|
|
| STARK CONTRAST:
In vivo brain images from optical imaging. Meningitis at 19h post
infection.
|
Traditional
survival-based endpoints are fast going out of fashion as a way of measuring
drug efficacy, in favor of in vivo imaging.
There is plenty of
choice among imaging techniques, but no single method does everything the
researcher wants. Moreover it’s hard to find a modality that’s suitable for
preclinical research and that can be carried over to the clinic for human
trials.
“In vivo
imaging in clinical trials has been growing ever since the FDA accepted
surrogate endpoints to prove drug efficacy,” says Mark Weinstein, CEO of
Bio-Imaging Technologies, a “core facility” that performs imaging studies on
behalf of pharmaceutical companies. In oncology for example–—
which represents about half of all drug research today–—
there has been a huge shift away from survival endpoints: “Instead
companies are using time to progression, which is essentially an imaging
endpoint.”
The reason for
this shift is both time and money. “Pick any of the drug studies going on now
and most of them relate to life-threatening or at least life-altering
conditions,” says Ken Faulkner, vice president of business development at
Synarc, another core imaging firm. “The concept of using survival as the
ultimate endpoint in these diseases–—
which means perhaps waiting until you see tumor growth or death–—
is increasingly disagreeable both for patients participating in the trial
and for regulatory boards and ethical committees.”
In vivo
imaging on the other hand provides “up-front” surrogate endpoints as a much
faster measure of efficacy–—
albeit usually in combination with biochemical markers in serum, says
Faulkner. For Phase II and even early Phase III trials this is usually be good
enough, though regulators often still insist on measurement of ultimate outcomes
in the late Phase III pivotal trials leading to approval.
“Imaging
doesn’t necessarily avoid having to do detailed, endpoint studies,” says
Faulkner. “But it sure moves them further down the road in the approval
process.”
Given time,
survival endpoints may themselves not survive: Bio-Imaging’s Weinstein claims
that about 80% of his company’s imaging work is now in pivotal studies. Much
depends on the particular therapeutic area: Imaging is usually a primary
endpoint in rheumatoid arthritis or osteoporosis, where the FDA will typically
require the Phase II work to be done using imaging and markers, and only in
Phase III demanding the “gold standard” of fracture data. In oncology,
though, imaging data could sometimes be merely a strong secondary marker used in
combination with survival and biochemical indicators.
“There is a lot
of talk that taking imaging to drug discovery is going to shorten [the discovery
process], and that has potential, but we need to be careful not to overplay
it,” says Weinstein.
Take Your Pick
So what’s the
favored imaging modality in clinical trials today? Again therapeutic area is
often the deciding factor. In musculoskeletal disease there are already a lot of
known X-ray biomarkers for bone and joint changes, and most of the work is
focused on X-rays and DXA-based bone densitometry says Faulkner.
|
|
| GETTING A
BETTER VIEW: High precision multi-modal co-registration capabilities
combined with simple multi-position imaging have been demonstrated to greatly
enhance localization of molecular signals in live animals. In this example
obtained on a Kodak Image Station, optical signals were generated from tumor
cells in the head area of the animal. Co-registration with X-ray images along
with multiple animal positions provide localization of the optical signals to
specific bones in the head region of the animal (i) Ventral view: Ventral
position imaging shows two optical signal generating metastatic lesions located
in the cranial region of the mouse (ii) Lateral view, right side: Subsequent
lateral position imaging clearly separates the two lesions to specific jaw and
skull locations.
|
In oncology, where
the key marker is tumor size, the key technologies have been high-resolution
anatomical imaging methods such as computed tomography (CT) and magnetic
resonance imaging (MRI). CT scanners are fast and comparatively cheap but poor
at soft tissue discrimination, whereas MRI works well with soft tissue. Both
techniques are favored in cardiology where formation and growth of plaques are
reliable markers of atherosclerotic heart disease, although ultrasound is also
much used because of its ability to measure blood flow.
