|Home | About | Journals | Submit | Contact Us | Français|
Rapidly developing, non-invasive, neuroimaging methods provide increasingly detailed structural and functional information about the nervous system, helping advance our understanding of pain processing, chronic pain conditions, and the mechanisms of analgesia. However, effective treatment for many chronic pain conditions remains a large, unmet medical need. In this review, we examine how the neuroimaging of pain has enhanced our understanding of the mechanisms of chronic pain, our current understanding of the central neural correlates of pharmacologic modulation of pain, and the role of neuroimaging in analgesic development, discussing both current limitations and future directions
Neuroimaging has rapidly changed the field of pain research. Electrophysiological, radiological, and magnetic resonance methods reveal increasingly detailed information about the neural correlates of pain perception and modulation in both animal and human models. These methods have helped identify the central neural structures involved in pain processing termed the “pain matrix” , and establish hypotheses related to the central mechanisms responsible for the generation and maintenance of chronic pain conditions. These methods have the potential to objectively diagnose central nervous changes in pain disorders and to identify better neural targets for treating chronic pain.
In previous reviews, investigators have discussed the application of functional magnetic resonance imaging (fMRI) to the research and development of novel drugs [2-4]. In this review, we examine how the neuroimaging of pain has enhanced our understanding of the mechanisms of chronic pain and the neural correlates of pharmacologic modulation of pain. We will discuss the current limitations and future directions of the role of neuroimaging in analgesic development.
The use of neuroimaging has enabled the non-invasive study of the central nervous system. Several techniques have been used to investigate pain including electrophysiological methods such as, electroencephalogram (EEG) and magnetoencephalogram (MEG), radiological methods such as positron emission tomography (PET) and single emission computerized tomography (SPECT), and magnetic resonance (MR) techniques such as magnetic resonance spectroscopy (MRS), structural magnetic resonance imaging (MRI), diffusion tensor imaging (DTI) and functional magnetic resonance imaging (fMRI). These neuroimaging techniques can be categorized as structural (revealing anatomical information, e.g., MRI, DTI), biochemical (revealing information regarding the local chemical environment, e.g., MRS) or functional (revealing signal changes related to neuronal activity, e.g., EEG, MEG, fMRI).
Investigators have employed structural imaging to associate anatomical changes in the brains of chronic pain patients. Although structural imaging gives no direct information about neural function, it provides indirect information about how chronic pain effects central plasticity and identifies the anatomical differences between pain patients and healthy controls. Structural imaging can be used to track longitudinal changes due to disease severity and progression and may potentially characterize changes following treatment. The regions of interest identified using structural imaging may then be targeted for follow-up investigations into brain function. Structural imaging methods include voxel-based morphometry (VBM) and cortical thickness analysis (CTA). Investigators have used VBM and CTA to identify structural differences in the gray matter densities of subjects with low back pain  and fibromyalgia . Decreases in prefrontal gray matter in these patients may help explain some of the cognitive issues noted in patients with chronic pain. Furthermore, white matter tract changes can be characterized using DTI tractography. Hadjipavlou et al. used this technique in healthy subjects to observe the physical connections of the periaqueductal gray (PAG) and nucleus cuneiformis to the amygdala, hypothalamus, prefrontal cortex, rostroventral medial medulla and thalamus . This method could identify associated differences in anatomical connectivity in chronic pain conditions.
Imaging of the biochemical environment of the central nervous system can help reveal abnormalities in chronic pain. Investigators have used magnetic resonance spectroscopy (MRS) to identify biochemical differences in brain regions of healthy individuals and chronic pain patients. Decreases in N-acetylaspartate (NAA) have been observed in the dorsolateral PFC of subjects with chronic low back pain  and complex regional pain syndrome type I  and in the thalami of neuropathic pain patients . NAA, considered a marker of neuronal activity, is found only in mature neurons [11, 12]. Investigators have proposed NAA as a possible biomarker of neuronal health and function that can be assessed non-invasively using MRS [8-10].
