Predicting therapeutic efficacy — Experimental pain in human subjects

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / b r a i n r e s r e v

Review

Predicting therapeutic efficacy — Experimental pain in human subjects Boris A. Chizh a , Tony Priestley b,⁎, Michael Rowbotham c , Klaus Schaffler d a

GlaxoSmithKline, Addenbrooke's Centre for Clinical Investigation, Hills Road, Cambridge CB2 2GG, UK Schering-Plough Research Institute, Kenilworth, NJ 08807, USA c Department of Neurology, UCSF Pain Clinical Research Center, USA d HPR-Human Pharmacodynamic Research GmbH, Peschelanger 3, D-81735 Munich, Germany b

A R T I C LE I N FO

AB S T R A C T

Article history:

The pharmaceutical industry faces tough times. Despite tremendous advances in the

Accepted 29 December 2008

science and technology of new lead identification and optimization, attrition rates for novel

Available online 31 December 2008

drug candidates making it into the clinic remain unacceptably high. A seamless boundary between basic preclinical and clinical arms of the discovery process, embodying the concept

Keywords:

of ‘translational research’ is viewed by many as the way forward. The rational application of

Translational research

human experimental pain models in early clinical development is reviewed. Capsaicin, UV-

Human experimental pain

irradiation and electrical stimulation methods have each been used to establish

Capsaicin

experimental hyperalgesia in Phase-I human volunteers and the application of these

UV-irradiation

approaches is discussed in the context of several pharmacological examples. However, data

Clinical trial

generated from such studies must be integrated into a well-conceived and executed series of

Vertex-EEG

Phase-II efficacy trials in patients in order to derive maximal benefit. The challenges involved in optimal Phase-II/III trial design are reviewed with specific attention to the issues of sample size and placebo response. Finally, the application and potential of cortical EEG studies are discussed as an objective alternative to more conventional pain assessment tools with specific examples of how this technique has been applied to the study of NSAID and opiate-based therapeutics. © 2009 Elsevier B.V. All rights reserved.

Contents 1. 2. 3. 4.

Why we need a more efficient strategy for developing next-generation pain medications. . . . . . . . Rational use of human experimental pain models in early clinical development. . . . . . . . . . . . . Proof of concept studies in patients: can the limitations of sample size, patient selection, and placebo be overcome? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proof of efficacy and dose-dependency of NSAID and opiate analgesics in Laser evoked potential paradigm — using capsaicin and UVB skin model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . response . . . . . . .

246

. . . . . . .

248

⁎ Corresponding author. E-mail addresses: [email protected] (B.A. Chizh), [email protected] (T. Priestley), [email protected] (M. Rowbotham), [email protected] (K. Schaffler). 0165-0173/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.brainresrev.2008.12.016

244 245

244

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

5. Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

1. Why we need a more efficient strategy for developing next-generation pain medications The pharmaceutical industry has benefited from considerable advances in technology over the past decade, or so, and has become very adept at identifying new molecular entities for clinical development. Modern day compound repositories, automated compound handling and high-throughput screening have combined to speed up the drug discovery process, despite the increased burden of additional safety and other ‘off-target’ assay activities. Solid preclinical efficacy and safety typically enables progression of the candidate molecule into Phase I clinical trials — where toplevel safety, tolerability and pharmacokinetic properties are established in humans. A green light here usually secures progression to more complex, and expensive, Phase II studies directed towards establishing some evidence for therapeutic efficacy together with additional safety/tolerability data and with some dose-ranging consideration. These early clinical studies, though generally conducted with ‘small’ cohorts of volunteers (Phase I) or patients (Phase II) are time-consuming and expensive, but this is nothing compared to the time and expense of subsequent multiple Phase III trials. These ‘pivotal’ trials involve large numbers of patients, usually multiple sites and usually lengthy enrollment times and two or more successful Phase III trials are required for product registration, regulatory-authority approval and launch. The problem with the current process, as described above, is that the success rate for progression of a preclinical candidate to a successful therapeutic is pitifully low, with the vast majority of promising molecules failing to meet the grade once they move into the clinical assessment phases. High attrition rates are nothing new and have been a part of pharmaceutical life for many years. However, the everescalating cost of drug development has prompted a closer look at the likely reasons for such high clinical failure rates. A now dated, but nonetheless revealing pharmaceutical benchmarking study by the Certified Medical Representatives Institute (CMR, reviewed in detail by Kola and Landis, 2004) assessed attrition by phase and by therapeutic area. This analysis revealed the major dropouts occur in Phases II and III — suggesting that the main causes of failure were lack of therapeutic efficacy and safety/tolerability. Though the CMR data revealed some variability between therapeutic areas the bottom line was a general probability of success of around 11%, in other words less than one in ten compounds achieve approval. The assessment, by therapy area, revealed arthritis and pain attrition rates to be marginally better than the rolling average for all therapeutic classes. However, this picture may be a little misleading. The data collection period was between 1991 and 2000, a period that saw the launch of Celebrex and Vioxx, two of the then-novel COX-2 inhibitors, and a period when multiple opioid reformulation and combination products were launched. Much-publicized, ser-

ious issues have since emerged for the COX-2 therapeutics and the added value of opioid-based medications to the pain armamentarium, while valuable, can hardly be considered a leap forward, at least with respect to mechanistic novelty. Arguably, the only true advance in therapeutic options for neuropathic pain during, the above period, was marked by the successful launch of gabapentin (Neurontin, Pfizer). Overall, it could be argued that little comfort should be drawn from the apparently better attrition figures for pain. Given the fact that the pharmaceutical industry has had to live with high attrition rates for so long, why a sense of urgency over recent years to address this? Apart from the obvious fact that inefficiency is indefensible if avoidable, there is little doubt that the sheer cost of failure, in monetary terms, is simply unsustainable. Failed clinical trials produce an enormous drain on company funds. Past trends indicate a general unwillingness to accept initial negative data, such that additional trials with the same failed compound were commissioned with the consequential additional costs. Spiraling clinical costs impose pressure on all other parts of the research organization — research budgets are squeezed, as are those needed to ‘fund’ preclinical safety evaluation of upcoming clinical candidates. The result — most companies appear to have taken an introspective assessment of their discovery process and are implementing measures aimed at improving efficiency. These include: • Increased reliance on more extensive preclinical pharmacokinetic/pharmacodynamic (PK/PD) assessments during lead-compound optimization stages of a program. The aim is to establish greater confidence in a molecular series by demonstrating a concurrence between efficacy and exposure (drug levels in plasma, for example). The simple principle is that more credence can be attributed to an efficacy endpoint, an ED50 for example, when the ‘therapeutic’ plasma level is consistent with the likelihood of a meaningful interaction with the suggested molecular target. Efficacy observations at dose levels that are too low to hit the intended target, or so high that they greatly exceed the known drug-receptor affinity, should serve as an alert — though this need not amount to abandoning a particular line of investigation. • Identification of validated biomarkers for the sought therapeutic effect and the implementation of the biomarker strategy as part of Phase I activities. Ideally, a biomarker should be directly related to mechanism but in many cases this is simply not feasible and a surrogate endpoint may have to be used. Confidence in novel surrogate biomarkers may take time to accrue and these initiatives may need to be developed/included as an activity during ongoing preclinical studies. • Early interaction between preclinical, commercial, regulatory and clinical branches of the organization. This avoids potential ‘surprises’ by ensuring that the various arms of

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

245

the drug discovery program aligned, both with respect to initial expectations and also in dealing more effectively with program issues as they emerge.

