Cerebrum Article

Transforming Drug Development Through Brain Imaging

After years of research and enormous expense, perhaps only one in a hundred potential drugs for a brain disorder will receive government approval and make it to the person who needs it. Neuroimaging may change this dramatically.

Published: November 15, 2006

According to the eighteenth-century French philosopher Voltaire, “The art of medicine consists in amusing the patient while nature cures the disease.” Even as late as the beginning of the early twentieth century, the great Oxford physician Sir William Osler wrote, “One of the first duties of the physician is to educate the masses not to take medicine.”

The subsequent years of the twentieth century saw a transformation of these pessimistic views. The discovery of penicillin was the most dramatic example of a “magic bullet” for acute, life-threatening infectious disease. However, although the stories are less well known, neurology paved the way for effective drug therapies for chronic disease. Success in alleviating the symptoms of neurological conditions has come in the use of drugs for treatment of epilepsy, Parkinson’s disease, migraine, pain, multiple sclerosis, amyotrophic lateral sclerosis, and even Alzheimer’s disease. With effective treatments to offer, neurologists have become more than simply spectators to the disorders of brain and mind. They are able to intervene with a scientifically informed expectation that the course of a disease can be changed.

But more than half a century since the modern age of pharmaceuticals was ushered in with the “magic bullet” of penicillin, no major affliction of the brain and mind has a cure. Many, such as dementia, depression, schizophrenia, and stroke, remain without satisfactory treatment. As our population ages, the burden of disorders of later life, frequently neurological, becomes ever greater. At least one in four of us will develop Alzheimer’s disease if we live to the age of 85. Stroke is the third leading cause of death and the major cause of chronic later-life disability in the United States. As our population grows overall, the cost to society in lost work days and lost human potential will increase.

What is keeping us from finding powerful new drugs to relieve patients with these major disorders? Most people in the pharmaceutical industry now estimate that typically $1 billion or more must be spent to develop a compound, test it in the laboratory and then in clinical trials, and finally obtain approval for it as a new drug.  Development of a new drug typically takes fifteen or more years from identification of a promising target to approval and sale of the drug. (To learn more about this arduous process, read The Long, Sometimes Bumpy Road of Drug Development.)  The number of hurdles that must be met is so great that, in recent history, only one out of ten compounds that entered initial human trials make it to market. For molecules targeting diseases of the brain, the record with conventional approaches has been even less encouraging, with perhaps only one out of 100 chemicals proposed as potential drugs receiving approval for sale. Moreover, many of the diseases for which treatments are now being sought have chronic courses too long to fit into conventional drug development timelines. Many are complex disorders involving interactions between several genetic, developmental, and environmental factors, so that drugs targeting any one process may be expected to have only a partial effect on the overall course of the disease.

The Potential of Brain Imaging

To help experimental medicine drive the development of drugs in humans faster and with greater confidence, one especially promising area for innovation lies in the use of new technology. Brain imaging methods that allow scientists to watch the living brain in action, non-invasively, are among the most promising of the new technologies. Used in exploratory ways, new imaging methods can better track the activity of diseases, providing more-sensitive measures of patient characteristics than is possible with usual clinical observations. Researchers can directly probe molecular interactions by using imaging to observe responses in patients’ brains, which will permit better selection of the right dose of experimental drugs. Information can be enhanced, and the effect of the drug being studied can begin to be predicted by measuring drug effects on brain functions that are relevant to the disease.

Modern imaging that allows scientists to watch the functioning brain now relies primarily on two approaches. Magnetic resonance imaging (MRI) uses harmless magnetic fields and radio waves to map brain structures and brain physiology. Positron emission tomography (PET) uses safe, tracer doses of radioactive materials to follow the fate of individual molecules as they travel in the human body. These techniques promise to change the way drugs are developed. Future drug development will spend less time studying “models” of disease in animals and move quickly to more-informative experimental medicine in humans. The benefits of using neuroimaging in this way can already be glimpsed in six aspects of drug testing: time and cost, confidence in targets, integration of information, dosage, drug combination, and understanding of the placebo effect. The FDA and other regulatory agencies are encouraging efforts to develop and validate these kinds of approaches.

