It’s no secret that the clinical trial process is riddled with inefficiencies that in certain respects has held the industry back. However, the emergence of new technologies in the past decade means the potential to streamline processes has never been greater. Even the ability to simulate and model clinical trials before they begin is a luxury that was not possible in the past. But have such tools been used to full effect? Einar Heiberg, CTO and Founder of Medviso AB, explains to Clinical Trials Arena why the ability to predict clinical trial outcomes could improve the way they are conducted.
Clinical Trials Arena: What are the biggest challenges the industry faces right now?
Einar Heiberg: When it comes to clinical trials, I’m not sure both the pharma and the medical device industries see that there’s a fundamental problem in the way clinical trials are conducted. The vast majority are not using adequate tools to design and plan the clinical trials ahead of time, and that is the bottom line. For instance, no one would design a car manufacturing plant without simulating or modelling it. However, the industry has been planning and conducting trials in this manner for decades.
CTA: So it appears there’s a basic lack of planning or forethought from an industry perspective?
EH: Precisely. With proper planning, there are so many mistakes that could be avoided. The only solution I see going forward is to do proper planning using computer modulation of the trials where you perform virtual trials in the computer with generated patients and see what is actually affects your statistical power. How do simplifications of normally distributed variables, dropouts or data missing points affect your required sample size? Patient crossover designs, adaptive designs – how do you ensure type 1 or type 2 errors? Is it possible to design the trial in such a way you halt it prematurely because your assumptions about the treatment effect were wrong? No one appears to be making informed decisions when carrying out trials by asking themselves, ‘what are the exact risks?’
CTA: So what improvements need to be made to improve processes?
EH: It all boils down to risk management. With simulations you can better assess what your risks are. Sponsors don’t tend to look at things like screening bias. Say, for instance, you want to enrol patients with high blood pressure (>140 mmHg). You screen patients and include the ones above the threshold. As everybody has a natural variation in blood pressure this process will include several patients that normally lies under the limit, but on that particular day some of your patients had high blood pressure. If you re-measure the enrolled patients in a month you would see a drop of about 10mmHg of just that screening effect without any treatment. There are many factors that need to be considered, such as ‘what are the consequences of my design?’ The tools are available – pharma companies are using it for Phase 1, Phase 2A trials where the dosage is low. But for pivotal trials – Phase 2B – there is very little usage, and the industry doesn’t seem to realise that this is a huge loss of potential.
CTA: You’ve talked a lot about the industry needing to plan ahead before the start of a trial. Do you see any signs of progress being made on that front?
EH: I have worked on both sides in medical devices, both in trials and as a researcher. The industry appears to be looking more into the modelling of clinical trials to understand further what’s being done and where the big opportunities lie. I also see a growing interest from the regulatory side to document trials more efficiently, and this is where simulation tools can play a significant role. Going forward, more regulatory agencies will look at trials to see whether sponsors have done proper modelling and simulation, and analyse what possible solutions can be drawn before the trial begins. A properly simulated trial helps you in the conclusion phase of a trial. Most trials are powered on the primary outcome and not Severe Adverse Events (SAE). Suppose that you have a trial with clear benefit on the primary outcome, but with more SAE's in the treated groups. What conclusions can you draw from that? Naturally, that depends on the magnitude of the difference, proving this is another area where modelling would provide detailed insights.
*Einar Heiberg is the CTO and Founder of Medviso AB