2. Rational use of human experimental pain models in early clinical development

Much of the above is embodied in the concept of ‘translational medicine’, in which ongoing interactions between basic and clinical research seek to reduce risk. At the heart of the issue is the concept of construct validity — how predictive of the human experience are the animal models used to identify and characterize the proposed therapeutic molecule? In the pain field there is a clear dichotomy here; acute pain models fair very well, chronic pain models less so. Commonly-used neuropathic pain models such as spinal nerve ligation (SNL) and chronic constriction injury (CCI) models, for example, have been valuable. They may even replicate elements of pathology seen in clinical pain states but, clearly, they have questionable construct validity as regards post-herpetic neuralgia or painful diabetic neuropathy — to cite just two common patient populations used to assess potential new therapeutics. Another, and more-general issue, that impacts the animal to human interface reflects decades of history and habit that have resulted in an ever-increasing drive for preclinical uniformity. Studies are almost always performed on same sex, age-matched, inbred, same strain rodents and these contrast markedly with the human heterogeneity that is part of everyday clinical practice. Correcting this situation, though certainly feasible, is likely to be unpalatable because it will involve more complex and lengthy preclinical studies, it will generate less clear-cut data and it will raise defensible ethical issues in terms of animal usage. Thus, it is reasonable to say that the majority of rodent pain models, and certainly those pertaining to chronic pain, are not readily ‘translatable’ to humans and, of more concern, they are optimized for success. There have been some attempts to address these issues and, indeed, the following articles are testament to this. Nevertheless, more needs to be done and, arguably, a more pragmatic way forward may be for the preclinical arena to devote more attention to what can be measured in the clinic, so-called reverse translation, rather than the other way around. In summary, there is a growing need for an early proof-ofconcept (PoC) signal in pain research — some indication that the desired pharmacological effect has been achieved. To be of value in real terms, i.e. in reducing the costs and attrition incurred during Phase II/III trials, this information must, ideally, be obtainable as part of the Phase I campaign. PoC, of course, dictates that a relevant disease population be used for the studies and this may be a tall order for a Phase I stage study for a chronic pain condition. Failing PoC, a reasonable next-best approach would be robust proof-of-activity (PoA) and this could entail either the use of direct or surrogate biomarkers of target engagement. The essential question being asked is ‘does the drug get to the target organ and in sufficient concentration to achieve the desired functional response?’ Given the right tools, this is certainly something that could be achieved in Phase I. The above approaches, as part of well-conceived and executed human clinical trial campaign, should have the combined effect of providing confidence on efficacy and doseranging and should, therefore, ultimately improve efficiency and reduce attrition.

Bringing novel treatments to patients suffering from chronic pain remains a challenge despite advances in basic pain science. The highest hurdle seems to be achieving clinical efficacy in proof-of-concept (PoC) studies in pain patients. Several potential explanations can be given, but species differences in target pharmacology or tissue distribution, inadequate preclinical models or markers, failure to predict therapeutic index and incorrect dose selection seem the most likely. This dictates the need to increase the confidence in new compounds in early clinical development, before committing to large-scale patient studies. The present article reviews how experimental pain models and markers can be used in early clinical phases to achieve this. In Phase I, safety, tolerability and pharmacokinetics (PK) of potential drugs is assessed in humans to support further clinical work. The informative value of this development stage could be much increased if evidence of pharmacodynamic (PD) activity could be generated, confirming in humans the target pharmacology and mechanisms observed in preclinical studies. Ideally, PD markers should assess the level of target engagement in nociceptive pathways. Several examples can be considered showing the utility of this approach. Thermosensitive Transient Receptor Potential (TRP) receptors, particularly the TRPV1, have attracted a lot of interest as pain targets, and several potent and selective antagonists are currently in development for pain indications (Gunthorpe and Szallasi, 2008). The TRPV1 receptor can be activated by heat, capsaicin and inflammatory mediators, and psychophysical (pain threshold and tolerance) and peripheral afferent (capsaicin-evoked flare) responses to the target-specific challenges (heat, capsaicin, inflammation) have shown the ability to detect TRPV1 antagonism in humans (Chizh et al., 2007b). Nociceptive axon-reflex flare (either capsaicin- or electricallyevoked) and secondary hyperalgesia appear to be sensitive peripherally- and centrally-mediated markers, respectively, of sodium channel blockade in humans (Wallace et al., 1997; Koppert et al., 2000, 2001; Gottrup et al., 2000; Ando et al., 2000). Importantly, these effects in humans are observed at plasma concentrations closely matching the efficacious systemic exposure levels of sodium channel blockers in neuropathic pain patients (Challapalli et al., 2005). Capsaicin-evoked flare has recently been used as a translational PD marker of CGRP antagonism in preclinical and Phase I studies (de Hoon et al., 2007; Salvatore et al., 2008) to identify clinically efficacious doses of a new anti-migraine treatment (Doods et al., 2007). In addition to demonstrating PD activity, human pain models could provide hints of effects on specific pain mechanisms. Central sensitization is a key mechanism of chronic pain and can be induced in humans by peripheral afferent stimulation (using capsaicin, heat or electrical current). The electrical hyperalgesia model developed by Koppert and colleagues offers a good control over test duration, stability, a demonstrated central mechanism of sensitization and sensitivity to a number of analgesics (Koppert et al., 2001, 2004, 2005; Klede et al., 2003; Chizh et al., 2004). The procedure involves the insertion of a bipolar stimulation electrode