Smaller, Faster Clinical Trials

Imaging tools can enhance the sensitivity of measurements of patient responses to individual drugs. Consider the example of Alzheimer’s disease. Because the clinical symptoms progress slowly, researchers using conventional clinical measures of the disease must study either very large numbers of patients or a smaller group over very long periods of time in order to have significant results. Both alternatives make clinical trials expensive and slow to complete.

However, elegant studies from the laboratory of Eric Reiman, M.D., in Arizona, among others, have shown that even early stages of Alzheimer’s disease are associated with reduction in the brain’s use of glucose, a natural fuel for the brain.1 More surprisingly, Reiman and colleagues have demonstrated that reductions in glucose utilization seen in a PET scan may precede noticeable symptoms of Alzheimer’s disease  even by decades and that these reductions can be measured with high precision.2 Calculations suggest that use of a PET outcome measure could decrease the number of patients required for early clinical trials by more than tenfold. While still expensive, PET scanners that can track glucose utilization in the brain are now widely available in major medical centers, making this a feasible approach for new trials.

Loss of brain volume is associated with the degeneration of nerve cells in Alzheimer’s. What is remarkable now is that using MRI scans, which provide extremely sensitive measures of brain structure, small changes in brain volume can be measured with extraordinarily high precision: out of the approximately one-and-a-third-quart brain, changes in volume as small as half a teaspoon can be seen. This allows the brain atrophy associated with Alzheimer’s disease to be plotted over periods as short as six months, using populations much smaller than those needed if researchers look only at conventional clinical outcomes.3

Increasing Confidence in Targets

With MRI measures, early clinical trials for some types of Alzheimer’s disease treatments could become more efficient, but how can possible new targets identified with genetic searches or other methods be validated as likely prospects? No simple, general strategy exists. One popular approach is to use transgenic technology to create animals that make either too much or too little of the gene product in question, looking for a change that would correspond to one associated with the human disease. But how does one assess dementia in a mouse? Poorly!

A promising new approach is to find specific measures of activity in the human brain that are informative about malfunctions related to a disease and then to look for changes in these brain functions that are associated with the slightly different forms of genes that are found between people.4 This approach uses normal human variation to identify mechanisms of disease.

The laboratory of Daniel R. Weinberger, M.D., of the National Institutes of Health, has pioneered this type of approach for psychiatric diseases. In an illustrative study, his group used functional magnetic resonance imaging (fMRI) to study brain signals in people who had taken the stimulant drug amphetamine.5 They learned that differences in the signals were associated with small differences in the structure of an enzyme responsible for inactivating neurotransmitters in the brain (COMT, or catechol-O-methyl transferase). Paradoxically, rather than improving thinking and memory, subjects with the more rare form of COMT experienced impaired performance after taking amphetamine. The use of modern imaging to relate genetic differences to drug response differences between people provides an efficient approach to pharmacogenetics—the tailoring of drugs to people who are more likely to respond well to them based on their personal genotype. This type of study also provides clues to genetic features that may predispose a person to addictive behavior or depression.

Integrating Information

Combining information from more than one imaging tool can provide further insights by relating changes at a molecular level to differences in the way large systems in the brain function. For example, researchers have directed considerable effort toward developing drugs that may help to reduce the liability to or severity of addictions, such as addiction to alcohol. The neurotransmitter dopamine (the same small molecule that when reduced in movement areas of the brain causes voluntary motion to freeze in Parkinson’s disease) is also implicated in this disorder of motivation.

Work by a research group in Cologne, Germany, illustrates this particular benefit of neuroimaging.6 They took a group of alcoholics, who had been abstinent from alcohol, and assessed the craving that they expressed when shown attractively displayed pictures of beer. As expected, the abstinent drinkers all expressed a desire for the beer (which they suppressed), but different individuals had different degrees of craving. The investigators then measured the concentration of the large molecule in the brain that allows dopamine to signal to neurons (or more precisely, to bind to specific receptors on the neurons) in a region known as the ventral striatum, which is a key element in the brain circuit controlling motivation. They found that the higher the craving, the lower the amount of the receptor that was able to bind a dopamine-like PET tracer. This demonstrated an association between abnormalities of dopamine and malfunction of its receptor.

