Here's a new editorial on clinical trials and drug development by Tomas Philipson and Andy von Eschenbach (former head of the FDA). It continues his earlier theme of scaling back Phase III trials (which I commented on here).
These Phase 3 clinical trials served us well in the past. Today, in an era of precision or personalized-drug development, when medicines increasingly work for very specific patient groups, the system may be causing more harm than good for several reasons.
First, because of their restrictive design and the way the FDA interprets their results, Phase 3 trials often fail to recognize the unique benefits that medicines can offer to smaller groups of patients than those required in trials.
Second, information technologies have created improvements in our ability to monitor and improve product performance and safety after medicines are approved for sale. Post-market surveillance can and should reduce dependence on pre-market drug screening in Phase 3 trials.
Third, reducing reliance on Phase 3 trials is unlikely to introduce an offsetting harm induced by more dangerous drugs, since evidence supporting safety is produced in earlier phases. Manufacturers also have powerful incentives to maintain drug safety, since they take enormous financial hits -- well beyond the loss of sales -- when drugs are withdrawn after approval.
I'm still of two minds about this proposal. The idea of moving to less preclinical study and more post-marketing surveillance is not a ridiculous one, but our current system (and the expectations it generates) do make a good fit with it. The nasty details I noticed being glossed over earlier are still with us: how will health insurance companies deal with this change? How do we keep unscrupulous gaming of the system, with companies rushing things to market and spinning out the postmarketing studies as thinly and cheaply as possible? What would keep the real bottom-of-the-barrel types from pumping out high-priced placebos for demanding diseases like Alzheimer's, which compounds would fly through safety studies and reap big profits until they (slowly) were proved ineffective? What would be the legal aspect of all this - that is, when would a patient have the right to sue if something goes badly wrong, and when would they have to just realize that they're taking an investigational drug and that they're part of a research study?
These are real problems, but you wouldn't imagine that they even exist when you read these editorial pieces. I'm a fairly libertarian guy, but these are the sorts of things that occur to me within the first few minutes of thinking about such proposals, which means that there must be many other wrinkles I haven't thought of yet. I agree that increasing the research productivity of the drug industry would be an excellent thing, but I'm really not sure that this is the way to do it.
College chemistry, 1983
The 2002 Model
After 10 years of blogging. . .
1. Jane Richards on March 4, 2013 9:46 AM writes...
Thanks for the article. For info on people using voluntary Libertarian tools on similar and other issues worldwide, please see the non-partisan Libertarian International Organization @ http://www.Libertarian-International.org ....
A good solution is wider use of private insured laboratories to set standards for which they're financially responsible.
Permalink to Comment2. SteveM on March 4, 2013 9:50 AM writes...
Re: "Today, in an era of precision or personalized-drug development, when medicines increasingly work for very specific patient groups, the system may be causing more harm than good for several reasons."
That appears to be the drug development business nut. Highly selective drug and biologic therapies with perhaps only marginal clinical value targeting very sick patients at a quarter million bucks a year. Subverting Phase 3 trials makes sense for desperate patients.
It will be interesting to see how the payers respond when there are scores of those drugs on the market with highly politicized demand (e.g. Avastin) for coverage.
Permalink to Comment3. RB Woodweird on March 4, 2013 10:15 AM writes...
Unfortunately for the pharmaceutical industry, "post-marketing surveillance" will involve ten times as many tort lawyers as clinicians.
Permalink to Comment4. weirdo on March 4, 2013 10:16 AM writes...
I think you're missing a "not" in the second sentence of the second to last paragraph: ". . . do not make a good fit . . ."
Permalink to Comment5. noko marie on March 4, 2013 11:01 AM writes...
"unscrupulous gaming of the system, with companies rushing things to market and spinning out the postmarketing studies as thinly and cheaply as possible"
Sounds like the nutriceutical industry...
Permalink to Comment6. Algirdas on March 4, 2013 2:45 PM writes...