246

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

arrangement intradermally into, typically, the volar forearm of volunteers. Stimulation parameters may be titrated to achieve a roughly equivalent and stable inter-subject pain intensity and the effect of drugs can be evaluated against the background pain or against secondary hyperalgesia (an index of central sensitization) that can be probed using punctuate stimuli. Importantly, the model has been shown to detect efficacy of standard treatments of neuropathic pain, gabapentin and pregabalin, at standard clinical doses (Segerdahl, 2006; Chizh et al., 2007a). Furthermore, the negative predictive value of the model, i.e. the ability to detect lack of efficacy of drugs known not to relieve pain in patients, has been shown with the NK-1 antagonist aprepitant (Chizh et al., 2007a). For targets involved in inflammatory pain, the model of UV-evoked inflammation hyperalgesia could serve as an early efficacy indicator in humans. The model has been shown to involve a release of a wide range of inflammatory mediators including the prostaglandins, neuropeptides, cytokines and chemokines known to be key drivers of inflammatory pain (Black et al., 1980; Hruza and Pentland, 1993; Benrath et al., 1995; Saade et al., 2000; Averbeck et al., 2006; Angst et al., 2008). The human UV hyperalgesia model has been shown to detect efficacy of standard treatments of inflammatory pain including non-steroidal anti-inflammatory drugs, selective cyclooxygenase-2 inhibitors and opioids (Bickel et al., 1998; Sycha et al., 2003; Gustorff et al., 2004; Sycha et al., 2005). There are preclinical correlates of the human model (Saade et al., 2000; Bishop et al., 2007; Saade et al., 2008), which makes it possible to perform translational studies. In conclusion, human experimental pain models and PD markers can be used successfully to demonstrate target engagement and interaction with pain mechanisms in humans, and provide confidence and guide dose-selection prior to large scale patient studies. Such models have mechanistic limitations (e.g. the duration and type of induced plasticity, types of afferents involved, reliance on evoked measures); therefore, extensive pharmacological validation is required before they can be applied as efficacy filters. Nevertheless, if used rationally under the guidance of preclinical pharmacology and mechanism of action data, and with attention to certain logistical issues discussed below, such models could help bridge the gap between preclinical and patient stages of analgesic drug discovery.

3. Proof of concept studies in patients: can the limitations of sample size, patient selection, and placebo response be overcome? Phase 2A studies present many challenges in clinical trial design. They represent the first test of any new therapy in the population of interest for the planned therapeutic indication. Although the preclinical package nearly always encompasses a variety of animal pain models, each model is performed in small groups of animals that are nearly identical in their genetic makeup. In contrast, a small Phase 2A study of 60 patients can be thought of as the equivalent of testing 60 different animal strains. Phase 1 studies have made the task of the Phase 2A trial slightly easier with the greater use of experimental pain models and other procedures designed to assure that the biologic target

of interest is engaged. However, patients are different than healthy volunteer subjects in a number of important ways. Healthy volunteers have money as their primary motivation, and the phrase ‘guinea pigging’ refers to the fact that some healthy volunteers participate in clinical studies as their primary occupation (Elliott, 2008). Patients with chronic pain rarely have money as the primary reason for participating in a study. They want help with their ongoing problem, sometimes desperately so if extensive prior treatment attempts have been dismal failures. Other psychological factors vary, as do expectations of the results of the experimental therapy. Phase 2A studies represent an important go/no-go decision point in development. Budgets are often constrained, and project timelines may allow little time for recruitment of a large number of subjects. In addition, there may be competing, sometimes mutually exclusive study goals. For example, finding the maximum analgesic dose poses the risk of producing unacceptable side effects that could halt further development. Should the protocol be a ‘pressure test’ of a possible Phase 3 design or can the study team implement a more creative approach that maximizes information about performance of the compound? If so, how many procedures and assays can be added within the limits of the budget, acceptability by subjects, and the limits of the skills of the different investigators? This last aspect is particularly important, as the multicenter trial approach is the norm even at the Phase 2A stage. Most of the study sites will be ‘for profit’ organizations emphasizing speed and volume of subjects recruited and may not have the staff or the interest in performing complex assessments that require extensive training and large amounts of time. One of the first steps in designing any clinical trial is determining the number of subjects needed. To do this, an expectation of efficacy must be determined, along with the anticipated variability in the outcomes in individual patients. The literature will be searched for results with a desired reference compound tested under reasonably comparable conditions. How good a guide is the literature? The recent publication by Turner et al. (2008) provides persuasive evidence that the literature is a biased guide because of selective publication. Turner's study took advantage of two developments. First, a dozen non-tricyclic antidepressants have received marketing approval in recent years. Second, the FDA now publishes the SBA (Summary Basis of Approval) for each new drug receiving marketing approval on its web site, usually about a year after approval is granted. The SBA contains comprehensive summaries of all registration trials performed with the compound, even studies that did not show efficacy. By comparing the data in the SBA with the published literature, the investigators were able to determine the likelihood of publication of trials with differing outcomes. They used the FDA's assessment of each trial as either positive, negative, or questionable. Of the 74 studies examined, 38 were positive and 97% were published. Of the 24 negative studies, 67% went unpublished. Selective publication had a significant impact on the apparent efficacy of each drug, as measured by the effect size. For every one of the twelve antidepressants, the effect size was larger based on the published literature than from the full dataset contained in the FDA's SBA document. The net increase in effect size was 32%.

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

The problem of selective publication hampering efficacy estimates and the understanding of the responsiveness of specific clinical syndromes almost certainly extends to new treatments for pain. For example, the SBA for the painful diabetic neuropathy (DPN) indication of Pfizer's Lyrica (pregabalin) lists 5 Phase 3 trials. Only the 3 positive trials have been published in full format in a peer-reviewed journal; limited results from the negative trials have appeared at multiple scientific meetings in poster format. The problem extends further because the FDA can only place the SBA on its website for drugs that have received marketing approval and few compounds with new molecular structures have been approved in the past decade for pain. Dozens of compounds with novel structures and/or targets have undergone testing in Phase 2 or 3 before development stopped. In addition, some compounds marketed for another indication have undergone extensive clinical trials for an additional pain indication. In both situations, if there is no SBA or publication, the data is for all practical purposes permanently unavailable. Furthermore, many single center studies funded by investigator-initiated grants, the NIH, or foundations are performed. It is unknown to what extent selective publication occurs in these studies, but suffice it to say that journals are reluctant to publish negative results and investigators naturally prioritize publishing their positive studies. Sadly, everyone is harmed. Companies trying to develop new therapies will set performance goals inappropriately high and underpower their trials. Patients, who have contributed their time and exposed themselves to potentially severe adverse effects, will not gain the satisfaction that comes from knowing they have helped advance medical science. The only bright spot is the advent of the clinicaltrials.gov website, a service of the NIH. Major journals will no longer publish study results if the trial was not registered prior to enrolling the first study subject and there is an expectation that study results (at the least the ‘top line’ results) will be placed on the web site within a year of study completion. The variation in the response to placebo complicates planning clinical trials. A review by Katz et al. (2008) of 106 published trials (66% with positive outcomes) shows broad variability in the placebo response. In general, the greater the response to active the greater the response to placebo. Clinical trials of antidepressants for major depression have been plagued by enormous variability in the placebo response across clinical trials with the same drug and substantially similar designs. In some cases placebo response rates have been as high as 50–70%. For neuropathic pain, variability appears less extreme. For example, the 3 large failed trials with topiramate for DPN had pain intensity reductions of 25–32% during placebo treatment (Thienel et al., 2004). The 5 pregabalin trials for the same indication listed in the SBA had pain intensity reductions of 13, 18, 21, 29, and 30% during placebo treatment. The pain reduction of under 2% during placebo seen in the crossover trial of riluzole by Galer et al. (2000) must be about the lowest recorded placebo response; suffice it to say the active drug failed to show benefit. Including positive controls (drugs known to be active in the condition, such as an opioid or a gabapentinoid) as well as negative (placebo) controls provides an important reference for evaluating study results. If both the study drug and the positive control fail to show benefit compared to placebo, the study drug can't be considered a

