|
It’s the worst surprise
a pharmaceutical company can face: A drug that’s
“clean” in preclinical development suddenly turns
out to be toxic when it reaches human trials — or
even worse, the market. It’s happened too many times
over the last few years, but the recent disastrous
TeGenero trial was a dramatic case in point. Although
the monoclonal antibody being tested appeared harmless
in rats, it caused a massive immune response in the
five volunteers, who ended up in the emergency room
facing organ failure and possible death.
Before that disaster,
most of the attention was focused on problems with
marketed drugs, such as Wyeth’s FenPhen, Bayer’s
Baycol, and Merck’s Vioxx. The situation is dire.
“We have failed to put a dent in the number of drugs
that fail during preclinical safety and human
testing,” said Rakesh Dixit, president of Safety
Sciences Advisors and previously a senior director of
toxicology at Johnson & Johnson, speaking at
CHI’s recent World Pharmaceutical Congress. “We
have to do better.”
The key challenge is
getting animal models that mimic the pathways you’ll
encounter in humans. “You are always wondering
whether an effect in an animal means an effect in
humans,” says Michael Bleavins, head of World Wide
Safety Sciences at Pfizer. “There are many
species-dependent metabolic and toxic effects,”
explains Dixit. As a result, traditional animal
screens sometimes provide misleading data.
Some markers do already
exist: for example, electrocardiograph QT signals for
cardiotoxicity and the transaminase enzymes ALT and
AST for liver toxicity. But these often have poor
specificity, and most arise only after damage has
occurred. As a result, they can’t be used after a
drug is marketed to identify patients at risk of side
effects.
The biggest hope is for
new predictors, such as animal serum troponins, for
example, which could be a promising alternative
biomarker for cardiotoxicity, or gene expression
changes for certain proteins (KIM-1 and clusterin)
that appear highly predictive of kidney necrosis,
according to Federico Goodsaid of the FDA’s Office
of Clinical Pharmacology and Biopharmaceutics.
The number of validated
new markers is small. One reason is that “Actively
searching for preclinical safety signals has been seen
as nobody’s job,” says Paul Rolan, a pharmacology
professor at Australia’s Adelaide University. He
says companies have regarded it as risky to develop
new safety biomarkers, fearing that the FDA would
demand they test every compound against the biomarker
and abandon work on any that fail — even though the
marker might not be a conclusive indicator of toxicity
in the clinic.
Companies are
scrambling to address the problem. Experts see three
big changes that need to take place: New technologies
for biomarker discovery have to be improved, more
mechanistic studies must be done, and companies have
to start pooling data.
Biomarker Discovery
Tools
Many tools are used for
finding biomarkers. Increasing in popularity is
high-content analysis, the automated fluorescent
imaging of cells to look for disruption of
intracellular processes. Such systems are primarily
used in the earliest stages of development,
particularly in gene toxicity screening where the
endpoints are changes in the cell chromosomes and can
thus be imaged. It is also a valuable aid in examining
cell metabolism, for example, to test whether a
compound inhibits mitochondrial function or the
cell’s ability to synthesize proteins.
Microscopic imaging of
histological changes is another increasingly important
source of biomarkers. In May, Bleavins presented an
example to the SCIpharm 2006 conference in Edinburgh,
Scotland, in which laser confocal microscopy was used
to measure changes in epidermal nerve volume in
diabetic rats treated with an aldose reductase
inhibitor. It is very hard to count nerves in ordinary
microscopic images of biopsies because the nerves
branch randomly. But with confocal microscopy, a
computer can construct a 3-D image of tissue by
combining hundreds of separate 2-D “slices”
stacked one upon another. It can thus follow an
individual nerve fiber as it moves through different
image slices, and so measure the volume.
Many new technologies
are being used, including proteomics and metabonomics.
Toxicogenomics has been struggling for recognition for
years and is finally reaching maturity (see
“Toxicogenomics Gaining Ground”).