The German group went on to use fMRI to probe the relationship of this specific biological effect to broader physiological responses in the brain. By alternately presenting pictures of beer or neutral pictures, they identified differences in brain activity in response to the alcohol cue, finding that these differences highlighted the same parts of the brain identified in the PET experiment. The researchers also observed a direct relationship between the magnitude of the fMRI signal and the binding of the PET probe to the dopamine receptor. Together, these observations further confirm a central role for the dopamine system in craving behavior, validating it as a target for drug therapy. While animal experiments have suggested this, here the information is coming not only from human studies but, more important, from one of the specific human diseases of interest for a possible new drug. Greater confidence in the validity of targets should limit wasted effort in drug development on targets that are not functionally important.

Getting the Dose Right

Once researchers have identified a validated target for a drug, they can develop molecules to interact with the target. How much of a potential drug must be given in order to have the desired effect? This is a particularly critical question, because too much drug could be harmful by leading to unwanted interactions or overly inhibiting an important pathway. Even with marketed drugs, ensuring that each patient gets the optimal dose remains a major problem. Consider anti-epileptic drugs. If an insufficient dose is administered, seizures may continue unabated; if too high a dose is given, the patient may experience unacceptable slowness of thinking or other neurological symptoms, such as blurring of vision or unsteadiness.

One clever way of addressing this problem is to develop a PET tracer molecule that sticks to the drug target if there is no real drug present but is knocked off when the drug is around. By measuring the amount of the PET tracer sticking to the target in people taking different doses of the real drug, it is possible to determine exactly the right amount of drug needed to get the desired result. The higher the dose of the real drug, the lower the amount of the tracer that binds to the target (because more drug is bound). This all can be explored non-invasively, very early in drug development (even in Phase I trials, which test drugs for safety in healthy individuals). This will allow a rational choice of the dose used in a Phase II clinical trial to be defined with confidence.

multiple-scan of brain
Positron emission tomography (PET) images the distribution of a safe, tracer dose of a radioactively-labelled molecule injected into the body.  This example shows where a drug that is intended to treat a neurological disease binds to have its action. The image shows strong binding in deep regions of the brain, confirming that it reaches the intended site of action. This type of information can be of critical importance early in drug development. Image courtesy of Prof. Irene Tracey and Dr.Richard Wise, FMRIB Centre, University of Oxford

An alternative new approach now generating much interest is the use of fMRI to assess brain responses to different doses of a potential drug. By looking at the relationship between brain response and drug dose, researchers can estimate the clinically effective dose. Recent studies have shown, for example, how emotional areas of the brain (in the limbic system) respond to anxiety-provoking fearful faces presented in an fMRI examination.7 Administration of an anti-anxiety drug such as Valium will reduce this response. Because fMRI responses can be measured relatively quickly, various doses can be tested over a short period of time, potentially allowing the selection of an effective dose after study of only a small group of healthy volunteers or patients.

Combining Drugs

Increasingly, doctors will treat major symptoms or diseases by using two or more drugs together. This allows each drug to be used at doses that minimize unwanted adverse effects, while combining to produce an additive therapeutic effect. In the brain, we can find many examples of how several functional systems act together to generate a particular behavior, symptom, or disease. Since we know too little about the use of this strategy for Alzheimer’s disease to use that as an example, let us consider instead the problem of treating pain.

Most broadly, pain is not simply a sensory phenomenon; it also involves an emotional response to a stimulus. Bad pain is not just strong but also deeply unpleasant and disturbing, or even frightening. In addition, the perceived intensity of pain or its unpleasantness is determined by our expectations of the pain. All nurses and general practitioners are well aware of this reaction—for example, when they administer injections to children, they try to do so quickly, before the child is aware of the needle coming.