Neuroimaging studies of pain have largely used fMRI and PET. Both techniques rely on hemodynamic responses in order to identify areas of metabolic change. Functional MRI uses an endogenous contrast mechanism, changes in the blood oxygenation to identify areas of neuronal activity. This is called Blood Oxygenation Level Dependent (BOLD) imaging . PET imaging requires injection of a labeled tracer, commonly H215O. SPECT imaging provides unique insight into brain function by providing a snap shot or frozen look at the brain state immediately after perfusion tracer injection. This is advantageous in the diagnosis of neuropathic brain states, as the dynamic nature of signal change in fMRI and PET may in some cases, yield complicated results .
These functional imaging methods have been used to characterize the complex network of nociceptive, affective, and cognitive elements that process pain. The thalamus, primary and secondary somatosensory cortex, insula, prefrontal cortex (PFC) and anterior cingulate cortex (ACC) have been identified to be involved with the perception and modulation of pain and are collectively known as the “pain matrix”. In addition, other regions such as the amygdala, caudate, amygdala, and parahippocampal area are known to influence pain perception. Ultimately, these activity maps may some day provide an objective measure of pain and its modulation in comparison to self report. This task, however, is complicated by the multiple cognitive and psychological factors that mediate pain perception. As a result, the neural picture of pain is also complicated.
FMRI, in particular, has revolutionized our understanding of the brain's role in the perception and maintenance of pain and the neural correlates of factors modulating it including: attention , anticipation , fear/anxiety , and direct control . Functional MRI has been employed to investigate differences in how healthy individuals process pain during normal and sensitized conditions , and its use has advanced our knowledge of the central neural correlates associated with a number of chronic pain conditions .
PET imaging is another functional neuroimaging method used extensively to characterize the neural correlates of pain processing and modulation. PET requires the use of radio-labeled ligands which restrict the duration of scanning and the quantity of scans that can be performed. PET can be used to investigate changes in metabolic activity related to a disease condition or to an evoked response. For example, Witting et al., has recently used PET imaging to identify the neural correlates of pain processing in chronic neuropathic pain conditions and shown it to differ from that observed during evoked pain in healthy subjects . Additionally, PET can be used to investigate the role and function of specific drugs and neurotransmitters. In these studies, investigators use sub-therapeutic doses of radio-labeled agents to investigate the functioning of the targeted systems. For example, Zubieta et al., has used PET to reveal opioid binding in brain regions associated with the perception of pain and has characterized specific binding sites related to the sensory and affective dimensions of pain . This group has also used PET imaging to characterize genetic and gender variability in opioids system activation [23, 24]. Additionally, PET imaging has identified decreases in opioid receptor binding in central and peripheral neuropathic pain [25, 26] and central poststroke pain . More recently, combined opioid and dopamine systems have been imaged, providing opportunities to investigate multiple neurotransmitters and their interactions . These studies shed light on the functioning of specific neurotransmitter systems in both healthy and in chronic pain conditions. While important in identifying potential targets for analgesia, these studies do not use therapeutic doses of PET ligands to investigate pharmacologic manipulation of the pain processing and modulation systems. Such studies are important in identifying the central mechanisms of effective and ineffective analgesia.
Although the integration of neuroimaging and drug administration has helped elucidate the neural correlates of pain and analgesia, neuroimaging techniques can further characterize the pharmacodynamics and pharmacokinetics of analgesics by identifying the sites, actions and time course of modulation. Whereas current analgesia methods include diverse drugs that act on a range of targets, if we can characterize the brain regions each drug acts upon, then we can more fully understand the neural mechanisms of analgesia. Table 1 lists the existing neuroimaging studies in which pharmacologic modulation of pain has been investigated.
