Using objective markers and imaging in the development of novel treatments of chronic pain

Chronic pain is an area of an immense unmet medical need and pharmaceutical companies have been putting a lot of effort into developing new pain treatments. In spite of this, attrition in this area has been high across the industry. With most candidates failing in early clinical development phases, the preclinical–clinical interface is the most likely reason for the low success rate. Currently, discovery of novel analgesics relies on animal models of pain; thereby selected candidates are then tested in proof-of-concept (PoC) patient trials. However, there are major concerns about the predictive value of animal models; one much discussed example being the tachykinin NK1 receptor antagonist class [1]. Conversely, PoC trials utilize subjective end points, mostly those based on patient’s report of pain intensity; these are inherently variable and require large numbers of observations to detect signals of efficacy [2]. Thus, the current approach to analgesic development is unsustainable, and development of tools enabling early detection of efficacy in humans seems key to success.

There are expectations that objective markers of nociception and pain can be used to provide evidence of pharmacological activity and therapeutic efficacy at early clinical stages. In addition, emerging imaging techniques have been contributing to a much better understanding of pain and analgesia mechanisms, helping describe the heterogeneity of pain and define patient selection criteria for efficacy trials. They also have the potential to minimize the preclinical–clinical interface by using similar approaches in animals and humans.

To date, the focus of preclinical target discovery has been largely on peripheral and spinal mechanisms of pain. Many promising new targets have been identified on nociceptive primary afferents, including the transient receptor potential (TRP) family, purinergic P2X receptor family, excitatory amino acid receptors, and sodium, potassium and calcium channels [3]; selective ligands are being developed for many of these targets. Objective assessment of nociceptor activation in the periphery can therefore be useful for detection of pharmacodynamic (PD) activity of such compounds in humans. This can be achieved using imaging techniques such as infrared thermography and laser-Doppler imaging (LDI). These methods detect minute changes in blood flow, for example in the skin, related to the release of vasoactive neuropeptides, (e.g., calcitonin gene-related peptide [CGRP]). Since many nociceptive fibers contain CGRP, their activation causes axon reflex-mediated retrograde release of neuropeptides and vasodilatation [4–7]; conversely, pharmacological agents reducing nociceptor excitability (e.g., sodium channel blockers and opioids) reduce flare evoked by noxious stimuli [8–10]. Recently, a novel TRPV1 antagonist has been tested in a Phase I study in humans and demonstrated by LDI to reduce flare evoked by capsaicin or ultraviolet irradiation, the effects being related to the drug plasma levels [11]. Thus, LDI and similar methods of quantitative assessment of nociceptor activation can provide important pharmacokinetic (PK)/PD information, as well as evidence of antinociceptive activity at the peripheral level.

For central pain mechanisms, imaging tools exist for assessment of activity-dependent changes in regional cerebral blood flow (using H₂¹⁵O positron emission tomography [PET]) and oxygenation (using blood oxygen level-dependent functional MRI [BOLD-fMRI]). These methods have provided extensive evidence on the involvement in pain processing of bilateral thalamus, insula, anterior cingulate, SII somatosensory cortex and several other supraspinal structures [12,13], sometimes collectively referred to as the ‘pain matrix’. There is increasing understanding of the role of cortical structures (such as insular, anterior cingulate and prefrontal cortex) in pain perception. Specifically, the role of the insula as an integrative structure receiving convergent nociceptive inputs has been discussed [14]. Activation of the insular cortex is directly correlated with the intensity of noxious stimuli as well as pain perception, indicating its role in encoding pain intensity (although the somatosensory cortex, e.g., SII, has also been implicated [15,16]). Importantly, pain-evoked insular activation has been shown in placebo-controlled studies to be modulated by intravenous analgesics, such as the opioid, remifentanil, and the N-methyl-D-aspartate antagonist, ketamine [17–19]. All of this implies that it may be possible to use measures of activation of the pain matrix, and the insula in particular, as objective markers of acute pain in humans. In addition, cortical and subcortical activation have been demonstrated in experimental and clinical conditions of pain and sensitizations associated with hyperalgesia (exaggerated painfulness of noxious stimulation) and allodynia (pain evoked by innocuous stimuli); for example, in capsaicin-evoked sensitization or in patients with neuropathic pain [20–22]. The pattern of activation, although distinct, partially overlaps with the pain matrix activated by acute noxious stimuli, with significant activations in the insular, anterior cingulate and prefrontal cortices, as well as brainstem structures [12,13,21,23]. While these data suggest that measures of pain-related cortical activation could be useful at early development stages as indicators of PD activity in humans, several potential caveats need to be considered. First, the specificity of pain-evoked cortical activation is not fully understood. Thus, several imaging studies have demonstrated areas of activation in prefrontal, insular and anterior cingulate cortices during tasks involving emotionally charged cognition, anticipation, reward and motion preparation paradigms [24–30], showing at least partial overlap with the areas related to pain perception. Second, variability of these responses and their sensitivity to analgesic treatments versus other objective markers (e.g., electroencephalography [EEG], see below) and standard pain intensity scales needs to be assessed. It has now become possible to record pain-evoked fMRI and EEG activation simultaneously [31]; data on the relative value of these objectives versus subjective pain markers are eagerly awaited.