247

failure. The downside of this approach, of course, is time and cost of recruiting more subjects. Furthermore, study inclusion/ exclusion criteria must take into account the risks of both the study drug and the positive control drug. Can the placebo response be managed? Pain intensity and pain relief are highly subjective constructs, especially in comparison to the outcome measures used in animal pain model studies where the outcomes are quantitative changes in evoked behaviors believed to represent pain. One possibility is to focus on disorders where sensory testing can be used to corroborate changes in spontaneous ongoing pain. In PHN, the vast majority of patients have touch-evoked pain (allodynia). In patients with allodynia, ongoing pain severity correlates with allodynia severity and showed little change over four weekly examinations (Rowbotham and Fields, 1996). There is some evidence to suggest that the placebo response is smaller for PHN than for DPN (Backonja et al., 1998; Rowbotham et al., 1998). My personal observation is that patients touch their area of pain and use the response as another way to judge the success or failure of therapy. However, PHN is much less common than DPN and large scale clinical trials have encountered increasing difficulty recruiting enough subjects to complete planned enrollments. Other modifications of the parallel study design might alter the response to placebo. A strategy that has been implemented in some studies is to eliminate all analgesic use prior to randomization. The limited data available thus far is that this strategy has only a slight beneficial impact on response (Dworkin et al., 2007). It complicates study recruitment by adding the time to withdraw study subjects from medications and presents the ethical issue of uncontrolled pain in those subjects randomized to placebo. The ability of study site personnel to subtly bias study results has received insufficient attention. There has long been a belief that providing a particularly supportive, even nurturing, study site environment might increase the response to placebo. Likewise, if the enthusiasm of investigators suggests to subjects that the study drug is particularly likely to be of benefit to them, their expectations of success could increase the response to placebo. Benedetti and colleagues have investigated the use of remote administration of study drug and placebo using an i.v. line extending into an another room so that subjects and study staff directly monitoring the subject do not know when treatment begins or ends (Colloca et al., 2004). This strategy appears to reduce the response to placebo (and also to active drug). The same strategy could be used in longer term trials of oral medications by using blister packs with variable starting and ending dates for active study drug, but such a design has not been adequately tested. Enriched enrollment designs have been employed in many clinical trials. The benefit is a gain in homogeneity of the subject population at the expense of generalizability of the results. Depending on the enrichment strategy, the added screening time and expense to select the final study population can prove disastrous if only a small minority of patients with a particular diagnosis meet the full set of criteria for entry. Genetic screens and pain phenotyping (specific sensory examination characteristics or setting limits on duration of disease are two examples) both have utility. Pharmacologic ‘probes’ are less often utilized. Three hypothetical variations are: (1) selecting

248

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

only those patients experiencing pain relief from iv infusion of a well known sodium channel blocker like lidocaine for a trial of an experimental subtype-selective blocker; (2) performing an iv fentanyl challenge before subjects enter an oral opioid trial; and (3) performing the ‘capsaicin response test’ before subjects enter a trial of a TRPV1 antagonist (Petersen et al., 2000). Attempts to use a single-blind placebo run-in period to find ‘placebo responders’ and eliminate them before starting the experimental drug have not been very successful. Trials of drugs for treatment of psychiatric disorders have successfully used a randomized withdrawal study design in which the study drug is first provided on an open label (non-blinded) basis. Subjects meeting response targets are then randomized into a more traditional parallel design, double-blind, placebo controlled trial. McQuay et al. (2008) have recently advocated for this design to be introduced into analgesic clinical trials. Crossover designs, although generally unacceptable to the FDA for Phase 3 trials, can be very useful in Phase 2A studies. The major disadvantage is the added time for each study subject, but subjects are more attracted to studies in which they know they will be exposed to the study medication. Each subject acts as their own control, and this tends to dampen both the response to placebo and active compound. Sample sizes can be reduced without loss of statistical power, with some caveats. The entire exercise is wasted if pain severity does not come back to baseline during the washout period between treatments and the study must then be analyzed as a parallel design trial using just the first study period data. A large number of drop outs during the second treatment period, especially if there is an asymmetry between active and placebo, also complicate analysis. Response may vary between the first and second periods because subjects are anxious as the study commences. In the second treatment period, subjects are conditioned by their experience during the first period. If they detect an obvious reduction in efficacy (or side effects) with the second study treatment, they may withdraw before completion. A crossover trial design that has rarely been used is the modified ‘n of 1’ approach in which each subject has multiple periods of exposure to the study drug. Laska and Sunshine (1973) demonstrated in a multi-session study the conditioning effect of the first exposure to treatment of pain on subsequent response to placebo in a highly influential 1973 study. The study published by Byas-Smith et al. (1995) represents a particularly complex example of a repeated exposures design. They studied clonidine for DPN through a two part approach. First, a three period crossover study was performed in which subjects received clonidine as one or two of the three treatments. Responders in that study were then invited to participate in a four period crossover study in which they received 7 days of clonidine at their highest previously tolerated dose during two of the four cycles. Final analysis showed that pain scores in this enriched group were significantly and consistently lower during clonidine treatment. An ingenious example of a novel approach to the placebo issue is the study by Fedele and colleagues, which has languished in obscurity since publication in 1989 (Fedele et al., 1989). They first exposed 152 women with primary dysmenorrhea to placebo at the appropriate time during their menstrual cycle under single-blind conditions. From this group, they selected