Investigators have primarily used two methods, PET and fMRI, to investigate the pharmacological modulation of pain. To date, PET studies that have integrated analgesia and pain have used 15O-water to identify areas in which changes in regional cerebral blood flow (rCBF) occur [29-31]. Remifentanil administration during painful heat stimulation has led to increased rCBF in the PAG and perigenual ACC, and decreased rCBF in the thalamus and somatosensory cortex . These results have important implications for our understanding of the pharmacologic modulation of pain and the central opioidergic system as it demonstrates regions that may be involved in the control of pain (PAG and perigenual ACC) and those that may be involved in perception of decreased pain (somatosensory cortex) . Casey et al. observed fentanyl induced rCBF changes in the mid-anterior region of the cingulate cortex during modulation of painful cold, but not during non-painful cold . Interestingly, an earlier study by Adler et al., found that the fentanyl induced modulation of painful heat, but not non-painful heat, resulted in rCBF increases in the supplementary motor area and frontal cortex . The variability amongst these two studies warrants further investigation to determine the influence of the type of pain and other variables which may contribute to the differences in the responses observed. As 15O-water labeled PET indirectly is sensitive to regional cerebral blood flow, interpretation of results must be performed carefully as any effect of the analgesic on the relaxation properties of the central vasculature may affect the hemodynamic response. A global relaxation in the responsiveness of the vasculature would result in a decrease in the ability to obtain further changes in signal. The use of fluorodeoxyglucose may more specifically identify areas of pharmacologic modulation by relying on glucose metabolism rather than hemodynamics for the source of the signal.
PET imaging has also been used to observe drug distribution. Otte et al. investigated the distribution of [11C] labeled eletriptan, an anti-migraine medication to observe the location and amount of medication that crosses the blood brain barrier .
One of the advantages of neuroimaging methods is the ability to identify the dose-dependent actions on specific neural systems, rather than on self-reports. Recently, Wagner et al. used PET to investigate the neural correlates of dose-dependent modulation of evoked pain by remifentanil in healthy subjects . They demonstrated dose-dependent reductions in the thalamus, basal ganglia, prefrontal, secondary somatosensory (S2), insular, temporal, parahippocampal, and occipital cortices and increases in the cingulofrontal cortex and PAG. They compared only two dosage levels of remifentanil due to the limitations imposed by the administration of radio-labeled tracers that restrict the time in which imaging can be performed and number of times a subject can be scanned. Functional MRI would be advantageous in such studies as it would allow greater flexibility in the number of times scans could be repeated in the same individuals, and it would enable the same study to be performed with smaller steps of increasing dosage. Similarly, repeated scans can be obtained in SPECT perfusion imaging if split doses are used . Functional MRI has been integrated with the pharmacologic modulation of brain activity using two approaches: one in which alterations in brain activity are observed following drug administration and a second in which the neural correlates of a task are observed under basal conditions and following drug administration. The utilization of these two methods have been previously discussed . Wise et al. used the latter method in a study in which they integrated a pharmacokinetic model into fMRI analysis (Figure 1) . In their paper, they estimated the effect site concentration of remifentanil to identify pharmacologic modulation of evoked pain. They extended this study to use fMRI to monitor the time course of actions of remifentanil in the insula of healthy volunteers during evoked pain . Using the insular cortical time course together with a simple pharmacokinetic model, they predicted the half-life of activity of remifentanil. They demonstrated reductions in BOLD activity in the right insula during remifentanil administration as compared to saline [34, 35]. Their work provided important insights and support the concept that pharmacologic fMRI (phfMRI) can be used to investigate both the central neural systems correlates of analgesia and its time course.
The integration of pharmacologic modulation in fMRI studies can also be used to characterize different qualities of pain perception. Wise et al. have investigated the anxiolytic action of midazolam during pain-related anxiety . They hypothesized that fMRI would reveal changes related to the modulation of anxiety related to a conditioned painful stimulus, however, pain related fMRI activity would remain unchanged. Percentage signal changes during pain were not significantly different between midazolam and saline infusions in the ACC, contralateral primary sensory cortex, bilateral anterior and posterior insula, or bilateral thalami. During anticipation of pain, significantly different percentage signal changes were observed in the contralateral anterior insula. Additionally, differences were observed in the ACC and ipsilateral S2. Studies such as this help characterize specific dimensions of pain.