Abbott, for example,
recently used toxicogenomics to identify the best
backup when the lead compound in a series showed
cardiotoxicity in the rat. The company had only one
lead series available for this project, so it was
important to find a better lead from that series. The
toxicogenomic study required only a small amount of
compound, and the results were available in a single
day. Other compounds were tested in the same model,
and a set of genes was identified that, even at low
doses, predicted cardiotoxicity, which was confirmed
in further in vivo studies. That signature went
into the company’s database and now may be helpful
when they are screening future compounds. At the same
time, they were able to find a better, nontoxic lead
to pursue.
The trick is, “To
develop useful gene expression signatures, one needs
access to a robust database,” says Eric Bloome, a
project leader in Abbott’s Cellular and Molecular
Exploratory Toxicology division.
Getting to
Mechanisms
More information is
desperately needed about what causes toxicity and how
it happens. “Mechanistic studies will become more
important in toxicology,” said Mathias Hukkelhoven,
senior vice president and global head of drug
regulatory affairs at Novartis, when speaking at the
Post-Approval Summit at Harvard University in May.
It’s very difficult
to get that information, of course. “You are trying
to work back from an effect to what caused it. There
could be hundreds of mechanisms that leads to that
endpoint,” says Bleavins. “Since you are dealing
with complex routes, finding the mechanism might be a
matter of looking at the right place and time.”
Sometimes a significant change in a plasma protein
level could have many causes — for example, the
inflammatory response cytokine IL-6. In that case,
says Bleavins, it is more appropriate to look for a
panel of biomarkers, each of which may not be very
specific but taken together can narrow down the
possibilities.
One new approach is to
use pathways analysis, from companies such as
Ingenuity. GlaxoSmithKline, for example, was able to
better understand why a PPAR-alpha agonist was causing
tumors in rats but not in monkeys, by using Ingenuity
Pathways Analysis (IPA). Gene expression analysis
determined that some genes were activated in one
animal, but not the other. “A network was generated
that included the significant genes,” says Deborah
Riley of Ingenuity. Information about known pathways
was then overlaid.
The researchers found
that the JUN, MYC, and NFB
families were downregulated in the monkey while they
were upregulated in the rodent. Other studies have
confirmed that rodents are particularly susceptible to
PPAR-alpha-agonist-induced tumors. IPA can be used to
generate biomarkers, and Ingenuity has developed a
Biomarker Workflow Guide.
Doing experiments that
explain why differences are seen between human and
animal responses to drugs “actually helps regulators
make decisions,” says Dixit. Which is better than
just “throwing out a lot of numbers at them
[FDA].”
Click to enlarge
The Data Dilemma
Everyone realizes that
to get good safety biomarkers, a huge amount of data
has to be collected and mined. “If idiosyncratic
toxicity only occurs in one in a thousand subjects,
you need to process a lot of samples to find a
biomarker to predict which one,” says Rolan.
That’s why so many experts believe that pooling data
is critical to advance preclinical toxicology.
Until recently, that
seemed like a pipe dream, but things are changing
thanks to the intervention of the FDA and its Critical
Path Initiative launched last year. FDA is encouraging
industry and academia to form consortia aimed at
discovering safety biomarkers that can be shared among
several companies. That way no one pharma company has
to risk pioneering a new biomarker and maybe end up
with the FDA’s arrows in its back. “Companies are
recognizing that if they are all in it together they
will all equally benefit or get a disadvantage,”
says Rolan.
In one major coup, the
FDA and The Critical Path Institute (C-Path) recently
announced the Predictive Safety Testing Consortium.
Several companies have already joined, including
Bristol-Myers Squibb, GlaxoSmithKline, Johnson &
Johnson, Novartis, Roche, and Schering-Plough. C-Path
is also sponsoring the Liver Toxicity Biomarker Study
being conducted by the FDA’s National Center for
Toxicology Research and BG Medicine (Waltham, Mass.).
That’s another big win. “Liver toxicity is where
we are failing worst of all,” notes Dixit.
Will biomarkers swoop
in and save preclinical toxicology? Not tomorrow, but
there is great hope that they will help make the field
much more predictive. That has to happen, or the whole
business model crumbles.
Click to enlarge
Additional
Online Content
FDA Tackles Drug Toxicity
By Pete Mitchell
|
