OR WAIT null SECS
The implications and challenges of the placebo effect on regulatory agency product approval.
The placebo effect has long confounded pharmaceutical company efforts in the development of new analgesic products. Experience has shown that while oral analgesic trials elicit some placebo response among subjects, placebo response rates tend to increase with transdermal gel or cream products and become excessively large with transdermal patch products. Thus, the placebo effect has become increasingly frustrating as drug companies seek to take advantage of the popularity and established safety of transdermal delivery systems. While transdermal opioids and amides have been successful in achieving market approval (likely because they are inherently stronger drugs and, in the case of opioids, are designed to work centrally and achieve a higher drug blood level) the same cannot be said of transdermal nonsteroidal anti-inflammatory drugs (NSAIDs). While, clearly, multiple factors influence the lack of success in NSAID transdermal trials, the outsized placebo response is a primary consideration (particularly in transdermal patch formulations). This has resulted in enormous financial expenditures as drug companies make repeated attempts to reduce placebo interference to achieve the two positive efficacy trials required for FDA marketing approval.
The FDA has approved few transdermal NSAID products for relatively narrow use in the United States. King's product Flector® (diclofenac) is the only NSAID patch approved for use in the United States and its path to approval was not without challenges: the pivotal trials suffered from such high placebo response rates that the two efficacy trials included in the initial submission did not demonstrate a significant treatment effect. According to the summary basis for approval, it was only after the addition of two more efficacy trials, redefinition of the endpoints, and the addition of two separate methods of imputation was this product approved. Additionally, the FDA reviewer noted that "the results were noticeable for quite high response rates in both groups (active vs. placebo), compared to what has been observed in previous trials of oral analgesics."
Numerous other companies have attempted to develop transdermal NSAID products in the past decade, but the requisite Phase II-III studies have thus far failed to meet the efficacy endpoints necessary for FDA approval, owing largely to a failure to separate the treatment effect from placebo. In some cases, this has resulted in abandonment of the product development altogether.
The reasons for the extremely high placebo response rates seen in the transdermal formulations of the NSAID drug class have been poorly understood, but examining the placebo response in the pain processing pathways of the central nervous system has shed some light on the phenomenon.1 This new data indicates that placebo response rates seen in analgesic trials are exacerbated when these drugs are delivered via a transdermal delivery system and may have such a strong effect on endogenous pain modulation that, particularly for NSAIDs, the treatment effect is relatively inseparable from the effect of the physical product itself. Our proposal is that the new research should prompt changes in the way topical analgesic products are approved. Agencies would be advised to consider changes in the way products are approved.
Table 1. The FDA has approved few transdermal NSAID products for relatively narrow use in the United States.
While the involvement of anticipation and planning centers in the brain have long been known to influence placebo responses and descending inhibition of pain, it is only recently that direct involvement of spinal cord neurons plays a role in the analgesic placebo response. The involvement of the spinal cord in placebo analgesia was demonstrated by an experiment wherein dorsal horn neurons were sensitized by heating the skin. Visual analog scale (VAS) pain intensity scores in response to punctate stimuli (hyperalgesia) and stroking touch (allodynia) on the sensitized skin revealed that placebo analgesia reduced not only the VAS pain intensity scores, but also the size of the hyperalgesic and allodynic skin areas. Because the area of hyperalgesia/allodynia is determined by intensity-dependent spinal segmental mechanisms, the results suggest that the placebo effect was expressed at the spinal cord level as well as at supra-spinal levels.2, 3, 4
These observations are supported by the recent work of Eippert et al. (2009), who showed direct evidence for spinal cord involvement in the placebo analgesia response using blood oxygenation level-dependent (BOLD) fMRI imaging. Pain was induced via a thermode on the skin of subjects conditioned to expect a placebo cream to reduce pain sensitivity to thermal stimulation. Reduction in VAS pain intensity scores coincided with a reduction in spinal cord dorsal horn responses to painful stimulus.1 The data suggest that the expectation of placebo analgesia influences the activity of pain signaling at the most basic spinal levels. It is this expectation of treatment effect that provides a key connection between the physiological responses to placebo analgesia and the high placebo response rate seen in NSAID transdermal patch trials. While the influence of expectation can be applied to many drug delivery formats, it is particularly impactful for transdermal formulations, as will be clarified below.
