There has been some discussion as to whether the statistics I used and referred to in previous blogs (positive and negative predictive value, sensitivity, and specificity) and the implications thereof were representative of the science of using animals as predictive models in general or perhaps merely toxicology. It is a fair, howbeit naïve question, as anyone acquainted with science in general and biomedical science specifically will be familiar with the terms and what they represent. Such knowledge, combined with a working knowledge of evolution is sufficient to answer the question.
Nevertheless, since the issue was raised, I searched PubMed for positive negative predictive value which resulted in 25,042 matches. Some of the titles are:
Evaluation of the efficacy and reproducibility of cholangiopancreatography by magnetic resonance for detecting biliary complications following orthotopic liver transplantation.
Predictive features of severe acquired ADAMTS13 deficiency in idiopathic thrombotic microangiopathies: the French TMA reference center experience.
Prediction of posttraumatic stress disorder among adults in flood district.
Predictive model of biliocystic communication in liver hydatid cysts using classification and regression tree analysis.
Evaluation of the diagnostic value of a Modified Liquid-Based Cytology using OralCDx (R) Brush in early detection of oral potentially malignant lesions and oral cancer.
Prediction of response and progression in multiple myeloma with serum free light chains assay: corroboration of the serum free light chain response definitions.
Immunohistochemical phenotype of breast carcinomas predicts the effectiveness of primary systemic therapy.
Performance value of high risk factors in colorectal cancer screening in China.
In silico evaluation of predicted regulatory interactions in Arabidopsis thaliana.
Capnometry in suspected pulmonary embolism with positive D-dimer in the field.
I then searched SCIRUS, a scientific search engine, for positive negative predictive value and found 414,952 hits. I found 89,053 hits when I searched "positive predictive value" in SCIRUS. PubMed yielded 18,356 for "positive predictive value." Clearly, the simple statistics used to determine whether a modality is predictive are ubiquitous in the scientific literature and are not confined to toxicology or discussions about using animals to predict human response to drugs and disease.
But let’s assume the only studies comparing the results from research in animals with the results from research in humans are from the field of toxicology. Let’s assume the only data at all that are available for determining whether animal models can predict human response are from this one field. What would the results tell us?
First, there indeed have been numerous studies in toxicology comparing human and animal outcomes. The studies clearly reveal that animal models are not predictive for human response. (See Animal Models in Light of Evolution for more on this.)
Second, we would then have to address how these results could be explained? An understanding of evolution and complex systems reveals that such interspecies variation is to be expected. (Intraspecies variation is also to be expected and that is what human studies reveal.) So even if the only data we had available was from toxicology studies, if we put that in the context of evolutionary biology and complex systems, we could rightly conclude that animals models would not be predictive for drug and disease response. (I again remind the reader that anecdotes do not count when determining whether a modality is predictive.)
Obviously, that is not the only way animals are used in research. If one wishes to generate hypotheses, then using animals can fulfill that need. If one wishes to study conserved processes, then animals can fulfill that need as well. Both of these uses however will only result in a hypothesis about humans, which must then be studied in humans.
This leads us to the question of why anyone would disagree with such scientifically obvious points. As I may have said once or twice, basic research using animals is sold to society on the basis of it being able to predict human response. But testing the hypothesis that animals can predict human response for drugs and disease has resulted in the hypothesis being falsified. Should that news get out, basic researchers who use animals would be forced to admit they are doing exactly what they are in fact doing: seeking knowledge for knowledge sake. Not necessarily a bad thing in and of itself! However, since society does not seem to appreciate animals like monkeys having electrodes forced into their eyes and brains when no cures are likely to be forthcoming, then the money flow would probably cease.
Simple statistics like positive and negative predictive value, sensitivity, and specificity are how biomedical science (and other areas of science) determines whether a modality is predictive. I encourage the reader to go to PubMed and judge for himself whether this is true. I also encourage the reader to put all this in context by reading Animal Models in Light of Evolution. (I receive no money from the sale of the book.)