"First, because of their restrictive design and the way the FDA interprets their results, Phase 3 trials often fail to recognize the unique benefits that medicines can offer to smaller groups of patients than those required in trials."
Can someone with more knowledge on phase 3 trials expand on this? Because I am looking at the following phase 3 study of VX-770:
http://clinicaltrials.gov/ct2/show/results/NCT00909727?term=cystic+fibrosis+vertex&phase=2&rank=2
It had 26 kids on the drug and 26 on a placebo. Does not get much smaller than this, does it? And yet, this was a phase 3 trial.
The small numbers are not surprising - the drug targets a single point mutation in a single-gene disease; the entire patient population is indeed very small. But the very fact of a phase 3 trial with 52 participants seems to contradict the claim in the original editorial as quoted by Derek. What am I missing? Are CF-related clinical trials an exception?
Permalink to Comment7. Anonymous on March 4, 2013 3:04 PM writes...
It seems to me If you don't understand the disease fully, who knows what exactly is going on in the phase III patient population? Is it any surprise those clean compounds don't work? Statisticians seem much more comfortable adding expensive n numbers to tease this out, but not in finding a way to make sure all the patients truly have the same disease.
Permalink to Comment8. hibob on March 4, 2013 7:48 PM writes...
@6: Yup. Phase 3 trials often fail to recognize the unique benefits that medicines can offer to smaller groups of patients because phase 3 trials are usually designed to capture as large a customer base as possible for the drug.
As has been pointed out in this blog in previous years, drug companies have the option of designing clinical trials for narrow groups of patients or (@7) trials that require specific diagnotic test results that preclude 70 or 80% of the patient population from the get-go... but that would preclude 70 or 80% of the profits as well.
The aim is to cast as wide a net as possible while still getting approval.
Permalink to Comment9. Cellbio on March 5, 2013 10:02 AM writes...
The opinion expressed in the editorial takes me back a few years. After many trials with no efficacy in the total population, people sit around and cut the population into "chunks" (sickest, those with marker X, fattest, etc.) and provide rationale for another trial preselecting the responder sub-population. New trial, same failure. Conclusion? Wrong marker, let's make a Personalized Medicine group, spend millions and years at finding the right marker. No progress, ahhh, let's make a PRECISION Medicine group. Do it all over again.
And is the idea running PhIII in just the population that would respond, or inferring response in a subpopulation in smaller studies? If the latter, this only makes the association with a subpopulation more tenuous (or statistical fluke, or worse, invisible). If the former, really nice idea, fits well with the other idea offered by consulting types, namely, only advancing the molecules that will work.
In my most in-depth experience in trial data analysis for sub-population identification, I noticed that there was no responder "group", but rather a movement of the disease score mean of the treated population. Unfortunately, that realization did not influence the hunt for markers of response.
Looking at the population response as a distribution is simple, and I assume it is done in fields I have not followed, but try to find clear data of specific patient groups that respond other than placebo vs. treated group differences (means) and inference of sub-populations. A ten percent delta over placebo does not mean 10% of the population responded.
I find the idea of precision medicine attractive from simple logistical and biological perspectives. However, my experiences (certainly narrow) tell me the hype is ahead of the ability to deliver, and the work requires more clinical trials rather than less. One would start without great knowledge of the sub-population, derive a hypothesis through trials, re-affirm with another trial. if this process could be supported with the idea of early marketing with post-marketing surveillance, that might help, but like Derek, it makes me a bit uncomfortable to allow a drug to be sold prior to solid proof of risk/benefit ratio.
To this last point, I think they are absolutely wrong that safety has been addressed prior to Ph3. You really need the numbers and time of exposure to reach a full understanding of risk (which is why we have post-marketing surveillance).
Permalink to Comment10. Jonathan on March 6, 2013 11:03 AM writes...
I think your question about how insurance companies will react can already be seen when it comes to diagnostics and things like exome or genome sequencing - they're quite open about not being at all interested in paying for evidence generation.
Permalink to Comment