Studies of the effect of expectation on placebo analgesia have demonstrated that, regardless of the conditioning agent used (opioid, NSAID, thermal conditioning, etc), the expectation of analgesic relief has a strong influence on the engagement of the endogenous opioid system to effect placebo analgesia relief.5, 6 Amanzio and Benedetti (1999) demonstrated that a verbal cue for expectation of pain relief increased the analgesic response to placebo when either morphine or ketorolac was used as a conditioning medium. Later, Zubietta et al. (2005) demonstrated that verbal cues used to enhance the expectancy of pain relief during placebo infusion effected a significant change in the μ-opioid activity of the anterior cingulate cortex and the dorsolateral prefrontal cortex, key brain regions involved in expectation and anticipation, and these results were supported by fMRI studies of the effects of expectations of placebo analgesia.7, 8, 9, 10
In a key study on the influence of expectancy versus conditioning procedures, Benedetti et al. (2003) demonstrated that verbal expectancy cues of analgesia or hyperalgesia in response to placebo injection could completely antagonize the effects of a conditioning procedure (ketorolac injection). In a similar study design, they showed similar expectancy effects on the improvement or worsening of motor effects in Parkinson's patients. However, hormone production showed no change in response to expectancy cues. These results suggest that expectancy cues have a much stronger influence on conscious responses, such as pain or motor movements, than on unconscious responses, such as hormone production.11
Table 2. Phase II-III studies for other transdermal NSAIDs have failed to meet FDA approval, owing largely to a failure to separate the treatment effect from placebo.
But what does this mean for transdermal NSAID clinical trials where verbal expectancy cues are limited to an informed consent that states only that the subject may receive active or placebo patches? Volkow et al. (2003) demonstrated that the expected administration of a drug has a more powerful effect on brain metabolism than an unexpected administration.12 Colloca et al. (2004) demonstrated that the open administration of a treatment in which the subject knows what is going on and expects an outcome is more effective than a hidden one, in which the subject does not know that any therapy is being given and thus does not expect anything.13 In addition, several studies have demonstrated that the procedure of surreptitiously reducing stimulus intensity before testing placebo analgesia (combining a conditioning stimulus with verbal expectancy) is more powerful than verbal expectancy alone.14, 15 In transdermal NSAID trials, subjects receive the expectation cue of pain relief from the informed consent and receive a conditioning stimulus in the combined form of a tactile stimulus (a gel/cream or patch) and previous pain relief experience after taking oral NSAIDs for treatment. Add in the conscious response effect (pain/pain relief), and transdermal NSAID trials are ideally designed to elicit a large placebo response rate.
Unlike oral analgesics, or even gels and creams, transdermal patches provide constant tactile and visual conditioning and expectancy reminders of treatment. As the above-mentioned studies have shown, treatment conditioning and expectancy stimuli influence the same CNS structures as active analgesic treatments; therefore, it may be impossible to separate the analgesia of the investigational drug from the analgesia produced by the method of drug delivery itself (the patch).
In addition to an acceptable safety profile, current FDA requirements for approval in the United States specify at least two studies that clearly demonstrate the superiority of the active product to placebo. However, even the few transdermal NSAID trials that have managed some separation of treatment effect from placebo have struggled against large placebo responses. While superiority to placebo is certainly the ideal and most direct method for determining efficacy, when the drug delivery device itself creates such a high response, then it may no longer be the best method by which efficacy is determined. In such cases, multiple studies incorporating indirect measures can serve to provide evidence of efficacy, and should be considered in the case of transdermal NSAIDs.
An elegantly designed study of Diractin®, a ketoprofen topical gel, provides a strong framework for the assessment of efficacy that could be applied to both patch and gel/cream formulations. The study utilized four treatment arms: topical active, topical placebo, oral active comparator (celecoxib), and oral placebo. Statistically significant non-inferiority results were revealed for all efficacy measures for Diractin compared to the oral active comparator. Diractin did not show a statistically significant difference compared to the topical placebo; however, both of the topical treatments (Diractin and placebo) and the oral celecoxib were statistically significantly superior to the oral placebo treatment.
This study design provides a useful combination of non-inferiority against a previously approved oral formulation while simultaneously addressing placebo responses for both formulations. As expected, the oral formulation produced a lower placebo response than the topical formulation, which provides a framework for examining efficacy versus the vehicle. Several other cream/gel NSAID products have used this study design for their Phase III trials, including the approved topical formulation of Voltaren, and could be productively applied to patch products as well. While some may argue that a study with a non-inferiority component and four treatment arms requires more patients, and therefore a greater financial expense, this should be viewed favorably in comparison to the financial and temporal expenditure associated with repeating Phase III efficacy trials to achieve separation from placebo.