MRI and CT are
also in favor because they can produce quantitative results, rather than
qualitative images that need to be interpreted by expert radiologists. The
latter can introduce an element of subjectivity that is unwelcome by regulators.
“FDA would like us to move to quantitative endpoints as much as possible,”
says Weinstein. “Advances in resolution will allow us to do that, and as the
standard deviation [i.e. error margin] comes down we may eventually be able to
run clinical studies with fewer patients. But that won’t happen for some time
because the agencies are so risk averse.”
Of course,
technology advances are changing the balance among preferred imaging methods all
the time. Bigger magnets for MRI are improving its resolution to the point where
it can sometimes replace X-rays. In rheumatoid arthritis, for example,
historically the gold standard has been X-ray detection of joint space narrowing
and bone erosion in hands and feet. But recently, Weinstein’s company has done
clinical sub-studies of anti-inflammatory drugs, where high-field MRI was able
to pick up changes in these biomarkers less than six months after therapy began,
rather than the two years usual with X-ray film.
According to
Faulkner, the biggest change is in the shift to functional imaging using
techniques like positron emission tomography (PET) and its close relative SPECT
(single photon emission computed tomography). In PET, patients are injected with
a radionuclide-labeled sugar–—
fluorodeoxyglucose or FDG–—
that is preferentially taken up by cells with high metabolic rate. The
radionuclide atoms then decay, generating high-energy photons. Collecting these
photons into an image highlights regions of metabolic activity rather than just
static structural information.
This is especially
useful in neurological diseases like Alzheimer’s that are challenging to
monitor in any other way. “It’s a wonderful combination of a chemical
biomarker with a medical image,” says Faulkner. “It tells us that the
chemical event is happening but it tells us where too.”
MRI can also
deliver functional images when used with specialized contrast agents that
gravitate to particular tissues, and this sort of functional MRI is on the rise,
says Faulkner. But typically PET and MRI machines cost millions of dollars,
limiting their application to core imaging facilities.
Preclinical
Paradox
What’s true of
the clinic is even truer of in vivo animal research. “The pharma
companies think imaging can to some extent short-circuit the discovery
process,” says Bill McLaughlin, director of R&D at Kodak, which supplies
small-animal optical imaging systems. The company counts among its users
companies like Merck, Bristol-Myers Squibb, Novartis, Aventis, BD Biosciences,
SuperGen, Genentech, and Zymogenetics.
|
|
| PHOTO FUSION:
Simultaneous in vivo imaging with spin echo MRI and 18F-FDG PET in a rat,
showing registered cross-sectional images through the head.
|
|
|
|
| AN ILLUMINATING
GLOW: Simultaneously acquired PET and CT images showing distribution of
18F-fluoride ion in a mouse. |
According to
McLaughlin, pharma researchers now want to take families of leads right into
animals in vivo instead of doing all their lead identification and
optimization in vitro and only then going into animals with the best
candidates.
“They can look
at efficacy, pharmacokinetics and safety testing all at once and get results
within days, as opposed to looking at the endpoints which could be weeks or
months”, McLaughlin says. And of course imaging techniques in animals don’t
have to be approved by the FDA.
Cost reduction is
another driving factor notes Mark Roskey, vice president of reagents and applied
biology at Caliper. Imaging often allows researchers to use fewer animals than
when doing histopathology on every sample and ultimately sacrificing the animal.
“You can image the functional biology over time using the same animals,” he
points out.
Caliper acquired
Xenogen in August and now has a portfolio of in vivo optical imaging
products based on fluorescence and luminescence. Roskey says optical imaging is
still the most important modality for work with small animals. It delivers
enough sensitivity and detection efficiency to see all through the animal.
Moreover, probe toxicity isn’t the touchstone issue in animals that it is in
humans and the variety of fluorescent probes now available means optical imaging
can measure a wide range of biomarkers in animals.