the 55 responders to placebo for participation in a double-blind trial of an NSAID during the next 4 menstrual cycles. In the first cycle of double-blind treatment, 84% of the placebo group responded compared to 96% of the NSAID group. In the second cycle, only 29% of the placebo compared to 83% of the NSAID group. During the 4th cycle, only 11% of the placebo group responded (with an additional 15% of the group refusing to participate due to futility). NSAID response stayed constant at 83%. The Fedele study shows in a particularly powerful manner that an initial response to placebo will gradually extinguish when repeat exposures prove to be a failure. Human experimental pain models are enjoying increasing use in Phase 1 studies to help assure target engagement and estimate dosage for Phase 2A trials. Although very imperfect as models of chronic pain, especially neuropathic pain, pain models can be incorporated into Phase 2A trials if they are relatively simple and non-invasive. One of the best examples is the recent placebo-controlled trial of smoked marijuana for HIV-neuropathy pain (Abrams et al., 2007). In this 7 day inpatient trial, subjects underwent the heat-capsaicin sensitization model (Petersen and Rowbotham, 1999) prior to smoking the first cigarette. The reduction in experimental hyperalgesia correlated with the reduction in ongoing neuropathic pain and provided a less subjective outcome to anchor interpretation of the primary efficacy measure. In summary, Phase 2A trials present great design challenges, but also provide an opportunity to creatively advance trial methodology. Selective publication represents missing data resulting in incorrect efficacy requirements and insufficient sample sizes. Placebo response risk cannot be completely eliminated but measures can be taken to increase the signal to noise ratio. There is no ‘free lunch’, however, because every technique to increase the information value of a study and reduce the response to placebo exacts a price in terms of study cost, subject recruitment rates, and time to study completion. There is no proven algorithm for managing the placebo response in Phase 2 trials; each molecule presents complexities best resolved by the clinical development team. Enriched enrollment schemes and crossover designs, including use of multiple periods of study drug exposure, can enhance study value. Simple pain models can be incorporated into large scale clinical trials, but the requirements of investigator training, time, and complex equipment mean that only small trials can utilize complex experimental pain models. The practical application of one of several possible human experimental pain procedures is discussed in detail below.

4. Proof of efficacy and dose-dependency of NSAID and opiate analgesics in Laser evoked potential paradigm — using capsaicin and UVB skin model Significant issues in the assessment of different clinical pain states arise mainly as a consequence of how patients' clinical impressions are measured. The quality and intensity parameters of an individual's sensory perception are only recorded in a subjective manner, usually via analogue or categorical scales or using pain diaries — but not in an objective/

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

quantitative approach. These measures are often confounded by co-medications and by additional pathophysiological, cognitive- and vigilance-based influences. Therefore such instruments are not ideally suited to the detailed investigation of pain processing — tracing its peripheral and central pathways, and elucidating the respective underlying mechanisms. The same holds for the evaluation of the general and specific efficacy of analgesic compounds. The human “blackbox” should allow additional enlightenment of diverse objectives — as efficacy principles (pharmacological targets), as well as time- and dose-efficacy of analgesic compounds. The application of laser technology as a novel, alternative or complimentary, technique in pain research (laser algesimetry) has resulted in a major advance in our ability to generate and interpret noxious thermal sensations. In pain measurements, the temperature at the skin surface has to be raised (N43 °C) in a very short time-frame to overcome the thresholds of the heat-sensitive pain receptors (thermonociceptors) and to open their heat-sensitive ionic channels and to avoid adaptation (desensitization). This may be achieved, very effectively, using a constant-duration (fixedinput length) nociceptive CO2-laser beam and with the added advantages of this being a contact-free approach that offers individually-adjusted intensity settings. The depth of penetration of the laser is low — resulting in a high receptorspecificity because the stimulus reaches only the free nociceptive terminals (features dictated by the beam's wavelength in the far infrared spectrum with a maximum absorption in water). These precise stimuli can be repeatedly applied, without habituation, on normal and irritated/hyperalgesic skin — e.g. by introducing UV-irradiation or topical capsaicin exposure. Analgesic and anti-hyperalgesic properties of drugs can be demonstrated objectively and quantitatively by alterations of the somatosensory evoked potential (SEP) parameters, measured using Vertex-EEG, mainly by reductions of signal amplitudes — e.g. vs. placebo. Stimulus-response specificity is determined by triggering, filtering and averaging of several artefact- and contact-free painful stimuli. The first two main EP-components (Fig. 1)

Table 1 – List of investigated typical and atypical analgesics by Laser-SEPs NSAIDs Celecoxib oral Dexketoprofen trometamol oral Diclofenac i.v. Dipyrone i.v. Etoricoxib oral Ibuprofen (raceme and enantiomer) Karprofen oral and i.v. Ketorolac oral Lysine clonixinate oral Naproxen Rofecoxib oral Parecoxib i.v. Valdecoxib oral ASA and ASA-like ASA oral EBA (Ethoxy-benzoic acid) solution Ethenzamide oral Eugenol solution Acetaminophen/Paracetamol Narcotics Cannabinoid agonists oral and i.v. Codeine oral Kappa agonist oral and i.v. Morphine oral Oxycodone oral Pentazocine oral and i.v. Remifentanil i.v. H1-antagonists Dimethindene oral Orphenadrine i.v. ReN1869 (amitriptyline-derived) oral Diphenhydramine oral Others Capsaicin antagonist oral Reboxetine (NARI) oral Ethanol oral and i.v. Flupirtine (pyridine derivative) oral Granisedron (5-HT3 agonist) ⁎) a Meprobamate oral a Propofol i.v. a Tapentadol oral Tramadol oral and i.v. Antiepileptics CM40907 oral Gabapentin oral Topicals Dimethindene gel 0.1% (antihistamine) DMSO solution with and without additives Lidocaine gel 2% Lidocaine solution and lozenges (oral cavity) ASA powder B1-antagonist (blocks bradykinin) a NK1-antagonist (blocks substance P) a a

Fig. 1 – Principle components of the Laser SEP (N1- and P2-peaks) with a typical overlay of analgesic vs. placebo waveforms.

249

Showing minor or no analgesic potency.

are evaluated with regard to their complex peak-to-peak (PtP) amplitude as well as with regard to the single N1component, mainly reflecting ‘peripheral’ effects (Schaffler et al., 1992), and P2-component, mainly reflecting ‘central’ effects in pain relief mechanisms (Schaffler et al., 1991). Analgesics of the peripheral type are preferably expected to depress the N1-amplitudes and to a lesser extent the P2amplitudes. Analgesics of the central type seem to depress

250

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

preferentially the P2-amplitude and to a minor extent the N1-amplitude. The laser-SEP can be used for different types of interpretations: • • •

Determination of sensory stimulus-intensity relationship (efficacy). Localizing type and pathways of pain-processing (peripheral and/or central preference). Quantification and comparison of the effects of different analgesics and analgesic combinations (general efficacy, onset, dose- and time-efficacy relationship).

Examples of the principal compound classes investigated, thus far, using the laser model include the following: NSAIDs (unspecific COX 1–2 and COX 2-selective inhibitors), opioids, opiates, antihistamines, antidepressants, antiepileptics, topicals and others (for additional details see Table 1).