To support an efficacy claim based on an indirect assessment such as the four-arm Diractin study, additional studies with a variety of indirect measures that assess treatment effect from different directions would be necessary. For example, employment of an epidemiological approach wherein the investigational product has to match a historical control level of a previously approved product could provide strong support for an efficacy claim. A pain model such as osteoarthritis (OA) of the knee could be employed in such a trial. Osteoarthritis of the knee is a well-characterized model such that, upon removal of standard anti-inflammatory treatment, patients experience a painful flare of their OA within two weeks of treatment cessation. In an enriched study design, patients with OA of the knee would be removed from their standard treatment, allowed to develop a flare, and patients with adequate flare would be enrolled. Following the flare, all patients would be assigned to an open-label active patch treatment for an adequate time frame and pain intensity recorded. At the end of the open-label period, patients would be assigned to either active or placebo patches and pain intensity and recurrence of flare would be recorded. This design would ensure that patients with adequate OA flares were enrolled while at the same time providing direct pain intensity measures and indirect flare prevention rates than can be compared to historical control levels.
Other indirect support of a transdermal NSAID product's efficacy could involve an imaging trial (e.g., fMRI) to demonstrate the product's ability to activate the key brain structures associated with descending pain control, namely, the PAG, RVM, and DLPT, and/or reduce activity in pain-responsive regions, such as the RACC, insular cortex, and thalamus. Zubieta et al. demonstrated the activation of endogenous μ-opioid neurotransmission in the brain using positron-emission tomography and a μ-opioid receptor selective radiotracer, which could be used to indirectly demonstrate efficacy through the activation of endogenous opioid transmission in the brain.6 At a more direct, functional level, BOLD fMRI can be used to visualize changes in pain signaling at the spinal level, which has been demonstrated by both Lilja et al. and Eippert et al.16
Yet another indirect method is a trial involving measures of product effectiveness, wherein patients are initially treated with the investigational product and then entered into an open-label period that allows the patient to treat a chronic pain condition (e.g., OA of the knee or low back pain) with either the investigational product or a rescue medication. This measures effectiveness through the patient's willingness to continue use of the product. There are numerous other indirect measures that, combined, can provide a body of evidence for a strong claim of efficacy.
Pharmaceutical manufacturers are attracted to transdermal formulations of NSAIDs because they provide a much stronger safety profile, particularly for gastrointestinal effects, than oral formulations, which is attractive to both patients and regulatory authorities. However, the safety advantage of these products cannot be realized if regulatory authorities resist the physiological impact of transdermal formulations on the placebo response and persist in applying an efficacy standard unsuitable for these products. Similarly, it is incumbent upon pharmaceutical manufacturers to develop new and supportable methods by which product efficacy can be judged by regulatory authorities.
The FDA and other regulatory authorities have a history of accepting alternative determinations of product efficacy when the use of a placebo arm has been deemed unethical. While in the United States, the use of a placebo arm in NSAID trials has not been deemed unethical, when data indicates that the method of placebo delivery itself dirties the response signal, it behooves all parties involved to open consideration of alternative methodologies. While the use of indirect measures would likely require a higher number of successful trials than the use of direct superiority-to-placebo trials, ultimately both regulatory agencies and pharmaceutical companies would benefit. Regulatory agencies would benefit through the increased variety of information on a product's activity upon which to base an approval and pharmaceutical companies would benefit through reduced expenditure on repeated Phase III trials in an effort to find just the right model, patient population, or inferential analysis that will ultimately provide clinically meaningful separation from placebo. In the end, the goal of pharmaceutical clinical testing is to demonstrate that the products brought to market actually deliver the efficacy claims made by the sponsoring company and to ensure reasonable safety of the products. Transdermal NSAID products can provide patients with both efficacy and safety if we focus on the goal of demonstration of efficacy rather than focusing on the red herring of separation from placebo, in which Phase III trials can become trapped.
Adrianne Ondarza*, PhD, is Project Specialist at INC Research, 3321 Bee Caves Rd. Austin, TX 78746, e-mail: email@example.com, Frederick Lewis, is an Independent Drug Development Consultant, and Tony Womack is Director at INC Research, Raleigh, NC.
*To whom all correspondence should be addressed.
1. F. Eippert, J. Finsterbusch, U. Bingel, and C. Büchel, "Direct Evidence for Spinal Cord Involvement in Placebo Analgesia," Science, 326 (5951) 404 (2009).