In addition to PET and fMRI, MEG has also been used to observe the pharmacologic modulation of pain . The greatest advantage of this technique may be that the source of the signal arises from magnetic dipoles due to neuronal activity and therefore is independent of any hemodynamic response. This method also offers advantages in terms of cost and temporal resolution. However, there are significant limitations of this technique in that imaging is restricted to cortical areas. Therefore, imaging of deep brain structures of interest such as PAG, thalamus, and ACC are not possible.
Despite the variety of available analgesics, many chronic pain patients receive inadequate relief. Physicians treating these challenging patients face daunting obstacles. To optimize compliance, minimize side-effects, and maximize analgesia, each one of these anti-neuropathic medications needs to be started at a low dose and titrated to effect over many weeks in a manner well-described as “labor-intensive” . Frustratingly, each anti-neuropathic medication produces pain relief for only a minority of patients with neuropathic pain . Not surprisingly, few physicians have the patience required to persist in a strategy involving months of frustrating, side-effect-laden medication trials. Therefore, novel methods to optimize the matching of a particular drug to a particular patient are needed.
These challenges not only affect the physician taking care of patients but also the pharmaceutical companies developing novel drugs. Drug development is an expensive process in terms of time and money and neuroimaging may improve the efficiency of that process. It is unknown why some chronic patients do not respond to some treatments. Neuroimaging could play an important role in discerning the neural differences between responsive and unresponsive patients that may lead to less expensive and more revealing clinical trials. Neuroimaging may also reveal potential targets that have yet to be explored. These techniques can also help identify more accurate dose response curves, thereby minimizing required dosages. The inclusion of neuroimaging may increase confidence in drug trial results by verifying that the desired targets are met.
The use of neuroimaging techniques in drug development may alter the way in which drug trials are designed and implemented [47, 48]. Drug trials are a lengthy and costly process. By identifying if drugs are successfully modulating target regions early on, neuroimaging techniques may decrease the time of drugs trials and may allow the earlier termination of ineffective drugs. Non-invasive imaging techniques enable repeated scans within individual subjects. As a result this can decrease the number of subjects that are required. Further longitudinal studies may provide insight into the central mechanisms of drug sensitivity.
Several remaining issues must be addressed to enable functional neuroimaging to be utilized successfully as a research and clinical tool to characterize the effects of analgesic agents in the human central nervous system. These issues include better characterization of the neural correlates underlying the inter-subject variability of pain perception, the heterogeneity of different pain conditions, appropriate experimental design including the selection of evoked pain stimuli and the value of applying our knowledge of acute pain processing in the context of chronic pain conditions. Additionally, appropriate controls must be defined in order to address placebo, expectancy and variables that may affect the neural-hemodynamic coupling. Finally the pharmacodynamics of a pharmacological agent must be considered in order to determine the most appropriate experimental method in which to study its actions.
Establishing the appropriate experimental conditions in which to study chronic pain can also be difficult. The nature of examining pain relief in chronic pain patients may include confounds due to placebo and expectancy. These can be reduced by appropriate blinding, cross-over studies and other sophisticated study design and statistical methods ]. It remains unclear as to whether placebo responders should or should not be included in studies. To confound the discussion, it may be that the placebo responders are also the pharmacologic responders. Appropriate application of neuroimaging may distinguish the true pharmacologic response from the placebo response.
