|
Drug development looks increasingly like an ecosystem
on the verge of collapse. It evolved in the good old days of plentiful low hanging fruit, at a time when patients
and doctors felt blessed to receive each new miracle drug, and third-party payers saw salvation in cheap
drugs replacing expensive medical procedures and hospital stays. Drug companies were saviors. Profits were
strong. Jobs were secure. Unfortunately, traits, strategies, and relationships that evolved under conditions of
plenty stop working when times get tough. And that’s what the drug industry is facing now.
In the natural world, failed ecosystems simply disappear
to be replaced by completely new ones. Many people think that could happen in this case. But I believe
that individual players can still escape and prosper by anticipating the new order. Punctuated
equilibrium, the mammalian radiation, railroads in the 1800s, automobiles in the
1900s, Microsoft and Google, all make the point: If you give up the past, you can
embrace the future.
So what should we be doing? Pharma companies themselves are already looking outside the proverbial
“box” for new solutions. Here we can talk freely about even radical ideas. I want to use this column to brainstorm
with you, the reader, about solutions. What would you do if you were the boss? If the FDA weren’t a
bottleneck, stockholders could see beyond the next quarter, politicians could look healthcare costs in the
face without ducking, and plaintiff lawyers weren’t such vultures? Suppose the only thing that mattered was the
science, the technology, and of course, the patients.
Since I’m the writer, I get to go first. Please write
back and tell me where I’m wrong. Perhaps together we can make some progress.
The Trouble with Clinical Trials
Let’s start with one of my hot buttons — the irrational
process we call “clinical trials.”
You know the standard design: Take a bunch of people
who are sick, give some of them a new drug (which may or may not work), give the others an old drug or
placebo, then sit back and wait to see who gets better. When the study ends, you hand the data to a bunch of
statisticians who do their magic and declare the new drug a success if the computed benefit reaches the
threshold of 95 percent significance.
The details strain credibility. Most trials do not enroll
all comers. The old, the young, those who are too sick, and those who are too healthy are all excluded.
The goal is to enroll perfect patients who are sick enough to show benefit but have no extraneous risk
factors that may complicate the results. This bears no relationship to the real world of clinical medicine,
where every human being is different and reacts differently to disease and treatment. The fact that on average
a drug helps or does not help a group of perfect patients may say little about whether it will help any specific,
imperfect patient.
A second problem is that statistical rigor is almost
always compromised by dropouts. You’ll see the throwaway line in many studies about handling dropouts using
the “intent to treat principle.” Sounds good, but it doesn’t address the real problem, which is missing data.
When a person leaves a study, there is no way to estimate what the missing data would be and no justification
for assuming that dropouts from both arms will have the same properties. There’s plenty of practical experience
showing that studies have lots of dropouts, and this has a huge effect on study outcomes.
A final problem is the short duration of most clinical
trials. For obvious practical reasons, trials have to be quick. It’s equally obvious that short trials cannot assess
long-term benefit and risk.While the FDA will increasingly mandate post-market surveillance to check
for long-term effects, this task has traditionally fallen to underfunded academic epidemiologists.
Post-Vioxx, of
course, that could change.
Historical Context
The early literature on clinical trials sheds light on how
this strange system came to be. The current paradigm emerged in the 1930s and ’40s in response to the very ad
hoc method then used to evaluate treatment options. In those days, new treatments were introduced by
strongwilled, famous clinicians on the basis of personal experience. This was unsatisfactory to many investigators
who felt that the scientific method should be applied instead with controlled studies, clear hypotheses, and rigorous
statistical analysis.
The new approach was certainly a step forward, but
in the rush to put clinical research on the same footing as basic science, important issues were swept under the
rug. People are not bacteria or even mice and cannot be studied using the same methods. People are a lot more
variable than model organisms and resist efforts to control their genetic background or environment. They
are also more persnickety, harder to recruit, less compliant, and more likely to drop out. But the most significant
problems are ethical. The health, welfare, and autonomy of human subjects must be given priority
even when this compromises the study design. Clinical trials may bear some resemblance to controlled experiments,
but they are not experiments and cannot be treated as such.
Despite these problems, the current genre of clinical
trials worked great in the old days with drugs that had whopping big effects, compliant patients, and less vigorous
attorneys. Those days are gone, and we need to devise new kinds of trials for the new era.
One Possible Solution
Let’s step back and ask what the real purpose of a clinical
trial should be.
My bedrock principle is that patients come first. It’s
not just because this is morally right, but because customer needs always drive the formation of new markets
and industries.
Patients want drugs that treat their disease with an
acceptable level of risk and at a price they can afford. A person suffering from a serious, untreatable illness will
rationally accept greater risk and cost. Someone dying of cancer, who has failed all approved treatments, may
be willing to accept an experimental drug with no evidence of efficacy — indeed that’s how cancer drugs are
tested. Someone with muscle aches and pains probably isn’t willing to take a drug that increases risk of heart
attack. It’s commonsense.
When you look at it this way, the current 95 percent
threshold for efficacy makes no sense at all. Surely, the correct efficacy threshold must vary as a function of risk
and benefit and may be a matter of personal judgment.
I work with patients suffering from a progressive and
ultimately fatal neurodegenerative condition — Huntington’s disease — for which no approved treatment
yet exists. I see no rationale for forcing these people to wait for drugs that are 95 percent certain to work when
without treatment they have a 100 percent chance of cognitive decline, physical disability, and early death.
I also work with patients who have type 1 diabetes,
the juvenile onset form of this disease, which requires lifelong insulin injections. This is an extremely serious
illness with a high rate of devastating complications (kidney failure, blindness, and amputation among
them) and risk of early death. Still, despite the pain of frequent blood tests and insulin shots, this community
seems to have little desire for unproven treatments.
The question is not whether a drug works in some abstract sense, but how likely it is to work for a given
patient. We need clinical trials that help doctors know when to start a treatment and when to stop it, how to
tell if the drug is working and what side effects it’s triggering.
For chronic diseases like Huntington’s and type 1 diabetes,
long-term individualized trials — sometimes called n-of-1 trials — are a plausible solution. In this
scheme, there is no arbitrary line between clinical trial and practice and no arbitrary declaration that a drug
works or not. The drug is simply used under appropriate supervision with patients monitored for benefit and
risk. If the treatment works for many patients, its use will grow, and the necessary level of supervision will be
adjusted to reflect growing knowledge of its safety profile. This has huge implications for the pharmaceutical
business model. Is it still a feasible approach?
I don’t claim that such methods are fully worked out,
but they offer a direction for the new ecosystem.
That’s my opinion. Now tell me yours.
Read and respond to Nat Goodman’s column at www.Goodman.PharmaDD.com.
|