2. R. H. LaMotte, C.N. Shain, D.A. Simone, and E.F. Tsai, "Neurogenic Hyperalgesia: Psychophysical Studies of Underlying Mechanisms," Journal of Neurophysiology, 66 (1) 190-211 (1991).
3. F. Cervero, R. Gilbert, R.G. Hammond, and J. Tanner, "Development of Secondary Hyperalgesia Following Non-Painful Thermal Stimulation of the Skin: A Psychophysical Study In Man," Pain 54 (2) 181-189 (1993).
4. D. Matre, K.L. Casey, and S. Knardahl, "Placebo-Induced Changes in Spinal Cord Pain Processing," Journal of Neuroscience, 26 (2) 559-563 (2006).
5. M. Amanzio and F. Benedetti, "Neuropharmalogical Dissection of Placebo Analgesia: Expectation-Activated Opioid Systems Versus Conditioning-Activated Specific Subsystems," Journal of Neuroscience, 19 (1) 484-494 (1999).
6. J. K. Zubieta, J. A. Bueller, L. R. Jackson, D. J. Scott, Y. Xu, R. A. Koeppe, and C. S. Stohler, "Placebo Effects Mediated by Endogenous Opioid Neurotransmission and μ-Opioid Receptors," Journal of Neuroscience, 25 (34) 7754-7762 (2005).
7. M. I. Posner, "Orienting of Attention," Quarterly Journal of Experimental Psychology, 32 (1) 3-25 (1980).
8. A. Allport, "Visual Attention," Foundations of Cognitive Science, M. I. Posner, ed. (MIT, Cambridge, MA, 1989).
9. R. Desimone and J. Duncan, "Neural Mechanisms of Selective Visual Attention," Annual Review of Neuroscience, 18, 193-222 (1995).
10. T. C. Handy, V. Green, R. M. Klein, and G. R. Mangun, "Combined Expectancies: Event-Related Potentials Reveal the Early Benefits of Spatial Attention that are Obscured by Reaction Time Measures," Journal of Experimental Psychology, Human Perception, and Performance, 27 (2) 303-17 (2001).
11. F. Benedetti, A. Pollo, L. Lopiano, M. Lanotte, S. Vighetti, and I. Rainero, "Conscious Expectation and Unconscious Conditioning in Analgesic, Motor and Hormonal Placebo/Nocebo Responses," Journal of Neuroscience, 23 (10) 4315-4323 (2003).
12. N. D. Volkow, G. J. Wang, Y. Ma, J. S. Fowler, W. Zhu, L. Maynard, F. Telang, P. Vaska, Y. S. Ding, C. Wong, and J. M. Swanson, "Expectation Enhances the Regional Brain Metabolic and the Reinforcing Effects of Stimulants in Cocaine Abusers," Journal of Neuroscience, 23 (36), 11461-11468 (2003).
13. L. Colloca, L. Lopiano, M. Lanotte, and F. Benedetti, "Overt Versus Covert Treatment for Pain, Anxiety and Parkinson's Disease," Lancet Neurology, 3 (11) 679-684 (2004).
14. G. H. Montgomery and I. Kirsch, "Classical Conditioning and the Placebo Effect," Pain, 72 (1-2) 107-113 (1997).
15. D. D. Price, L. S. Milling, I. Kirsch, A. Duff, G. H. Montgomery, and S. S. Nicholls, "An Analysis of Factors that Contribute to the Magnitude of Placebo Analgesia in an Experimental Paradigm," Pain, 83 (2) 147-156 (1999).
16. J. Lilja, T. Endo, C. Hofstetter, E. Wetman, J. Young, L. Olson, and C. Spenger, "Blood Oxygenation Level-Dependent Visualization of Synaptic Relay Stations of Sensory Pathways Along the Neuroaxis in Response to Graded Sensory Stimulation of a Limb," Journal of Neuroscience, 26 (3) 6330-6336 (2006).