The caveat of limited specificity of pain-related cortical activation outlined above suggests that multilevel assessments of activity in nociceptive pathways may offer a robust supplementary or alternative approach. It is now possible to detect and quantify activation along the whole nociceptive pathway (e.g., in the trigeminal system) [32,33]. Thus, noxious heat stimulation of the face in healthy subjects has been demonstrated to evoke activation in the trigeminal ganglion [34], as well as the second and third-order neurons located in the spinal trigeminal nucleus and thalamus, respectively, and in the somatosensory cortex [35]. Variances from normal activation in the trigeminal system have been described in neuropathic pain. It has been demonstrated that experience of brush and cold allodynia, and heat hyperalgesia in patients with unilateral trigeminal neuropathy is associated with significant activations in the trigeminal ganglia, thalamus and SI, as well as outside the trigeminal pathway (brainstem and cortical activations) [36]. Thus, fMRI correlates of exaggerated brain responses to painful innocuous (allodynia) and noxious (hyperalgesia) stimuli have the potential to serve as objective markers of sensitization in nociceptive pathways. Importantly, the activations are reproducible between visits [36], making this methodology amenable to use in controlled drug studies involving repeated efficacy assessments and crossover designs.

In contrast to fMRI measures of acute pain and hyperalgesia/allodynia, quantitative assessment of ongoing chronic pain with imaging methods has been challenging. This is due to the specifics of the fMRI analysis whereby temporal changes in BOLD signal are related to the timing of experimental stimuli. One interesting recent approach utilizes temporal profiles of spontaneous fluctuations of ongoing pain for signal analysis [37]. Using this methodology in patients with chronic back pain and post-herpetic neuralgia, pain-related activations have been observed in affective and sensory-discriminative areas (thalamus, SI–SII, prefrontal, anterior cingulate, insular cortices) [38,39]. Interestingly, while short-term analgesic efficacy of lidocaine-patch treatment was associated with attenuated activation in these areas, long-term efficacy also reduced activation of areas involved in reward and punishment, such as the ventral striatum and amygdala [38].

Complementary information to that gained from blood-flow imaging techniques, such as H₂¹⁵O PET and fMRI, can be attained from EEG and magnetoencephalography (MEG) studies. The principles of EEG/MEG rely on the basis of detecting fluctuations in the electromagnetic activity generated by cortical neurons [40]. Both techniques have excellent temporal resolution (millisecond) meaning that brain activity can be monitored in real-time, as opposed to PET and fMRI, which detect changes that are secondary to neural activity. This is a particularly important feature when studying pain, as objective assessment of pain transmission and cortical processing may require differentiation of stimulus-specific and stimulus-related components [41].