•

The laser-SEP technique offers several advantages that are particularly relevant to the issue of increasing clinical trial efficiency, by decreasing overall risk: The laser approach enables us to work in small groups of healthy subjects (18 to 24 people) – due to a low variability –

Fig. 3 – (A) Laser-SEPs in normal skin conditions showing low hyperalgesia development over time and the lack of discrimination between drug (APAP) and placebo. (B) Laser-SEPs in UV-B irradiated skin conditions showing developing hyperalgesia and distinctly better discrimination of drug (APAP) effect vs. placebo.

Fig. 2 – (A) Time course of VAS pain score in post UV-irradiation hyperalgesia development (congruent to LSEP). (B) Time course of peak-to-peak (PtP) amplitude of Laser-SEP in post UV-irradiation hyperalgesia development (congruent to VAS).

and in an ethically acceptable intra-individual crossover approach, without additional impacts on suffering patients in clinical pain situations. • The risks of local adaptation, or sensitization, are low in this highly-selective stimulation of heat-sensitive (nociceptive) ionic channels — due to random positioning of the noxious laser beam. • Pharmacological effects of compounds can be evaluated on up to 3 different skin types simultaneously in each subject participating in the trial, and each person can contribute as his own control (low variability intra-individual crossover trials) in homogeneous repeated measurement designs. • Excellent congruency of subjective with objective-quantitative measures. An example is shown in Fig. 2, in which the temporal aspects of both subjective (VAS) pain score and objective (LSEP amplitude) measurements are compared in a UV-induced hyperalgesia paradigm for more than 24 h. This figure also illustrates the various options for analgesic treatment interventions: drug administration in

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

•

•

251

a pre-emptive, in an acute or steady-state of pain/ hyperalgesia conditions. Ability to discriminate analgesic and antihyperalgesic actions. This is nicely illustrated by comparing laser-evoked pain (VAS scores) for acetaminophen/paracetamol (APAP) in normal and UV-sensitized skin conditions. On normal skin (Fig. 3A) there is no obvious hyperalgesia development after multiple Laser exposures over the course of a day and acetaminophen/paracetamol (APAP) does not present effective analgesia in this non-sensitized skin condition. After application of UV-B (Fig. 3B) there exists a distinct hyperalgesic development with successive laser exposures over the assessment day and a significant antihyperalgesic effect of acetaminophen/paracetamol is discriminated vs. placebo. Ability to discern additional mechanistic information from drug action. The comparison of normal and UV skin conditions in laser PtP amplitudes adds additional information on drug action. For example, when compar-

Fig. 5 – (A) Regression of LSEP amplitude vs. VAS pain score reduction in capsaicin-sensitized skin (p.a. means). A meta-analysis of available data was performed to cluster diverse analgesics with known clinical efficacy to define the threshold of clinical efficacy. The threshold is about twice the limit of significant detectable reductions in LSEP amplitude, illustrating the sensitivity of the LSEP technique. (B) Same regression, but extended to a highly potent opiate (remifentanil) — demonstrating that the analgesic “clusters” show good alignment with treatment definitions of mild, moderate and severe pain states.

Fig. 4 – (A) Laser-SEPs in normal (non-inflamed) skin conditions. Note the stable placebo LSEP signal and the lack of habituation to thermo-nociceptive laser stimulation. Opioids show robust efficacy whereas NSAIDs are ineffective on normal skin nociceptive thresholds. (B) Laser-SEPs in UVB (inflamed) skin conditions. Note the time-dependent increase in placebo signals during thermo-nociceptive laser stimulation (indicating the development of hyperalgesia) and the good efficacy seen with both oral NSAID and oral opioid.

ing a NSAID and an opioid (Fig. 4A and B), the NSAID is – as to be expected – was not effective in the normal (noninflamed) skin paradigm, but opioid efficacy was evident. Furthermore, the stability of the responses to laser stimulation over the course of a day is nicely demonstrated in the normal skin/placebo condition (plateau), confirming the lack of habituation/tolerance development over the day. In contrast, both drugs were effective in the UV-sensitized skin condition, showing that the NSAIDs principal therapeutic domain are inflammatory states, but (as expected) the opioid was also effective in raising basic nociceptive thresholds. The application of LSEP to mainstream experimental pain studies is relatively new but the technique is gaining interest

252

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

and momentum. Benchmarking LSEP against more traditional ‘instruments’ will be crucial to the further development and application of this procedure — and some progress has been made in this regard. For example, a meta-analytical approach comparing peak-amplitude mean values of several established analgesics in studies using laser stimuli on capsaicin-sensitized skin were re-evaluated (regression) with respect to their dependency on laser amplitude and VAS pain score reduction (Fig. 5A). All drugs in the sample set clustered above a (N1–P2) peak-to-peak amplitude reduction of 2.5 μV or more (and a VAS score reduction of more than 10%). Statistically significant peak-to-peak amplitude differences were already reached with about half of these changes in LSEP amplitudes. The declared ‘starting level’ of clinical relevance for the respective parameters in this test paradigm was set at two-fold the statistically significance level (about 2.5 μV) and the data revealed that clinical experience fits quite well with the ranking of the drugs (Fig. 5A). Furthermore, Fig. 5B shows that the 3 groups/clusters are adequately treating the respective stages of pain (mild–moderate–severe) — e.g. Tramadol 50 mg–Tramadol 100 mg–Remifentanil (Schaffler, 2006). Thus laser-SEP represents a path to proof of concept by facilitating go/no-go decisions and distinctly reducing preparation time and the expenses of the subsequent, and more extensive, patient studies required for registration. The following pharmacological effects can be investigated in an ethically acceptable mode in healthy humans with an experimental (objective-quantitative, high-resolution) algesimetric approach — using a standardized and reproducible (thermonociceptive) pain induction with Laser-SEPs in different skin conditions: • • •

•

Efficacy and mechanisms in general Dose- and time-efficacy, onset of efficacy Mode of drug administration regimen (single- and multiple dose administrations or pre-emptive, acute, and (sub-) chronic intervention) Drug combination evaluations

The laser paradigm can also be considered as a tool of choice for a translational approach (animal–healthy subject– patient). Additional support for the laser model was recently bestowed by the “European Federation of Neurological Societies” (citation from the Laser Workshop at the NeuPSIG congress June 2007 in Berlin, see also Cruccu et al., 2004) – which indicated and recommended the usefulness of LSEP application in diagnostics of neuropathy – especially to distinguish between large and small fibre neuropathy. There are certain issues with the laser approach that warrant consideration. For example, application of the procedure to chemically-sensitized skin may be compromised in cases where the topical agent has a high water content (e.g. unguents and gels) because of the high absorption of the CO2-laser's infrared energy by water. However, this is usually a minor problem that may be overcome by washing and drying the skin prior to the stimulation procedure and if a more “permanent” sensitization is desired, the topical agent may be re-applied after the test session. The most significant limitation of the laser approach is that is cannot test the “mechan-

ical” aspects of pain (i.e. the mechano-sensory modality changes in an area of secondary hyperalgesia, for example). However, laser stimulation may be augmented by separate mechanical stimuli and both may be quantified using vertex SEPs, permitting both MSEP and LSEP readouts.