In order to appreciate the recent findings, the central nervous system (CNS) pathways involved in pain transmission, perception, and modulation should be examined. The CNS structures involved in the placebo-pain response comprise a complicated web of pain signaling pathways, midbrain pain-modulation nuclei (rostroventral medulla [RVM], dorsolateral pontine tegmentum [DLPT], periacqueductal gray area [PAG]), limbic system centers (hippocampus, amygdala, hypothalamus, nucleus accumbens), and higher processing loci of the cortex (dorsolateral prefrontal cortex [DLPFC], anterior cingulate cortex [ACC], insular cortex, neocortex). These structures are involved in both the ascending transmission and perception of pain signals, and the inhibition of pain transmission through descending pathways back to the spinal cord, most likely through the inhibition described by the Gate Control Theory of pain, introduced by Melzack and Wall in 1965.1
Spinal dorsal horn neurons receive pain input from cells in the dorsal root ganglia and send ascending projections to the RVM, PAG, hypothalamus, thalamus, and amygdala (both via direct projections and indirect projections through the parabrachial nucleus),2 among other neural structures. This initiates a complicated feedback loop of conscious and unconscious reactions that will lead to descending modulation of the pain signal.
The midbrain PAG and the RVM are two of the primary nuclei associated with the modulation of both ascending and descending control of pain signals and receive various inputs from both the limbic system and the planning/expectation cortices. The PAG receives input from the anterior cingulate and insular cortices (limbic system/attention/expectation); the amygdala (emotion, fear), which receives significant input from both the hippocampus (memory) and the neocortex (higher processing); the hypothalamus (hormonal control and behavior); the dorsolateral prefrontal cortex (attention, memory retrieval, complex cognitive behavior, anticipation/planning); and the nucleus accumbens (drive/reward/fear via the lateral hypothalamus and amygdala).2, 3, 4, 5 These brain areas form a network of reciprocal connections that, when activated by a nociceptive input (pain), work to create a descending response to the stimulus based on the emotional content of the stimulus, experience with the stimulus, the amount of attention paid to the stimulus, and the person's expectation of the response to the stimulus.
The PAG then projects to the RVM and the DLPT, and the RVM projects to the dorsal horn of the spinal cord to inhibit the transmission of peripheral pain signals, most likely via the Gate Control system described by Melzack and Wall using serotonin as a neurotransmitter in the spinal dorsal horn. The PAG-RVM connection is critical for pain modulation; the PAG projects only minimally to the spinal cord dorsal horn, and the pain-modulating action of the PAG on the spinal cord is relayed largely, if not exclusively, through the RVM. The PAG also projects rostrally to the medial thalamus and orbital frontal cortex, which raises the possibility of ascending control of nociception as well as descending control from the PAG, both of which could be key factors the placebo analgesia response.4, 6
Figure 1. Simplified diagram of the brain’s pathway connections that process the analgesia placebo response.
From the RVM, there are several possible circuits whereby RVM neurons could inhibit nociceptive transmission in the dorsal horn. There is evidence that descending pain modulatory neurons in the RVM and DLPT inhibit nociceptive transmission by several mechanisms: direct inhibition of projection neurons, inhibition of transmitter release from primary afferents, excitation of inhibitory interneurons, and inhibition of excitatory interneurons.5 One possibility is that brain stem neurons directly inhibit rostrally projecting nociceptive dorsal horn cells: electrical stimulation in the RVM produces a monosynaptic, inhibitory, postysynaptic potential in spinothalamic tract neurons, most likely through the release of serotonin.5, 7
While painful stimuli activate the above described pain-modulatory circuitry, analgesics also have an effect. The contribution of endogenous opioid peptides to pain modulation was first suggested by reports that stimulation-produced analgesia in animals and humans was reduced by the narcotic antagonist naloxone.5 In related experiments, morphine produced an analgesic effect when microinjected into the PAG, RVM, amygdala, or anterior insular cortex and microinjection of opioid antagonists into these areas reduced the analgesic effect of systemically administered opioids.8 The three classic opioid receptors (μ, δ, and κ) are each present in the insular cortex, amygdala, hypothalamus, PAG, RVM, and spinal cord dorsal horn. The highest levels of opioid receptors are found in the neocortex, hippocampus, amygdala, striatum, and superficial laminae of the dorsal horn. 9, 10 Endogenous opioids involved in pain modulation include leucine-enkephalin, methionine-enkephalin, ß-endorphin, dynorphin, endomorphin-1 and -2, and nociceptin.5 ß-endorphin containing neurons are present in the ventromedial hypothalamus and project to the PAG.11 Met- and leu-enkephalin containing cells are widely distributed in the amygdala, hypothalamus, PAG, RVM, and superficial dorsal horn.12 Thus, the physiological network for higher-order processing of pain stimuli has inherent checks in place, in the form of endogenous opioids, to modulate the conscious and sub-conscious response to painful stimuli.