The method in which to study pain in a chronic pain group is also difficult. Some of this difficulty is related to the heterogeneity of chronic pain conditions. This contributes to a significant variance between subjects with different pain conditions. There is a need for larger number of studies in homogeneous patient populations that have been carefully phenotyped. Additionally, the manner in which pain is elicited in chronic pain patients can be a significant confounder. Commonly, an evoked stimulus is used. The central changes that take place in chronic pain conditions may result in differences in the processing of evoked pain as compared to unaffected individuals. Additionally, it is not clear how an experimentally evoked stimulus relates to a patients spontaneous or dynamic pain. One recent option is to directly measure the patient's own endogenous or dynamic pain while simultaneously measuring the resulting brain activity. Apkarian et al., has performed this in low back pain patients and was able to use a patients changing pain measurement as a regressor to identify specific brain regions involved with the perception of their spontaneous back pain . Ultimately, this may find utility in combining this dynamic pain measurement with an analgesic. Even this method, however, has limitations as the very process of cognitively evaluating one's pain involves specialized brain regions that may confound interpretation of the specific analgesic activity.
An alternative to studying heterogeneous chronic pain conditions is to attempt to mimic chronic pain in healthy individuals. Topical capsaicin has been used to establish a sensitized state in healthy individuals . Iannetti et al. used this experimental model to compare pain processing in healthy subjects under normal conditions and following capsaicin-induced secondary hyperalgesia as well as analgesia by gabapentin (Figure 2) . These experimentally induced pain conditions clearly do not reproduce the true altered neural processing present in a chronic pain condition however, it may provide a model in which aspects of chronic pain can be studied under more controlled conditions.
There are technical limitations to consider in the application of neuroimaging to study pharmacologic analgesia. The main issue is the physiological source of the signal that many techniques depend on. Methods such as PET and fMRI are dependent on activity related neural-metabolic changes (e.g. glucose or oxygenation) and hemodynamic responses. Uncoupling of this relationship by pharmacologic actions will therefore adversely affect the observed signal. Changes in signal observed in fMRI and PET can then be the result of not only changes in neuronal activity, but global changes in cerebral blood flow or cerebral blood volume. Therefore, it is important to consider other sources of signal and verify that other factors are not major contributors. A recent review has identified key problem areas that should be considered in order to improve phfMRI . Study designs must include controls to verify that the effects of the drugs are not on the hemodynamics, and are not global. One method of identifying the effects of pharmacologic agents on the properties of vasodilation is a breath holding task. Pattinson et al.  employed a breath holding task to demonstrate that the hemodynamic response observed was not a global effect and was in fact localized to target regions and were related to neural activity.
Some studies have included additional functional tasks that target regions that are known not to be modulated by the drug of interest. This delineates any global effects of the agent. A third control is to employ a crossover study that observes a tasks during analgesic and placebo administration such as that used by Iannetti el al. . This design can control confounding effects due to expectancy and placebo and allows each subject to be used as their own control. However, often same day scanning is impossible due to the latency of drug action.
There are some experimental limitations to be aware of that are governed by the properties of the analgesic of interest. The method of administration, drug action, latency, and period of effectiveness are all factors that need to be considered. The administration and latency of the drug may determine if studies are longitudinal or done over a single session or day. Medications administered intravenously have an advantage in that they have shorter latencies and so baseline scans can be acquired minutes before analgesia is achieved. There can also be variation in metabolism rates of medications across individuals. One approach taken to address this issue is the sampling of serum levels to identify the circulating level of analgesic.
Neuroimaging can aid the identification of the neural correlates of analgesia, the identification of new targets, and quicker and more accurate assessment of novel analgesics. The use of neuroimaging can help supplement the experimental design used to study the treatment of poorly understood chronic pain conditions such as central neuropathic pain. The central mechanisms underlying neuropathic pain are believed to include hyperexcitability of neurons. Neuroimaging may first play a role in enhancing elegantly designed studies of pharmacological pain management. Siddall et al. recently observed several clinical measures of reduced neuronal hyperexcitability in spinal cord injury patients with neuropathic pain treated with pregabalin in comparison to placebo . The use of neuroimaging would be helpful in elucidating the mechanisms and sites of modulation.