Studies using fMRI have allowed the complexity of pain to be probed in increasing detail. In one experiment, led by my Oxford University colleague Irene Tracey, D.Phil., some years ago, either pleasantly warm or painfully hot thermal stimuli were applied to the back of the hand soon after a particular colored light was turned on—one color for the pleasant sensation and another for the painful one.8 Tracey and her colleagues identified sensory, emotional, and attention-focusing regions of the brain that were more active during the periods associated with the burn than with the gentle warmth. Together, these regions form the parts of the brain that create the perception of pain. All of the fMRI signals related to this perception could be turned off if a powerful pain-relieving drug, such as morphine-like remifentanil, was given. The extent to which the signal was suppressed was related directly to the dose of the drug.9

Tracey and her colleagues were able to extend their observations beyond simply understanding pain by investigating what went on in the seconds just before the painful heat or the gentle warmth was applied, when the study participants viewed the colored light that signaled what was to come. The fMRI response during this period provided a brain measure of anticipation. As we might expect, the period before the painful heat was associated with real anxiety—even the most hardened research volunteer does not like being burned. Tracey found that distinct areas in the front part of the brain were active during this time. Use of an anti-anxiety drug such as Valium could block the signal from these regions selectively. This result illustrates how imaging can provide measures of pharmacodynamics (a technical term for the response to a drug), which can be more informative than conventional measures because of the selectivity and precision with which they can be measured. By better understanding precisely which aspects of pain processing are affected by different drugs, it is possible to more rationally think of ways in which drugs can be used together to get maximal benefit.

Understanding the Placebo Effect

A particular complexity in developing treatments for brain disorders lies in the way in which activity of the brain can change its own responses. We often observe a strong placebo effect (a benefit from inactive agents) in disorders of the mind. Recall Hamlet’s words: “Nothing is either good or bad but thinking makes it so.”

The placebo effect can be thought of as a response changed simply by expectation. In a drug treatment trial, this response complicates efforts to determine the true effectiveness of a drug. Even if neither the study participants nor the investigator knows which group of subjects is getting the active drug and which is getting an inactive comparator (placebo), the placebo effect tends to reduce the potential magnitude of difference between responses from the two groups.

A frontier area for brain imaging, particularly in pain studies, lies in trying to establish which brain systems are involved in placebo effects so that we can determine whether markers exist that could distinguish placebo responses from drug responses.10 Finding such markers, particularly if they could be applied easily to a larger population, could result in substantial reductions in the size of populations needed to test drugs, enhancing the speed and reducing the cost of the tests.

top view scan
Functional magnetic resonance imaging (fMRI) can be used to define drug effects. Here the brain response to a painful heat stimulus is shown during infusion of saline placebo or after administration of remifentanil, a morphine-like analgesic.  After the remifentanil, most of the brain activity signaling pain during the heat stimulus disappears, along with the subject’s perception of pain. Image courtesy of Dr. Tony Gee, GSK Clinical Imaging Center, UK

New Tools, New Potential

What will the future bring for imaging technology in drug development? A frontier area for partnerships between engineers, doctors, and pharmacologists lies in finding ways of translating sophisticated, expensive methods such as fMRI and PET into cheap, portable tools that can be used in large trials in all regions of the world, including the developing world. One example already in place is optical coherence tomography (OCT). OCT can be performed with a relatively inexpensive desktop device that uses a low-energy beam of laser light to rapidly scan across the retina of a subject looking into a computer-driven optical scanner. From sensitive measurements of differences in the way that reflected laser light bounces off the retina, a tomograph, or picture, of the back of the eye can be generated that allows the thickness of the retinal nerve layer and its distribution across the retina to be mapped with exquisite precision.

OCT was developed originally to assess nerve degeneration in diseases such as glaucoma, in which fluid pressure in the eye is increased, causing degeneration of the optic nerve. However, it also can be used for brain diseases. For example, David Miller, M.D., and his colleagues at University College, London, have recently shown how the degeneration of the optic nerve in multiple sclerosis can measured sensitively. 11   This advance presages a potential for the technique to be used to assess new drugs that might limit nerve degeneration in multiple sclerosis, which is thought to be the major cause of progression to full disability in that terrible disease.

New imaging technologies that allow direct study of the human brain and human brain disease promise to make drug development for diseases of the brain and mind faster and cheaper, delivering both more drugs and more-effective ones. However, we must also be cautious in the use of this technology. What we want to treat is the disease—the dis-ease of the patient—and that ultimately can be assessed only from the clinical effects—the symptoms and the outcome. Imaging is a tool to help move toward this goal with greater confidence, not a replacement for careful clinical observation of the patient.

