We spoke to Professor Michel Goldman, Executive Director of the Innovative Medicines Initiative (IMI), about the eTOX project, which has the potential to reduce the number of animal tests needed in drug development.
The eTOX project is part of IMI, a collaborative effort between the European Union and the pharmaceutical industry association EFPIA aiming to speed up the development of better and safer medicines for patients.
What is the eTox project and who is involved?
The eTox project is conducted by a unique consortium made up of industry, academia and SMEs. It also involves the regulators. The goal is to take advantage of the new models that are making use of computers to better predict the potential risks of new drugs.
In the four years that the project has been running, we have seen the development of databases that consolidate information, meaning that when someone asks questions about the potential risks associated with a compound there is one place where all of the information is available. The other aspect is to use the power of computers to simulate the action of drug candidates on humans or human tissues or cells, so that companies can make informed decisions as early as possible regarding moving a drug forwards, whilst trying to avoid the use of animals.
How reliable are computer models in comparison to animal models?
The demonstration of reliability is a real challenge; sometimes a drug can reach the market following clinical trials and, still, some years later major problems can become apparent. The real problem is that there will always be uncertainty in the drug development and safety testing process. When taking a drug under real world conditions sometimes the patient may forget to take their drug, sometimes they might take two pills instead of one, or they may take other pills alongside which will affect the action of the drug. It’s not ideal but it happens, this is life.
I must be careful when saying how reliable computer models are. Are they more reliable than animal models? I think that I am quite confident to say that they will be as reliable as many animal models, for which we know there are many caveats. This doesn’t mean that they will be able to predict everything, or that they will necessarily be able to predict more accurately than the models we use at the moment, but they should be as reliable.
If adverse effects can be predicted in vitro or using computer models then we should take advantage of that. I am quite confident that, if the use of cells lines or the use of computers can provide some evidence that there might be a risk with a certain drug candidate, this will lead companies to stop developing this drug further as the chance of something happening is significant. I think the major goal is to reduce the number of drugs with potentially unacceptable risks being further developed. If you can predict a potential problem earlier without exposing animals, obviously it’s good for animal welfare, but it also reduces the costs dramatically.
What sort of information is needed to be fed into a computer model before it can start making predictions? Do you need information from animals?
Not necessarily. Let me give you one example; the risk of drugs to the heart. We know that heart muscles can be sensitive to the action of drugs and that this sensitivity can result in arrhythmia. If a patient takes a drug and we do an ECG on that patient before and after they take the drug, we can find the acute interval. This is a measurement on the ECG which can reveal the risk that the patient will have arrhythmia, including fatal arrhythmia. We know that this risk is linked to the action of several ion channels. In this instance, we don’t need to expose animals to the drugs being tested as we can mimic the action of the drug in vitro using ion channels in myocytes. The interesting thing is that this approach mimics the actions based on the knowledge of human cells not animals.
Does that mean that this approach is more predictive of the human response to the drug?
Again I must be careful; I cannot say that you do not need animals. The dream would be that one day the regulators would say, ‘We want to see the effect of your new drug on the computer without the need to go through animals’. The regulatory requirements are critical. But we are not there yet. I think that there will be an intermediate phase where animals are still needed to convince the regulators of a drugs safety. However, I think that already computer models can make a difference, because if a computer is able to show a significant risk then the drug will not progress to further trials, and therefore savings in animal use will be made. That being said, it is very important that people understand that at the moment animals are still needed for most areas of research.
What sort of limitations do the computer models have?
The limitation is that we will never get results that are as predictive as if we were able to test the action of a new drug candidate in a human being. Even if we were able to test on humans there would still be limitations. Every person is different. One patient might respond well to a new drug and their sister might still develop problems. It will never be 100%, there will always be limitations whichever way we test drugs. The point is that we should do everything we can to test drugs as thoroughly as possible.
All decisions should be based on fact and not on economic consideration or whether a company is willing to sell a product. The real question is trust. If there is a problem with a drug that makes it to market we need everyone to be convinced that everything possible has been done to try and predict and prevent that problem happening. That’s why it is so important that we have several companies working together in consortia like eTOX. It limits the risk that financial considerations take precedence. I think this is what the public would question first if there was a problem with the way a drug had been tested.
In terms of the eTOX consortium, what are the practical considerations in combining data from such a wide range of sources?
I think there are two important points. Data sharing must be efficient. It is important to be sure that the data is shared for the purpose of giving more precise and more reliable information. The other aspect is to ensure that data is shared in a responsible manner. To be efficient and reliable data must only be combined if they are expressed the same way. It is not possible to compare pears, apples and oranges and assume that it must be useful simply because data is being shared. Any data that is entered in the system must be compatible and harmonised. Some people think that all data should be public and everything should be shared. But it’s not easy to organise. There will always be claims that industry should disclosure more information. But the other aspect is that, outside industry, some people just want to share data for political reasons and this is not in the public interest either. I think it’s in everybody’s best interest that the people who pool the data and make the analysis do it in a responsible way.
Is it possible to measure the animal reduction in animal use as a result of the eTOX project?
It’s a difficult question, but it’s an important one. Putting numbers on our achievements is a challenge for the future. I think collecting this type of metric is possible, but it is a lot of work and we have to think about the methodology. Much of this will only become concrete a few years from now, because first the new approach must be accepted and endorsed by the companies and by the regulators. Sometimes it is possible to decide, for example, that if this computer shows this then we will not need to expose this number of animals to that. We can make a start in measuring reductions but for this we need companies to agree to tell us about their strategies, so it’s not an easy question. However, it is an important to ask this. I think we have established the basis, we have demonstrated that it is possible to have people working together and we have the first results. Now in the second phase we must translate these results into this type of metrics. It will take time to get answers, but it is feasible.
What are your hopes for the future of the project?
My hope is very simple; it is that in the not too distant future the regulatory agencies, perhaps starting with the EMA, will include some output of this project in their own guidelines for the assessment of the safety of new drugs.