Narrative Adapted by: Facundo Zaffaroni, Statistician; Director, Business Development – SME in Biostatistics & AI
Original Authors of Technical Blog: Marc Buyse, ScD – Chief Scientific Officer, IDDI and Everardo Saad, MD – Medical Director, IDDI
Remember your first childhood crush? I do. I was just a small kid living in Uruguay, but my heart raced every time I saw her. Then I got the courage to say hi… and she said it back! We gave each other flowers and drawings with lots of hearts and sent love letters (well, our parents did, since we were still learning to write). The butterflies, the certainty, the joy… It made me think “She’ll become my wife, and then we’ll live together forever”. Such an amazing feeling… and you’ve probably felt that too for someone. But then we grow up. We learn more. And over time, that perfect match didn’t seem so perfect anymore (and we broke up).
Clinical research works in a similar way.
Most sponsors fall in love with their successful phase 2 results. They see a positive signal and feel ready to move to a phase 3 clinical trial and prove that they have something that can be valuable to patients, regulators, and the world. It is easy to imagine the excitement.
However, just like the first childhood crush, the data shows just how few candidates move from a successful phase 2 trial to a successful phase 3 clinical trial. In fact, only 32% of oncology drugs that move from phase 2 into phase 3 succeed. That is even lower than the 41% success rate for non-oncology indications.
So why do so many phase 3 clinical trials fail?
One big reason: phase 2 trials often make treatments look better than they are.
This does not mean that there were errors in the phase 2 data analysis. It is because the early picture was incomplete.
Small trials sometimes show very strong results, but these results are often outliers. In other words, they are exceptions, not the kind of result you would expect if the same trial were run again in a larger group. That is why phase 3 trials are needed.
When the larger phase 3 trial happens, the result becomes more representative of what can be expected in a broader patient population. As a consequence, results are more realistic and not as exciting as the phase 2 result.
Statisticians call this “regression to the mean”. This is a fancy (and sometimes pessimistic) way to say that when something looks great (or awful), it’s probably too good (or too bad) to be true, so it is probably not real.
Think of this as like when a kid has an amazing day playing the sport they love, and a parent thinks, “Wow! That was amazing! My kid is awesome!” but the next week, they are back to normal, at their usual performance level – still good, but not outstanding.
Even if the treatment is truly effective, the best phase 3 adaptive design clinical trial is built, and then supported by a skilled biostatistics programming team, the results might not be as strong as those from the phase 2 trial.
In addition, there are other potential reasons for phase 3 clinical trials to fail:
- The outcome measured in phase 3 may be different (most likely, it will)
- The patient population may be different from phase 2
- Research sites may not follow the protocol exactly
- The data could be disorganized, inconsistent, or incomplete
- The study may have included too few patients (by design, or through recruitment challenges)
- And sometimes, results are cherry-picked (meaning, only the best-looking ones are highlighted)
All these factors add up.
Then, suddenly, one wonders why a phase 3 trial doesn’t deliver as expected…
What does this mean for you, as a sponsor?
- If you are designing a phase 3 trial based on phase 2 results that look too good to be true… they probably are, unfortunately.
- Before starting a phase 3 trial, revisit your assumptions. Surround yourself with partners who are experts in their fields and ask questions to qualify their experience.
- If you are feeling rushed and excited about your phase 2 results, take a moment to step back. Consider your next moves carefully, identify the best strategy, and only then proceed.
- Keep in mind that only 32% of phase 3 oncology and 41% of non-oncology clinical trials yield “positive” results.
- Don’t fall for early data just because it makes your heart race, but do remain hopeful of finding love!
If you would like to learn how IDDI’s team of biostatistics experts can help you design a phase 3 trial that avoids false expectations, sets realistic assumptions, and delivers credible results that regulators can trust, get in touch.