New techniques in neuroimaging could also be applied. An emerging field in fMRI is the use of resting state fMRI, or investigation of the default mode. Such experiments examine fluctuations in the signal changes during a task-free state and identify regions of neural activity by data driven analysis such as independent component analysis. To date this work has identified possible circuitry that is active when the brain is “resting” or not focused on a particular task. Differences in default mode networks have recently been observed in disorders such as depression  and anxiety  in comparison to healthy individuals. The technique could potentially identify differences in the general neural networking in chronic pain conditions.
A natural progression of phfMRI research is to look beyond individual regions of modulation and examine the networking between them. Connectivity analyses are used to establish the temporal relationship between areas of fMRI signal change observed and provide estimations of causality. To date, one animal study has measured cerebral blood volume (CBV) changes following administration of D-amphetamine and fluoxetine in order to delineate the functional connectivity of the dopamine and serotonin neurotransmitter systems . The application of connectivity analyses in pain models may aid in identifying the network changes that lead to chronic pain conditions.
Structural imaging techniques may also aid in identifying new targets for analgesia. These methods could aid in identifying longitudinal structural changes following treatment, without susceptibility to hemodynamic signals that functional methods are reliant upon.
Neuroimaging may play an important role in the pain clinic in providing a tool for directing personalized medicine. The combination of neuroimaging and genetic biomarkers may indicate which analgesics a patient is most likely to respond to. Furthermore, these techniques may also identify patients who are likely to suffer side effects. A recent study used cerebral blood flow effects to investigate genotype-based differences in the response to remifentanil .
To this point, we have focused the applications of brain imaging. It is also important to consider applications at the level of the brainstem and spinal cord. The periaqueductal gray, for example, is known to be an important structure in the descending modulation of pain. It is therefore of interest to not only examine regions of the brain involved in chronic pain and pharmacologic modulation, but also the spinal cord. PET has been employed to investigate morphine-induced cholingeric activity in the monkey spinal cord . The development of fMRI for assessment of spinal cord function is relatively new. While it has been used to investigate sensory processing in subjects with spinal cord injury , the investigation of spinal cord activity in chronic pain patients has yet to be performed. These potential applications for spinal fMRI in chronic pain conditions and pharmacological interventions have been discussed in a recent review . Investigation of this and other pain conditions and analgesia at the level of the spinal cord remains uncharted territory.
Finally, the use of multi-modal imaging combines the advantages of several techniques in order to gain more comprehensive information. For example, one may utilize the temporal resolution of EEG and the spatial resolution of fMRI. These techniques have been combined to observe the responses to noxious electrical stimulation  and laser-evoked responses . Such studies used to investigate chronic pain and the modulation by analgesics pain help elucidate the time course of pharmacologic actions.
Neuroimaging can play an important role in directing the development of novel analgesics. Structural imaging techniques can identify regions of the brain in which anatomical differences exist in chronic pain conditions. Functional techniques such as fMRI and PET can identify the regions of interest in which there are metabolic differences. Techniques such as MRS may identify biomarkers that can be used to identify biochemical changes in these regions of interest. Neuroimaging methods can aid in identifying the brain regions in which modulation is required to achieve analgesia and which areas remain unaffected during inadequate pain relief in order to identify new targets for novel analgesics. In future, fMRI shows promise as a clinical tool for identifying specific analgesic needs of chronic pain patients, as well as in research as a method of assessment of analgesic action. In order for neuroimaging techniques to be utilized to their full potential, further development of proper controls is required to accurately interpret the obtained results. With established protocols in place, neuroimaging methods may lead to quicker and more efficient analgesic development strategies.
The authors wish to acknowledge support by a grant from the John and Dodie Rosekranz Endowment (SCM) and grant NS053961from NIH NINDS (SCM).