“We can do 3D
tomography, looking at a particular fluorescent probe or a glowing group of
cells deep within a mouse on our standard IVIS platform,” says Roskey. “This
can give a 3D image of the tumor and monitor tumor growth as a biomarker. Or it
can monitor other biomarkers such as a particular protease or cathepsin or some
other enzyme activity that is related to tumor formation.”
Optical imaging is
also relatively cost-effective compared with CT and PET, requiring only a CCD
camera and reagents or probes. “Taking a picture of mice and using software to
quantify the signal is fairly simple for research scientists to do, and it
doesn’t involve million-dollar machines or big magnets,” says Roskey. “And
you can start to quantify as few as 500–
1,000 cells, so for example you can look at very small, early-stage
tumors.”
Biomarker
expertise is the limitation: “At the moment there are probably fewer
imaging-related biomarkers than there are serum-based or other biomarkers,” he
says. “But that’s an issue for biomarkers generally. Increasingly people are
working on developing more of them.”
Subjectivity of
results has been a challenge though software is providing an answer, says Roskey.
“So far it has not been easy to quantify fluorescence in in vivo
imaging because of the effects of autofluorescence in the animal and attenuation
of the signal through the tissue,” he says. “But the development of 3D
imaging and the ability to do spectral deconvolution and transillumination of
the source has allowed us to effectively quantitate fluorescence in small
animals, and is really reducing the subjectivity.”
Kodak’s
McLaughlin stresses that although optical imaging has advantages–—
it is easy and convenient, and the fluorochrome reagents used have long
shelf life and low cost compared with radionuclides–—
it isn’t a panacea. Because it is mostly a dark-field technology where
only the target cells are illuminated, the signal appears on its own, without
any anatomical context. Also, because fluorescence emitted from cells deep
within a mouse is scattered and diffused on its journey to the surface, it
provides only weak positional information: in short, the image is blurred.
“It is very
clear to us that it is not just optical imaging that people want,” McLaughlin
says. “What everybody wants now is multimodality–—
metabolic information combined with an anatomical structure and
contextual information.”
Multimodal
Mission
This makes sense,
especially as multimodal imaging is already widely accepted in the clinic. As
imaging expert Simon Cherry of University
California, Davis, notes, “You often hear people say they’d like an imager
with the spatial resolution of MRI, the temporal resolution of ultrasound, and
the sensitivity of PET.”
That may be
unrealistic but it’s an ideal that can be approximated. The usual approach is
to combine a modality that excels at providing structural information (such as
CT, MRI, and ultrasound) with one that supplies functional or molecular
information (i.e. optical imaging, PET and SPECT). This, says Cherry, allows
enabling function and structure to be examined simultaneously in the same
individual and for the two types of information to be correlated to identify the
exact location of particular cellular events.
In oncology for
example, “hot spots” with suspiciously enhanced uptake of a tumor-seeking
radiotracer can be accurately localized and correlated with anatomy.
PET/CT and SPECT/CT
whole-body imaging systems are now standard in clinical research, says
Bio-Imaging’s Mark Weinstein. “New machines are routinely sold with both
modalities,” he says. “It’s used a lot in early oncology work, because the
big challenge in oncology is to see whether the drug, and maybe its delivery
vehicle, are getting to the right place at the right time and affecting the
malignant cells. With PET/CT you can see both in real time.”
Similar multimodal
approaches are being developed for preclinical imaging in small-animal models.
Kodak has within the last year launched a small animal imaging system combining
three modalities: optical (fluorescence or bioluminescence), radioisotopes, and
X-ray.
The optical or
radioisotope image gives information on molecular interactions–—
binding of antibody to particular receptor or cells, or activation of
smart probes by enzymes–—
while adding X-ray imaging allows more precise location of the signal.
“It puts the target in an anatomical frame of reference,” says McLaughlin.