Figure 2. Ascending pain pathway from the spinal cord to the rostroventral medulla, the periacqueductal gray, and the thalamus, and the descending inhibition pathways from the cortex back to the spinal cord.
The neuronal circuitry implicated in the mechanisms of the placebo response is similar to the pathways described for pain processing and pain modulation. Using blood oxygenation level-dependent (BOLD) fMRI, Wager et al. (2004) demonstrated an increased neural response to intense shock that revealed activation of the classic pain matrix: thalamus, neocortex, anterior cingulate cortex, and ventrolateral PFC. After application of placebo analgesia, there was a reduction in the magnitude of activation in pain-responsive structures (rACC, insula, thalamus) that was consistent with reductions in VAS-reported experiences of pain intensity. This provided evidence that placebo reduces brain activity in the pain matrix. Then they took the experiment a step further to show that placebo analgesia increases activity in areas of anticipation of pain stimulus (DLPFC and OFC). The negative correlation between the increased activity in the DLPFC and OFC and the decreased activity in the pain matrix in response to placebo analgesia implies a complimentary interaction of reduction of pain afferent signaling coincident with expectation of pain relief.13 That these experiments also found an increase in activity in the PAG, which contains a high concentration of opiate neurons with descending spinal efferents5 indicates a feedback mechanism that influences pain signaling at the spinal level, in addition to the cognitive effects.
1. R. Melzack and P. D. Wall, "Pain Mechanisms: A New Theory," Science, 150 (3699) 971-979 (1965).
2. C. Gauriau and J. F. Bernard, "Pain Pathways and Parabrachial Circuits in the Rat," Experimental Physiology, 87 (2) 251-258 (2002).
3. D. M. Hermann, P. H. Luppi, C. Peyron, et al., "Afferent Projections to the Rat Nuclei Raphe Magnus, Raphe Pallidus and Reticularis Gigantocellularis Pars Alpha Demonstrated by Iontophortic Application of Choleratoxin (Subunit B)," Journal of Chemical Neuroanatomy, 13 (1) 1-21 (1997).
4. A. A. Cameron, I. A. Khan, K. N. Westlund, et al., "The Efferent Projections of the Periaqueductal Gray in the Rat: A Phaseolus Vulgaris-Leucoagglutinin Study. I. Ascending Projections," Journal of Comparative Neurology, 351 (4) 568-584 (1995).
5. P. D. Wall and R. Melzack, Textbook of Pain, 5th ed., Chapter 7: China, 125-142, (Elsevier, Churchill Livingstone, Oxford, UK, 2005).
6. J. A. Coffield, K. K. Bowen, and V. Miletic, "Retrogarade Tracing of Projections Between the Nucleus Submedius, the Ventrolateral Orbital Cortex, and the Midbrain in the Rat," Journal of Comparative Neurology, 321 (3) 488-499 (1992).
7. W. D Willis Jr, "Anatomy and Physiology of Descending Control of Nociceptive Responses of Dorsal Horn Neurons: Comprehensive Review," Progress in Brain Research 77, 1-29 (1988).
8. T. L. Yaksh and T. A. Rudy, "Narcotic Analgesics: CNS Sites and Mechanisms of Action As Revealed by Intracerebral Injection Techniques," Pain, 4 (4) 299-359 (1978).
9. T. Darland, M. M. Heinricher, and D. K. Grandy, "Orphanin Fq/Nociceptin: A Role in Pain and Analgesia but So Much More," Trends in Neurosciences, 21 (5) 215-221 (1998).
10. A. Mansour, C. A. Fox, and S. J. Watson, "Opioid-Receptor Mrna Expression in the Rat CNS: Anatomical and Functional Implications," Trends in Neurosciences, 18 (1) 22-29 (1995).
11. M. J. Millan, "Descending Control of Pain," Progress in Neurobiology, 66 (6) 355-474 (2002).
12. H. Akil, F. Meng, D. P. Devine, et al., "Molecular and Neuroanatomical Properties of the Endogenous Opioid System: Implications for Treatment of Opiate Addiction," Seminars in Neuroscience, 9 (3-4) 70-83 (1997).
13. T. D. Wager, J. K. Rilling, E. E. Smith, A. Sokolik, K. L. Casey, R. J. Davidson, S. M. Kosslyn, R. M. Rose, and J. D. Cohen, "Placebo-Induced Changes in RMRI in the Anticipation and Experience of Pain," Science, 303 (5661), 1162-1167 (2004).