[Editor's Note: The October issue contains an article entitled "Experimental Drugs on Trial" (not free) that discusses the pros and cons of approving experimental drugs for individual use in certain cases. The following is some additional thoughts from Dr. Richard Miller, CEO of Pharmacyclics, a developer of cancer drugs who has had his own struggles with the issue.]
When I read Beryl Lieff Benderly's article entitled "Experimental Drugs on Trial" (not free) about the Abigail Alliance lawsuit, a topic I've been following for some time, I was glad to see it included a broad discussion of clinical trial methods and policies.
Much of the current debate on access to experimental drugs has been dominated by two points of view:
Individual dying patients have a constitutional right to early access
Drugs should be made available only after they have been proven to be safe and effective in definitive controlled clinical trials
Unfortunately, each of these extremes fails to address the most important issue in this debate: the need to rethink and reengineer clinical research methods and FDA policy for drug approvals.
Patients, and now many prominent researchers, are increasingly calling for ways to conduct and evaluate clinical research that are more efficient and more relevant, especially for individuals with deadly or incapacitating diseases. It makes little sense to get stuck on outdated clinical designs and statistical approaches when the science and the practice of medicine have advanced far beyond the state of knowledge that existed when many of these approaches were first introduced.
For example, as it stands now, the FDA too frequently applies a one-size-fits all approach to drug approval. What is needed is a context based approach to drug approval where the data, severity of the illness, safety, availability of alternate therapies and other factors are considered together in context. This is the very concept behind the growing trend in medicine to personalize treatments that best fit the biologic and clinical needs of a patient.
Benderly's article eloquently concludes, and concurs, that the rules governing clinical research and analysis may need to be reengineered--a point that the medical and scientific communities are slowly beginning to speak out about as well.
In a recent JAMA article (need subscription), statisticians Kent and Hayward wrote that "the results of clinical trials might not apply in a straightforward way to individual patients, even those within the trial." They admonish current statistical analyses widely used in clinical research because they only describe average effects of a treatment and fail to address the needs of the entire distribution of patients. Even though on average there is a benefit, some patients may benefit and some may be harmed by a therapy.
Kent and Hayward propose the use of a risk-stratified analyses looking at multiple variables so that we can best identify which patients will benefit from a treatment. Although intuitively obvious in many ways, this is a procedure that would be currently frowned upon by FDA, which usually requires that clinical trials meet a single pre-defined endpoint. This position is somewhat incomprehensible given that the "average" patient doesn't really exist; he or she is a hypothetical construct. In medical practice, rather than making a treatment decision based on the average results, it would be more helpful to be able to determine how patients with specific characteristics had performed in a trial.
Similarly, in a recent editorial in the Journal of Clinical Oncology, Sargent and Grothey suggested that "in the absence of randomized data on a certain issue, retrospective, exploratory analyses, with all their limitations, are preferable to obtain information for a clinical decision, rather than refusing to offer any opinion, provided that these analyses are accompanied by a full discussion of limitations." Current FDA policies prohibit use of any post-hoc or subset analyses regardless of the degree of biologic and clinical plausibility, and patient need.
Some of these issues were confronted by AIDS patients who led the charge to make changes in drug approval regulations, such as accelerated approval. Unfortunately, accelerated approval (pdf) has been whittled away by FDA so that it has lost its original intent, which was to allow faster access to promising new drugs.
The FDA's overly cautious approach can be felt elsewhere--from physicians baffled by new warnings against generics that have been prescribed for decades to biotechs and industry observers shaking their heads at surprising delays of drugs that have shown strong evidence of efficacy and safety. Some physicians are even laughing out loud in disbelief by the delays of some approvals of cancer treatments for indications which they have been prescribing off-label for years [hat tip to Kevin MD].
To progress beyond the dueling perceptions that the FDA is lax on clinical trials or that the FDA is too conservative, we need an FDA that has adequate funding, and internal and external resources to review drug approval data completely and in proper context where the risk versus reward relationship can be thoroughly evaluated.
We need a culture at FDA where the regulators can assume appropriate risk in permitting drug approvals based on the use of novel endpoints and clinical trial designs for those patients who can assume more risks and who need faster access to therapies.
And we need to make sure that physicians understand the potential limitations of these data.
Whether we use Bayesian analyses, adaptive trial designs or surrogate endpoints, for the public and many researchers, this debate is not really about permitting access to experimental drugs after phase 1, 2 or 3 trials; it's a challenge to the existing biomedical and clinical research system to exhibit a more contemporary scientific understanding in the procedures and methods used to evaluate drugs.
As Perry Cohen, founder and director of the Parkinson's Pipeline Project, was quoted in Benderly's article, on the whole there is still a need for "scientists to realize that there's a crisis of confidence" in clinical trials among the public, which is why I am very grateful to SciAm for the opportunity to blog about the topic here--and looking forward to hearing what others have to say.