The system can
also image PET and SPECT isotopes and thus indicate areas of high metabolic
activity like tumors. “Our system gives a 2D planar image rather than 3D, and
it doesn’t give the resolution of typical microPET, but it gives good general
distribution information in the animal,” he says. It also costs under
$100,000, compared with a few million dollars for microPET.
The system uses
specialized phosphor screens–—
one for X-ray imaging, and another one for radioisotope imaging. These
screens slide in and out of the optical plane under the animal without the need
to move anything (movement of the subject is a classic problem in multimodal
imaging). Different images are all taken at the same optical plane by the same
camera and so they “snap together” to make a single multimodal image without
needing any sophisticated image-fusion software, says McLaughlin. “The images
can be overlaid to within 80 µm, giving excellent co-registration of optical
and isotope molecular signals with the X-ray background.”
But this isn’t
ideal, says Cherry. “What’s needed is a single detector system for optical
and PET imaging,” he says. That could allow simultaneous monitoring of the
expression of two different reporter genes, for example. For now, that ideal
remains distant.
(SIDEBAR)
Found in
Translation
A key challenge
facing imaging researchers is the need to translate methods from one stage of
drug discovery to the next.
In animal models,
says UC Davis’ Simon Cherry, there is an urgent need to correlate in vivo
whole-body images with ex vivo images from high-resolution techniques
like histopathology and autoradiography. Researchers need tools for registering
excised biopsy or post-mortem tissue with its exact location in an in vivo
imaging study, so they can correlate cellular ‘micro’ changes with data at
the ‘macro’ level.
Moreover, there is
the usual difficulty in bridging preclinical work to clinical trials.
“We know animal
models don’t always behave like humans. That’s one reason so few drugs make
it through the pipeline,” says Faulkner. “That can happen when using imaging
biomarkers as well and we don’t have a great answer on how to stop it
happening. You can’t put mice into an MRI machine and do the same things to
their small bodies that you could do with a human.”
Toxicity of
contrast agents and probes is a thorny problem for optical imaging in
particular. As a result, says Caliper’s Roskey, there’s no direct
translation of molecular imaging into humans. A few reagents being developed for
small animal imaging can be used in the clinic, including one fluorescent probe
that tracks angiogenesis by binding to VEGF, “but it’s early days yet,” he
says.
An even more
fundamental problem for optical imaging is its limited depth. Light cannot
penetrate more than a few centimeters through tissue, so while it can escape
from deep inside a small animal; the same isn’t true of humans except at the
extremities or near the surface. Moreover, the longer the emission path, the
more interference is present from background fluorescence. All optical imaging
companies are working on this problem.
One route is
tomographic imaging, collecting signal from all around or through the animal.
But the solution most likely to be translatable to humans will be epifluorescent
mode, where the signal is collected on the same side as the incident light.
Kodak is following
two main epifluorescent routes, says McLaughlin. One is multi-spectral imaging,
using optical probes emitting at many different wavelengths. Sophisticated
algorithms can “unmix” the resulting multi-colored signals from one another
and also from the autofluorescence background, boosting the signal-noise ratio
and improving sensitivity and depth.
The second avenue
is nanotechnology. Kodak has designed fluorescent nanoparticles that emit
extremely strong signals but do not contain the toxic heavy metal atoms present
in many current high-brightness fluorophores. This means they can potentially be
used in humans, says McLaughlin. The nanoparticles provide a scaffold for
binding both target-specific probes as well as highly fluorescent moieties. They
also provide for improved distribution through the body, which can be a problem
for probes based on heavy metal ions. Imaging systems using these new
fluorophores will be tried out in pharma companies and top academic labs
“within months,” he says. “Combining brighter fluorophores with
multi-spectral imaging really improves on how far you can see into animals and
humans,” he says.
Maybe Cherry’s
ideal of “spatial resolution of MRI, the temporal resolution of ultrasound,
and the sensitivity of PET” provided with a single detector system is still
distant, but at least some elements of it are in sight.
P.M.
|
