning to see animal modelers use the terms complex and complex system more and more, but they do not seem to really understand what complex means. The fact that two systems being compared are both examples of systems that are differently complex means the probability of extrapolating between one for the other, at higher levels of organization, such as where disease and drug reactions occur, is very low. The characteristic of complexity is not, contrary to what these authors seem to think, yet another item to be added to a list of commonalities between species thus adding substance to the expectation that such models will be of predictive value.
This goes directly to their comment that there already exists a vast body of knowledge on marmoset pathology. Such is meaningless unless the marmoset pathology is predictive for human pathology. It is not. (Again, I remind the reader that I am addressing animal models of disease and drug response, not animals as heuristics or marmoset as models of other marmosets, where pathology of the model is of obvious value.) The proof for this lies in part in the track record for using marmosets and other NHPs to predict drug targets for humans. ALL animal models have a horrible predictive value for humans in terms of drug efficacy and safety; two ways target data from animal models is ultimately evaluated in drug development. If the authors have data that contradicts this, they need to produce it. But in all the drug development literature I read, nowhere do I see Pharma saying: “Gosh darn, if only we could get more marmosets, then we could cure diseases.” I read researchers at primate centers, like these authors, saying that but not the people who actually use the data generated by animal modelers like the authors.
In books and articles we have written, we have listed what probably amounts to hundreds of references from scientists in Pharma saying animal models do not predict human response. At best, from the animal modelers’ perspective, the Pharma scientists go on to say something like, “but animal models are still of great value,” without saying why such is the case or making an argument for such a claim. But most of the time they either say nothing more or equate “not predictive,” implicitly or explicitly, with “should therefore be abandoned.” I would rather have everyone read and understand the scientific literature in order to evaluate the problem for yourselves. But since that is unrealistic, I would ask you to consider the source when making a decision. People with a vested interest tend to lie in order to protect that interest.
/t Hart et al continue:
Traditionally, the most frequently used NHP species for preclinical in vivo pharmacology, safety studies and toxicology assessment has been the cynomolgus monkey (Macaca fascicularis; 80%), the marmoset (15%) and the rhesus monkey (Macaca mulatta; 5%) (Weinbauer, G., Covance Laboratories, pers. commun.). The main reason for the current preference of the large cynomolgus monkey ~3 kg) is the abundant background data that can be used as a reference repository. In several disease areas, however, including immune-mediated disorders, available models in macaques are suboptimal because they demonstrate a poor resemblance to the corresponding disease in humans.  (Emphasis added.)
Again, abundant background data does not seem to be working out for Pharma, as the attrition rate, based on all animal models, not just NHPs, is not sustainable. An attrition rate of >90% of all drugs that begin human clinical trials is not something the basic research community should be bragging about.
/t Hart et al continue:
The sequencing and annotation of the human genome is an essential step toward a better understanding of the genetic basis of human disease. The subsequent translation of genetic into functional information is a major challenge for the future. Much groundbreaking work has already been done in mice that transgenetically express human genes. However, a human gene placed in a mouse cell can behave differently than it does in a human cell and the marmoset can be of use here.  (Emphasis added.)
This is farcical. The authors acknowledge that genes behave differently depending on the environment/species and yet they go on to claim that genetically modified NHPs will give reliable data for humans when genetically modified mouse model do not. This is the type of claim that makes me wonder if the animal model community really has any understanding whatsoever of complexity science or evolution.
/t Hart et al continue:
The evolutionary proximity of marmosets and humans is reflected by their comparable brain morphology, making the marmoset a potentially useful model for neuroscience and neuropathology research. The growing importance of marmosets for research in neuroscience and related fields is demonstrated by the increasing number of brain atlases based on histology and magnetic resonance imaging that enable positron emission tomography (PET) to associate drug-induced changes in behavior with anatomically specific alterations in cerebral glucose metabolism. [2
This is an excellent example of confusing phylogenetic closeness with construct validity of a model. We state in Animal Models in Light of Evolution:
It is certainly true that from an evolutionary standpoint, we expect there to be fewer differences between humans and chimpanzees than between humans and mice, or humans and yeast. But since humans and our closest phylogenetic relatives are complex, organized, interactive systems where small differences can be of great biomedical significance, it is far from clear what follows from this observation concerning the degrees of phylogenetic closeness between humans and mice or monkeys. [p321]
(See Animal Models in Light of Evolution for more on the problem of using phylogenetic closeness in animal modeling.)
The article by /t Hart et al is representative of the genre. In reading it, I found the author’s had little appreciation for complexity theory or evolutionary biology and absolutely no critical thinking skills. The only thing that should be more embarrassing to the scientific community than the authors' profound ignorance of general science is the fact that essentially no one in the scientific community will speak out and correct the factual errors the authors' made in this essay and that other scientists made in similar essays. No one in the evolutionary biology department will criticize the authors for their frank incompetence. No one in the mathematics, physics, or electrical engineering departments will point out the author’s ineptitude regarding complex systems.
But, then again, the evolutionary biologists, mathematicians, physicists, engineers, and physicians who are employed by the universities where animal models are widely used, profit financially, directly or indirectly, from the money brought into the university by these animal modelers. So they must weight their financial and career stability against a few fallacies and errors of fact. “No big deal! Everybody does it, right?”
Even when the lives of sick children are at stake?
The scientific community should be ashamed.
1. Mueller, S, D Wang, MD Fox et al. (2013) Individual Variability in Functional Connectivity Architecture of the Human Brain. Neuron 77:586-595. http://linkinghub.elsevier.com/retrieve/pii/S0896627313000044.
2. T Hart, BA, DH Abbott, K Nakamura et al. (2012) The marmoset monkey: a multi-purpose preclinical and translational model of human biology and disease. Drug Discovery Today 17:1160-1165. 10.1016/j.drudis.2012.06.009. http://www.ncbi.nlm.nih.gov/pubmed/22728226.
3. Manson, JM (1987) Biological Considerations for Risk Assessment in Developmental Toxicology. In: McLachlan, JA, RM Pratt, CL Markert (eds) Developmental Toxicology: Mechanisms and Risks. Banbury Report 26. Cold Springs Harbor Laboratory, p 307-322.
4. Markou, A, C Chiamulera, MA Geyer et al. (2009) Removing obstacles in neuroscience drug discovery: the future path for animal models. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology 34:74-89. 10.1038/npp.2008.173. 2651739. http://www.ncbi.nlm.nih.gov/pubmed/18830240.