5.

Concluding remarks

A general consensus throughout the pharmaceutical industry favors implementation of a range of measures to reduce the high levels of attrition currently associated with drug development. This will likely involve rethinking both preclinical and clinical paradigms and a closer alignment of these to implement a more efficient translational approach. The ideas discussed above represent only some of the thoughts and approaches currently being developed. An important component, not discussed, will undoubtedly come from the many molecular and tissue imaging strategies currently being applied to pain research and several promising studies/ approaches have been describe recently (Ianetti et al., 2005; Borsook et al., 2007; Baliki et al., 2007). The pharmaceutical industry clearly has a number of options for PoA and PoC, and though there are hurdles ahead, the successful implementation of a revised approach to drug development will be driven by a combination of incentives that will include fiscal responsibility and scientific innovation.

REFERENCES

Abrams, D.I., Jay, C.A., Shade, S.B., Vizoso, H., Reda, H., Press, S., Kelly, M.E., Rowbotham, M.C., Petersen, K.L., 2007. Cannabis in painful HIV-associated sensory neuropathy: a randomized placebo-controlled trial. Neurology 68, 515–521. Ando, K., Wallace, M.S., Braun, J., Schulteis, G., 2000. Effect of oral mexiletine on capsaicin-induced allodynia and hyperalgesia: a double-blind, placebo-controlled, crossover study. Reg. Anesth. Pain Med. 25, 468–474. Angst, M.S., Clark, J.D., Carvalho, B., Tingle, M., Schmelz, M., Yeomans, D.C., 2008. Cytokine profile in human skin in response to experimental inflammation, noxious stimulation, and administration of a COX-inhibitor: a microdialysis study. Pain, in press. doi:10.1016/j.pain.2008.02.028 Averbeck, M., Beilharz, S., Bauer, M., Gebhardt, C., Hartmann, A., Hochleitner, K., Kauer, F., Voith, U., Simon, J.C., Termeer, C., 2006. In situ profiling and quantification of cytokines released during ultraviolet B-induced inflammation by combining dermal microdialysis and protein microarrays. Exp. Dermatol. 15, 447–454. Backonja, M., Beydoun, A., Edwards, K.R., Schwartz, S.L., Fonseca, V., Hes, M., LaMoreaux, L., Garofalo, E., 1998. Gabapentin for the symptomatic treatment of painful neuropathy in patients with diabetes mellitus: a randomized controlled trial. JAMA 280, 1831–1836. Baliki, M.N., Geha, P.Y., Apkarian, A.V., 2007. Spontaneous pain and brain activity in neuropathic pain: functional MRI and pharmacologic functional MRI studies. Curr. Pain Headache Rep. 11, 171–177. Benrath, J., Eschenfelder, C., Zimmerman, M., Gillardon, F., 1995. Calcitonin gene-related peptide, substance P and nitric oxide are involved in cutaneous inflammation following ultraviolet irradiation. Eur. J. Pharmacol. 293, 87–96.

B RA I N RE SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

Bickel, A., Dorfs, S., Schmelz, M., Forster, C., Uhl, W., Handwerker, H.O., 1998. Effects of antihyperalgesic drugs on experimentally induced hyperalgesia in man. Pain 76, 317–325. Bishop, T., Hewson, D.W., Yip, P.K., Fahey, M.S., Dawbarn, D., Young, A.R., McMahon, S.B., 2007. Characterisation of ultraviolet-B-induced inflammation as a model of hyperalgesia in the rat. Pain 131, 70–82. Black, A.K., Fincham, N., Greaves, M.W., Hensby, C.N., 1980. Time course changes in levels of arachidonic acid and prostaglandins D2, E2, F2 alpha in human skin following ultraviolet B irradiation. Br. J. Clin. Pharmacol. 10, 453–457. Borsook, D., Moulton, E.A., Schmidt, K.F., Becerra, L.R., 2007. Neuroimaging revolutionizes therapeutic approaches to chronic pain. Mol. Pain 3, 1–8. Byas-Smith, M.G., Max, M.B., Muir, J., Kingman, A., 1995. Transdermal clonidine compared to placebo in painful diabetic neuropathy using a two-stage ‘enriched enrollment’ design. Pain 60, 267–274. Challapalli, V., Tremont-Lukats, I.W., McNicol, E.D., Lau, J., Carr, D.B., 2005. Systemic administration of local anesthetic agents to relieve neuropathic pain. Cochrane Database Syst. Rev. CD003345. Chizh, B.A., Dusch, M., Puthawala, M., Schmelz, M., Cookson, L.M., Martina, R., Brown, J., Koppert, W., 2004. The effect of intravenous infusion of adenosine on electrically-evoked hyperalgesia in a healthy volunteer model of central sensitization. Anesth. Analg. 99, 816–822. Chizh, B.A., Gohring, M., Troster, A., Quartey, G.K., Schmelz, M., Koppert, W., 2007a. Effects of oral pregabalin and aprepitant on pain and central sensitization in the electrical hyperalgesia model in human volunteers. Br. J. Anaesth. 98, 246–254. Chizh, B.A., O'Donnell, M.B., Napolitano, A., Wang, J., Brooke, A.C., Aylott, M.C., Bullman, J.N., Gray, E.J., Lai, R.Y., Williams, P.M., Appleby, J.M., 2007b. The effects of the TRPV1 antagonist SB-705498 on TRPV1 receptor-mediated activity and inflammatory hyperalgesia in humans. Pain 132, 132–141. Colloca, L., Lopiano, L., Lanotte, M., Benedetti, F., 2004. Overt versus covert treatment for pain, anxiety, and Parkinson's disease. Lancet Neurol. 3, 679–684. Cruccu, G., Anand, P., Attal, N., Garcia-Larrea, L., Haanpaa, M., Jorum, E., et al., 2004. EFNS guidelines on neuropathic pain assessment. Eur. J. Neurol. 11, 153–162. de Hoon, J., de Lepeleire, I., Hickey, L., Kane, S.A., Murphy, M.G., Palcza, J., Sinclair, S., Vanmolkot, F., Willson, K., Xiao, A., Xu, Y., 2007. Inhibition of capsaicin-induced increase in dermal microvascular blood flow by the oral CGRP antagonist, MK-0974. Cephalalgia 27, #F060. Doods, H., Arndt, K., Rudolf, K., Just, S., 2007. CGRP antagonists: unravelling the role of CGRP in migraine. Trends Pharmacol. Sci. 28, 580–587. Dworkin, R.H., Thakur, R., Griesing, T., Sharma, U., Young, J.P., 2007. Randomized clinical trials of pregabalin for neuropathic pain: methods, results, and implications. Prog. Neurotherapeutics Neuropsychopharmacol. 1–21. Elliott, C., January 7, 2008. Guinea-Pigging. The New Yorker. Fedele, L., Marchini, M., Acaia, B., Garagiola, U., Tiengo, M., 1989. Dynamics and significance of placebo response in primary dysmenorrhea. Pain 36, 43–47. Galer, B.S., Twilling, L.L., Harle, J., Cluff, R.S., Friedman, E., Rowbotham, M.C., 2000. Lack of efficacy of riluzole in the treatment of peripheral neuropathic pain conditions. Neurology 55, 971–975. Gottrup, H., Hansen, P.O., Arendt-Nielsen, L., Jensen, T.S., 2000. Differential effects of systemically administered ketamine and lidocaine on dynamic and static hyperalgesia induced by intradermal capsaicin in humans. Br. J. Anaesth. 84, 155–162. Gunthorpe, M.J., Szallasi, A., 2008. Peripheral TRPV1 receptors as targets for drug development: new molecules and mechanisms. Curr. Pharm. Des. 14, 32–41.

