Problems in the structure of Clinical Trial

It's a great talk.   I agree that the status quo in BC clinical trial is not acceptable.   Each semi-successful drug takes 15 years from bench to approval, 100s million dollars, not to mention the failed drugs.    Combination drugs may well help, (Herceptin, Everolimus, Pertuzumab all succeed in combination treatment) but there is a big problem with mathematics.

Usually a good treatment (not a specutacularly good 70-100% response treatment, maybe a 30-50% response treatment) takes 500 patients in a randomized blinded controlled trial to get statistically significant proof that it actually is good, not just a statistical fluke, a placebo.   I explained this in another thread:  A clinical trial on 23 patients is like 23 coin tosses. After 23 coin tosses, say I get 15 heads vs 8 tails. I could say that this coin is weighted heads/tails ratio almost 2, that's very different from 1. But I am not very confident of this conclusion because the sample size is too small. Now if I toss 500 times, and still gets heads/tail ratio of 2, then I become very confident of this ratio.

That's why randomized double blinded clinical trial with 500+ patients is still the gold standard in clinical trial. That's expensive and requires lengthy recruiting. When you consider that different subtypes of BC (TNBC, claudin-low, normal-like, basal) may have only 2%-10% patient population, you realize how tough it would be to find 500 patients with this small subtype to test a drug specifically targetting this type. Add to that cancers evolve to become drug resistant, it become even tougher. 

Now, let's move on to the even more difficult tasks of testing combination treatment.  Suppose we have 3 drugs (A,B,C) that has cleared phase II trial with good results, each requires 500 patients to go through phase III trial for approval, that's 1500 patients required total.   Now if we want to explore combination treatment, then there are at least 4 combos, AB, AC, BC, ABC.   That would require 2000 patients for a full test.

Suppose we have 4 drugs (A, B, C, D), there will be 15 combos+singles to fully test, requiring 15*500.   Suppose we have 5 drugs, we will need 31*500 patients.

It's exponential, doubling with every new promising drug to test and we do not have this many patients.     But "In clinical trials, 97% of patients in Europe are not in a trial", I think numbers are similar in US.   That is a HUGE problem.

Traditional clinical trials especially on early stagers require long time commitment (like 5 years) from patients. So traditionally late stagers are the main force of clinical trials.  Late stagers are fewer in numbers (50000/year), the ones with the right subtype of BC rarer (say 5000/year for TNBC, which has more subtypes), the ones with good performance statuses are fewer still (say 2000/year), the ones with good performance statuses living near a research facility are rarer yet (say 1000/year), the ones with good performance status who are willing to serve in clinical trials are still fewer(I don't know how many, but certainly less than 50% from my observation), and so it's hard to recruit enough of them to test new drugs, much less combinations. End result is: traditional clinical trial process takes 15 years and 1 billion dollars for any good drug to make through.

This is why I'm so excited by ispy-2.org. Newly dxed early stagers needs to commit only months to either control or experimental arm. Basically a couple of dose to see if it (standard chemo or standard+new drugs) works, if not, they change chemo or go ahead to surgery. The new drugs are generally the most promising drug that has already been tested for safety. 200k/year newly dxed patients with good performance status could potentially be recruited which means quick recruitment. A new drug if demonstrating efficacy, could go from bench to approval in as little as 5-10 years. Even FDA is excited.

If Henry Ford is on Titanic, he would not say: let's do a handmade lifeboat for everyone. He'd make an assembly line that churns out lifeboats. Ispy2 is the kind of assembly line for new cancer drugs that he'd be proud of.

ispy-2 has potential to increase patient participation dramatically by reducing patient commitment.   That said it only could test certain type of drugs.   In general, we need much greater patient participation across the board, and we need a different clinical trial process (phase I, II, III may need to be changed),  we need to invest in better animal model/preclinical model, better blood/imaging tests for progression, different end-point, otherwise, we will not get the MBC cure.    

If a targeted therapy (whether single or combo) is very specutacularly successful (90% response), then maybe 100 patients is enough to be statistically significant.   FDA already expedite approval for these kinds of therapies, no change is needed.  Look at Xykori for example.

However, a less successful therapy (whether single or combo) is NOT going to be able to get statistically significant response on 100 patients.    So, they need to do better to either: a) run trials faster like ispy2.net  b)  do better clinical design to watch the patients more extensively and do better controlled matching of the patients, c) get FDA to approve based on alternative end points

Notice how currently end points is fuzzy and make the statistically significance criteria harder to reach.   Radiological progression is often different if looked on by different radiologist (imagine the tumor had an area of necrosis, or size decreased but SUV increased on PET, or liver tumor decreased but 1 more bone tumors, do you call this progression?) , not to mention expensive.   Overall survival takes long time to reach, and it also sometimes means that researcher have to tell the controlled arm patients to "Drop Dead" when cross-over on progression should be allowed to save patients' life.    

Basic science needs massive investment:  better animal models, better understanding of mechanism of every thing.   Nonradiological technologies to track tumor growth and correlate with OS should be allowed as alternative end point, think CTCs.   Clinical trials like ispy-2 should be expanded to reduce time to market.   Combos needs to be explored, but trial design need to be better otherwise you get a whole bunch of failed experiments due to statistical insignificance.   New patients need to be extensively recruited for clinical trials, existing patients need to join more clinical trials.   That's my cato bit.