With modern imaging, genetics, genomics, and information technologies we should now be in a position to rethink current drug development paradigms and ask, Can we move from the stage of having a new target to a new drug in just five years? Can we dramatically cut the high cost of drug development (which benefits no one)? Can we search for cures, rather than contenting ourselves with symptom management? The challenges of brain diseases are enormous. But with new imaging tools and an increasing understanding of brain function, there is hope that—even within our generation—new approaches to drug development will allow us to progress toward realizing real brain health all through life.


  1. Alexander GE, Chen K, Pietrini P, Rapoport SI, and Reiman EM. Longitudinal PET Evaluation of Cerebral Metabolic Decline in Dementia: A Potential Outcome Measure in Alzheimer’s Disease Treatment Studies. American Journal of Psychiatry 2002; 159(5): 738-745.
  2. Reiman EM, Chen K, Alexander GE, Caselli RJ, Bandy D, Osborne D, Saunders AM, and Hardy J. Correlations Between Apolipoprotein Epsilon4 Gene Dose and Brain-Imaging Measurements of Regional Hypometabolism. Proceedings of the National Academy of Science 2005; 102(23): 8299-8302.
  3. Schott JM, Price SL, Frost C, Whitwell JL, Rossor MN, and Fox NC. Measuring Atrophy in Alzheimer Disease: A Serial MRI Study Over 6 and 12 Months. Neurology 2005; 65(1): 119-124.
  4. Matthews, PM, Honey, GD, and Bullmore, ET. Applications of fMRI in Translational Medicine and Clinical Practice. Nature Reviews Neuroscience 2006; 7: 733-744.
  5. Mattay VS, Goldberg TE, Fera F, Hariri AR, Tessitore A, Egan MF, Kolachana B, Callicott JH, and Weinberger DR. Catechol O-methyltransferase val158-met Genotype and Individual Variation in the Brain Response to Amphetamine. Proceedings of the National Academy of Science 2003; 100(10): 6186-6191.
  6. Heinz A, Siessmeier T, Wrase J, Hermann D, Klein S, Grusser SM, Flor H, Braus DF, Buchholz HG, Grunder G, Schreckenberger M, Smolka MN, Rosch F, Mann K, and Bartenstein P. Correlation Between Dopamine D(2) Receptors in the Ventral Striatum and Central Processing of Alcohol Cues and Craving. American Journal of Psychiatry 2004; 161(10): 1783-1789.
  7. Paulus MP, Feinstein JS, Castillo G, Simmons AN, and Stein MB. Dose-Dependent Decrease of Activation in Bilateral Amygdala and Insula by Lorazepam During Emotion Processing. Archives of General Psychiatry 2005; 62(3): 282-288.
  8. Ploghaus A, Tracey I, Gati JS, Clare S, Menon RS, Matthews PM, and Rawlins JN. Dissociating Pain from Its Anticipation in the Human Brain. Science 1999; 284(5422): 1979-1981.
  9. Wise RG, Rogers R, Painter D, Bantick S, Ploghaus A, Williams P, Rapeport G, and Tracey I. Combining fMRI with a Pharmacokinetic Model to Determine Which Brain Areas Activated by Painful Stimulation Are Specifically Modulated by Remifentanil. Neuroimage 2002; 16(4): 999-1014.
  10. Petrovic P, Dietrich T, Fransson P, Andersson J, Carlsson K, and Ingvar M. Placebo in Emotional Processing–Induced Expectations of Anxiety Relief Activate a Generalized Modulatory Network. Neuron 2005; 46(6): 957-969.
  11. Trip SA, Schlottmann PG, Jones SJ, Li WY, Garway-Heath DF, Thompson AJ,Plant GT, and Miller DH. Optic Nerve Atrophy and Retinal Nerve Fibre Layer Thinning Following Optic Neuritis: Evidence That Axonal Loss Is a Substrate of MRI-Detected Atrophy. Neuroimage 2006; 31(1): 286-293.