Stimulus-specific aspects are generally represented as the phase-locked early components of the pain evoked response. These waveforms, or evoked potentials (EPs), result form the arrival of the initial afferent volley from the periphery in the cortex [42]. Measures of the amplitude and latency of these components can inform us as to the conduction velocity and sensitivity of the nociceptive pathway and can be particularly useful in characterizing neuropathic pathology and pharmacological analgesic effects [43–45]. Cortical neural activity, which is time-but not phase-locked to the stimulus, can also be detected by monitoring changes in the oscillatory frequency of neural activity within specific brain regions. For instance, a 13–30-Hz β-rhythm is predominant in sensorimotor regions and this ‘idling’ activity is significantly suppressed during experimental pain [46]. Stimulus-related components reflect higher order processing of pain signals and these components can often be readily modulated by changing the context or state of arousal of the subject [41]. Importantly, these later EP components have shown some clinical significance in differentiating patients with different etiologies of pain [43].

This type of spectral analysis has also been utilized to assess the pharmacological modulation of cortical neuronal function [47]. A study using benzodiazepine revealed a number of changes in oscillatory power across the cortex. Most notable were the increases in β (15–25 Hz) synchronous power in bilateral regions of the frontal cortex, sensorimotor cortex and occipital cortical areas. The ability to directly and accurately measure drug-induced changes in cortical neural function within specific cortical networks is a concept that lends itself to a wide spectrum of applications relating to the understanding of normal brain function, the treatment of neurological disorders and the targeted design of specific therapeutic agents.

An example of how this may be relevant to pain comes from a recent study by Stern and colleagues [48]. Using EEG, it was shown that patients with neuropathic pain have increased low β-band activity (12–16 Hz) when compared with healthy controls, and this activity localized to a host of cortical structures within the pain matrix. Interestingly, surgery to create a therapeutic lesion in the thalamus reduced the oscillatory power within this frequency band, allowing the authors to conclude that alterations in thalamo–cortical rhythmic activity may play an important role in the generation of symptoms in patients with neuropathic pain conditions.

It is also possible to record EEG/MEG responses to visceral pain [49]. These studies have demonstrated that the visceral EP is predominantly mediated via Aδ-fibers, and can be reliably evoked by both electrical and mechanical (distension) stimulation [50]. As with somatic pain models, EPs have been used to provide objective neurophysiological evidence of increased afferent sensitivity after a sensitizing intervention (esophageal acid infusion, for instance) [51]. This approach has also been used to differentiate patients with visceral hypersensitivity in a manner similar to somatic pain studies [52,53]. It has also been possible to show that visceral and somatic pain are processed in a broadly similar manner within the cortex, with activity occurring in parallel with the insula and somatosensory cortices [54,55].

In addition to the functional markers of pain and sensitization assessing neural activity, structural imaging techniques can provide important information on CNS plasticity and neurodegeneration. Evidence is emerging on brain remodeling and neuronal atrophy in certain chronic pain states. Thus, MRI studies have demonstrated cortical and thalamic gray matter loss in patients with chronic back pain and fibromyalgia, correlated with the duration of the disorder [56,57]. If confirmed by longitudinal studies and linked to information on analgesic efficacy, such data could help understand the neuroanatomical basis of pain heterogeneity and potentially stratify patient populations for efficacy trials.

Other important areas of application of imaging tools in analgesic development are target-occupancy studies using PET. They can be used to demonstrate drug interaction with its target sites in the brain and link target occupancy with pharmacological effects. In addition, PET ligands can be utilized to provide information on the importance of a particular target in various pain states. This can be information on receptor density and distribution, endogenous neurotransmitter release, or both. As an example, the role of the endogenous opioid [58,59] and dopaminergic systems [60–62] in pain modulation and chronic pain states has been documented, with potential implications for new therapeutic approaches.

In summary, imaging techniques such as MRI, PET and MEG have specific roles in the discovery and development of novel analgesics. Each approach has strengths and weaknesses, and careful consideration should be given when designing pain experiments that the correct technology is utilized to answer specific biological questions. Importantly, cross validation data are beginning to emerge; for example fMRI, PET and EEG/MEG studies confirming localization of the main structures of the pain matrix and addressing specific methodological concerns. To date, these methods have been primarily useful to increase our understanding of mechanisms of pain and analgesia in humans. Some methods are beginning to be considered as potential objective markers of pain and/or PD activity that may be more sensitive to treatments than clinical end points. Validation with standard treatments, utilizing randomized, controlled study designs and dosing regimens, still needs to be carried out before any novel drugs can be explored. Principally, analysis of sensitivity versus standard clinical efficacy end points is required for any such objective pain markers.