Skip to main content

Liars and Statistics. Part III

In Liars and Statistics Part I and II, I presented comments from scientists linking the drug failure problem to animal models and began my analysis of Lovell-Badge’s essay that attempts to separate the two. I complete my analysis in this essay. (I remind the reader that all comments in quotations that are in bold are my emphasis not the original author’s.)

Lovell-Badge continues, stating: “The first thing to note is of those drugs which pass animal tests, 94% will fail during human clinical trials stages (Phases 1 – 3)*.” This is also in line with previous figures. Lovell-Badge continues: “So the failure rate is actually higher than even the animal rights organisations suggest (since they are using data from before 2006).” I fail to see why a difference of a couple of percentage points is relevant but if you have no relevant facts then I guess you try to make a mountain out of a molehill. Lovell-Badge: “Is this damning for animal research? Consider that of all the drugs which pass Phase 1 clinical trials in humans, 86% will fail in later stage human trials**. Yet, we do not hear activists suggesting that humans are an entirely inappropriate model for drug development (though we should note that one human is not a perfect model for another).”

There are four rejoinders to the above. First, Phase I human trials usually test for safety not efficacy. It is dishonest to say that since X% of drugs that tested positive for characteristic Y (safety) failed when tested for characteristic Z (efficacy), therefore the test for characteristic Y is flawed. Second, on the other hand, drugs that test safe in Phase I do sometimes fail in Phases II and III and this is one reason patient advocates have asked that higher numbers of volunteers or patients be tested in Phase I as well as Phases II and III and or that approval be gene-based. Third, as I have said, Pharma needs new tools: tools that actually work. In vitro, and Lovell-Badge sort of has a point here, do not work very well either in terms of efficacy and safety. I have criticized animal rightists for claiming that current in vitro and in silico can do everything that needs to be done in drug development. But there are no scientists with a vested interest in in vitro tests out there campaigning and selling in vitro-based methods to society based on false claims. Basic researchers that use animals sell animal models as being of predictive value even when applied researchers have acknowledged they are not predictive. The reason for this is explained in our essay Is the use of sentient animals in basic research justifiable? Briefly, basic researchers cannot get funded unless they claim animal models are predictive and universities cannot get their cut of the money if basic researchers are not funded.

Fourth, patient advocates along with scientists have asked for genotype-based testing as opposed to the testing system we currently have. So, contrary to Lovell-Badge, animal activists and others have pointed out that any given human is, or a small group of humans are, inappropriate for representing the general population, as test results vary by genotype. This is the basis for personalized medicine and readers of this blog know I bring it up quite often.

Lovell-Badge continues:

Before the preclinical animal tests there are a large number of pre-preclinical non-animal tests done on all manner of research tools including computer models, automatic screening, cell cultures, microbial studies and more. These methods are used to (relatively) cheaply remove many potentially toxic, or obviously non-starting drugs from reaching the more expensive animal testing stage – greatly reducing the amount of animal research required for a drug to reach market. So contrary to animal rights claims of alternative methods being better, the truth is that 94% of drugs that pass animal AND non-animal preclinical tests will fail in human tests. So rather than damn just the animal tests, have animal rights activists managed to damn all of preclinical research? In short, no.

In short, no! The above is very misleading. The tests performed prior to animal testing are not performed to ascertain safety and efficacy, for the most part. They are performed to ascertain the physico-chemical properties of the drug: things like pH and pKa and these are not why drugs fail. Animal tests are not designed or performed to test physico-chemical properties, rather they are supposed to test primarily safety and efficacy. If a test is supposed to predict safety and efficacy and the biggest reason new drugs fail in human clinical trials is safety and efficacy then it is misleading to clam that the REAL problem with drug development is back at the in vitro and in silico stage.

Lovell-Badge: “The role of preclinical animal tests is to check if the drug offers any potential therapeutic value and, importantly, if it is safe enough to move to Phase 1 trials in humans. This does not even mean free of all side effects, but to learn whether a drug can safely be given to humans and at what approximate dosage.” If by “preclinical,” Lovell-Badge means animal then this is correct. But if he is lumping all preclinical together then this is the fallacy of distribution again. Different stages of development are meant to accomplish different things. Animal tests are primarily meant to test safety and efficacy and safety and efficacy are why drugs fail in human clinical trials. Lumping all stages of drug development into one stage and saying that the chemistry tests are equally as bad as the animal tests in terms of drugs failures is fallacious.

Lovell-Badge then goes on to tout the claim that Phase I human clinical trials have had no deaths in the UK. But this is a claim without substance. Consider the following: Speaking of toxicity trials for new drugs in humans, an unnamed clinician quoted in Science stated, “If you were to look in [a big company’s] files for testing small-molecule drugs you’d find hundreds of deaths [1].” Chapman states in 2011: “. . . but other incidents of harm [besides TGN1412], even death, to participants in Phase I trials, some then known and other unpublicized, had taken place” [2] So much for animal testing protecting those undergoing clinical trials, in the UK or elsewhere.

