Animal testing fails more than 90% of the time!

Science cannot predict which animals, if any, will have the same reaction as humans.

Animal tests are often conducted to try and prevent dangerous or ineffective new drugs from reaching human trials. We can see how well this is working by looking at the overall failure rates (the rate at which a promising drug from animal tests end up failing in human trials).  

‍Spoiler alert: it isn't working well.

How new drugs are tested and approved

While the exact parameters vary between countries, most require a number of animal tests for any drug before human trials can be started.¹ Including New Zealand. These usually include several different species:

• Several studies aiming to study the effect the drug is supposed to have, like curing a disease

• Toxicity studies, giving the animals way higher doses of the drug over a longer time than what is calculated a human patient would need

• Tests on pregnant animals to see if the drug has negative effects on development before birth.

To progress to human trials, a drug would not only have to clear all the toxicity testing, it would also need to show enough promise for the pharmaceutical company to invest the millions of dollars for human trials.

Human trials happen in stages, starting with minimal doses given to a few healthy people. If that goes well, more people are recruited and eventually, the drug is tested for its actual function to cure a certain disease or ease a symptom.²  

Regardless of prior animal testing, most new drugs fail human trials.

‍

Studies, reviews and reports

There are a considerable number of publications backing the often-cited over 90% failure rate.

In 2012, figures reported by the U.S. Food and Drug Administration (FDA) showed a 96% failure rate despite efforts to make animal studies more effective. They summarise:³

‍“Half fail because they simply do not work - efficacy testing in animals was wrong. [184] Another 30% fail because they are unsafe - safety testing in animals was wrong.[185] And 20% fail because they are no better or safer than other available drugs. [186]
Additionally, about half of those few drugs that succeed in clinical trials and receive FDA marketing approval are later relabeled or withdrawn for serious or lethal adverse effects not detected during animal testing. [187]”

A 2014 review concluded that more than 90% of new drugs that seem to be safe and efficacious in animal tests are not effective in human trials and do not make it to clinical use. This review also states:⁴  

‍“Given the large amount of animal research being undertaken, some findings will extrapolate to humans just by chance.”

Alzheimer's disease research using animals resulted in 99.6% of new medications developed in animals failed in humans between 2002 and 2012.⁵ To date, there is no cure for Alzheimer’s, and the few approved medications to slow it down have questionable efficiency and can cause dangerous side effects.⁶

Using biomarkers was found to decrease failure rates in a 2015-study using over one thousand drugs in the U.S. National Institutes of Health (NIH) drug database. Yet, that only brought the failure rate to 76% as opposed to 94% for compounds developed without biomarkers. Even using biomarkers, more than three quarters of drugs emerging as promising from animal models failed.⁷

Using public sources and commercial databases, researchers gathered data about late-stage clinical trials between 1998 and 2008 (meaning these drugs had already passed initial tests on healthy humans at lower doses). A total of 54% of these pre-screened drugs failed at these later stages, mostly due to inadequate efficiency (57% of the failures) and safety concerns (17% of the failures). It is noteworthy, that only 40% of these failures were actually published, adding to the impression of animal models leading to successful medical developments.⁸

These examples are just some of many studies we found that demonstrate a high failure rate.

Pushback

Defenders of the animal model often claim the “>90%” to be false or misleading. They frequently reference isolated studies to support their position, such as those reporting an 86% agreement between animal and human studies, and use this to dismiss concerns, suggesting, in effect, "see, the agreement is high."⁹ While it is true that such a percentage was calculated, a closer examination of the underlying data¹⁰ reveals that their definition of "agreement" is highly questionable and includes the following:

• Substantial differences in effect size were still considered to be in agreement (for instance, a large positive effect in animals and only a minimal positive effect in humans were both counted as supporting the same outcome).
• All results were reduced to a simplistic binary classification - studies were labelled as either positive or negative. This meant that if all animal studies showed a positive result, but only 55% of human studies did, the data was still considered to be in agreement.
• All interventions were weighted equally, regardless of clinical relevance or complexity. For example, the use of cannabis for pain relief was treated as equivalent to interventions for cancer treatment.

Therefore, we cannot rely on these claims of high agreement as credible evidence of the predictive value of animal studies. When the definition of “agreement” is this broad and uncritical, it undermines the strength and validity of the conclusions being drawn.