Furthermore, the reason Phase I studies are as safe as they are, is attributable to the very small doses that are used. This is not due to animal studies but rather the comfort level of the people responsible for the trials. Giri and Bader illustrate this in their 2011 article:

Often, toxicity of a drug candidate is not discovered in preclinical stages until clinical trials are conducted. As an up to date example, TGN1412 (also known as CD28-Super- MAB) is an immunomodulator for the treatment of rheumatoid arthritis. In 2006 clinical trials on six volunteers were carried out using a 500-times lower dose than the dose found safe in animals [3] Nevertheless, four volunteers suffered from multiorgan failure. Clearly, drug testing on animals is unrealistic and causes unforeseen reactions in human clinical trials.[4]

Nobody trusts animal tests to assign dose. I have advocated a system for starting all doses at 1ng and working up from there. [5]

Opar, writing in Nature Reviews Drug Discovery, 2012, supports my position regarding the lack of predictive ability for animal models for establishing FIH doses:

The damage hit without warning. Of the 40 healthy volunteers who had enrolled in a Phase I trial for panadiplon, an experimental anxiety drug, 11 quickly showed liver injury, even though rat, dog and monkey studies hadn’t hinted at any toxicity issues. “That was really scary,” says Roger Ulrich, who worked on the study in the 1980s, “we had no idea what would happen to these people.” The individuals recovered, but the agent was killed. Fast forward to 2009, when liver toxicity popped up in a Phase I trial for Gilead’s cancer drug CAL-101. “It was the same boat — we didn’t know what was causing it, or whether these people were in dire straits or not,” says Ulrich, who has 30 years of drug development experience at big and small pharma companies. The elevated liver enzyme levels turned out to be asymptomatic, and the agent has since progressed into Phase III trials. [6]

Needless to say, things like this are not highly publicized.

Lovell-Badge continues: “Furthermore, when a drug is licensed for use, it is on the basis of the clinical trials in humans, not the preclinical animal tests which exist to ensure that a drug is safe enough to move into Phase 1 trials. So when animal rights activists claim that adverse drug reactions can be blamed on animal tests approving the drug, remember that it is the clinical trials in thousands of people which provide the evidence of its safety.” This is not entirely true. It should be true but animal results often do influence regulators even though such results should not. The FDA can even base approval on animal tests. From the FDA website:

Sec. 314.610 Approval based on evidence of effectiveness from studies in animals.

(a) FDA may grant marketing approval for a new drug product for which safety has been established and for which the requirements of 314.600 are met based on adequate and well-controlled animal studies when the results of those animal studies establish that the drug product is reasonably likely to produce clinical benefit in humans. In assessing the sufficiency of animal data, the agency may take into account other data, including human data, available to the agency.

(Anecdotally, I have heard that the FDA does consider animal data when contemplating final approvals for certain drugs that do not meet the above criteria. I have no references for this, however.)

Regardless, no one that I know of (save Pharma and the FDA) has ever claimed that testing a small group of humans was adequate for releasing a drug to the general population. Society actually needs genotype-based drug development and if that is too much to ask then longer and broader clinical trials should be required. Spedding et al., writing in Nature Reviews Drug Discovery 2005, state: “Animal models often cannot be transposed to Phase I and Phase II clinical testing, and Phase I/II clinical testing is often not transposable to Phase III trials and the general population.” [7]

Lovell-Badge’s essay is wrong on several levels:

1. He has certain verifiable facts wrong. For example, the reason animals are used as opposed to the reasons in vitro and in silico are used. He presents so much misleading or simply wrong information that no one can counter it all, even in three essays.

2. His reasoning is fallacious.

3. His position is at odds with the opinions of the experts I quoted.

4. But most of all, his position is at odds with the clinical data. Expert opinion is meaningful in some instances but when empirical data contradicts the experts, the experts are wrong. In this case, both the data and the experts agree.

5. The way Lovell-Badge presented his case is, however, the most egregious violation for a scientist. He asked the reader to accept claims without offering evidence. This is anti-science and as it concerns human safety and health it is anti-human.

In the final analysis, the only predictive model for you is you, or at least your genotype. Pharma is going to have to test on humans much earlier in the drug development process. Not because Pharma is wicked and wants to torture humans, but because the facts of the material universe do not allow any other way to currently gain the information Pharma needs (and that you and I need when we are patients). Maybe someday science will know everything about the genome and simply removing some DNA with a cheek swab will allow your physician to know everything about how you will respond to a new drug. That type of thing is happening for some patients suffering from some diseases. But for the time being, human clinical testing is the best we have. Although such testing can be safe and ethical, society is going to have to address this issue and understand what it entails.

(I have addressed many aspects of the above in various blogs, articles, and books. If you are new to this issue I suggest you read FAQs About the Use of Animals in Science: A handbook for the scientifically perplexed. It is good starting place and there is more to read after you have digested it. This is not an issue that can be adequately addressed in blog form. Sorry, but you really do have to read books!)


1.         Marshall, E (2000) Gene therapy on trial. Science 288:951-957.

2.         Chapman, AR (2011) Addressing the Ethical Challenges of First-in-Human Trials. J Clinic Res Bioeth 2:113. 10.4172/2155-9627.1000113.

3.         Suntharalingam, G, MR Perry, S Ward et al. (2006) Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. The New England journal of medicine 355:1018-1028. 10.1056/NEJMoa063842.

4.         Giri, S, A Bader (2011) Foundation review: Improved preclinical safety assessment using micro-BAL devices: the potential impact on human discovery and drug attrition. Drug Discovery Today 16:382-397.

5.         Greek, R (2013) Animal Models in Drug Development. In: Gowder, S (ed) New Insights into Toxicity and Drug Testing. InTech, Manhattan, p 124-152. 980-953-307-844-6.

6.         Opar, A (2012) Overtaking the DILI Model-T. Nat Rev Drug Discov 11:585-586.

7.         Spedding, M, T Jay, J Costa E Silva et al. (2005) A pathophysiological paradigm for the therapy of psychiatric disease. Nat Rev Drug Discov 4:467-476. nrd1753 [pii] 10.1038/nrd1753.


Popular Video