253

Gustorff, B., Hoechtl, K., Sycha, T., Felouzis, E., Lehr, S., Kress, H.G., 2004. The effects of remifentanil and gabapentin on hyperalgesia in a new extended inflammatory skin pain model in healthy volunteers. Anesth. Analg. 98, 401–407. Hruza, L.L., Pentland, A.P., 1993. Mechanisms of UV-induced inflammation. J. Invest. Dermatol. 100, 35S–41S. Iannetti, G.D., Zambreanu, L., Wise, R.G., Buchanan, T.J., Huggins, J.P., Smart, T.S., Vennart, W., Tracey, I., 2005. Pharmacological modulation of pain-related brain activity during normal and central sensitization states in humans. Proc. Natl. Acad. Sci. 102, 18195–18200. Katz, J., Finnerup, N.B., Dworkin, R.H., 2008. Clinical trial outcome in neuropathic pain: relationship to study characteristics. Neurology 70, 263–272. Klede, M., Handwerker, H.O., Schmelz, M., 2003. Central origin of secondary mechanical hyperalgesia. J. Neurophysiol. 90, 353–359. Kola, I., Landis, J., 2004. Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug Discov. 3, 711–715. Koppert, W., Ostermeier, N., Sittl, R., Weidner, C., Schmelz, M., 2000. Low-dose lidocaine reduces secondary hyperalgesia by a central mode of action. Pain 85, 217–224. Koppert, W., Dern, S.K., Sittl, R., Albrecht, S., Schuttler, J., Schmelz, M., 2001. A new model of electrically evoked pain and hyperalgesia in human skin: the effects of intravenous alfentanil, S(+)-ketamine, and lidocaine. Anesthesiology 95, 395–402. Koppert, W., Wehrfritz, A., Korber, N., Sittl, R., Albrecht, S., Schuttler, J., Schmelz, M., 2004. The cyclooxygenase isozyme inhibitors parecoxib and paracetamol reduce central hyperalgesia in humans. Pain 108, 148–153. Koppert, W., Ihmsen, H., Korber, N., Wehrfritz, A., Sittl, R., Schmelz, M., Schuttler, J., 2005. Different profiles of buprenorphine-induced analgesia and antihyperalgesia in a human pain model. Pain 118, 15–22. Laska, E., Sunshine, A., 1973. Anticipation of analgesia. A placebo effect. Headache 13, 1–11. McQuay, H.J., Derry, S., Moore, R.A., Poulain, P., Legout, V., 2008. Enriched enrolment with randomised withdrawal (EERW): time for a new look at clinical trial design in chronic pain. Pain 135, 217–220. Petersen, K.L., Rowbotham, M.C., 1999. A new human experimental pain model: the heat/capsaicin sensitization model. Neuroreport. 10, 1511–1516. Petersen, K.L., Fields, H.L., Brennum, J., Sandroni, P., Rowbotham, M.C., 2000. Capsaicin evoked pain and allodynia in postherpetic neuralgia. Pain 88, 125–133. Rowbotham, M.C., Fields, H.L., 1996. The relationship of pain, allodynia and thermal sensation in post-herpetic neuralgia. Brain 119, 347–354. Rowbotham, M., Harden, N., Stacey, B., Bernstein, P., Magnus-Miller, L., 1998. Gabapentin for the treatment of postherpetic neuralgia: a randomized controlled trial. J.A.M.A. 280, 1837–1842. Saade, N.E., Nasr, I.W., Massaad, C.A., Safieh-Garabedian, B., Jabbur, S.J., Kanaan, S.A., 2000. Modulation of ultraviolet-induced hyperalgesia and cytokine upregulation by interleukins 10 and 13. Br. J. Pharmacol. 131, 1317–1324. Saade, N.E., Farhat, O., Rahal, O., Safieh-Garabedian, B., Le Bars, D., Jabbur, S.J., 2008. Ultra violet-induced localized inflammatory hyperalgesia in awake rats and the role of sensory and sympathetic innervation of the skin. Brain Behav. Immun. 22, 245–256. Salvatore, C.A., Hershey, J.C., Corcoran, H.A., Fay, J.F., Johnston, V.K., Moore, E.L., Mosser, S.D., Burgey, C.S., Paone, D.V., Shaw, A.W., Graham, S.L., Vacca, J.P., Williams, T.M., Koblan, K.S., Kane, S.A., 2008. Pharmacological characterization of MK-0974 [N-[(3R,6S)-6-(2,3-difluorophenyl)-2-oxo-1-(2,2,2-trifluoroethyl) azepan-3-yl]-4-(2-oxo-2,3-dihydro-1H-imidazo[4,5-b]pyridin-

254

B RA I N R E SE A R CH RE V I EW S 60 ( 20 0 9 ) 2 4 3– 2 5 4

1-yl)piperidine-1-carboxamide], a potent and orally active calcitonin gene-related peptide receptor antagonist for the treatment of migraine. J. Pharmacol. Exp. Ther. 324, 416–421. Schaffler, K., 2006. Drug effects on laser-evoked nociceptive eeg-responses obtained from uv- and capsaicin-irritated skin of healthy humans. Keystone Symposia: “Pain Mechanisms And The Development of Analgesics” 11.-16.06.2006. Keystone Resort, Colorado USA. Schaffler, K., Wauschkuhn, C.H., Gierend, M., 1991. Analgesic potency of a new anticonvulsant drug versus acetylsalicylic acid via laser somatosensory evoked potentials. Randomised placebo-controlled double-blind (5-way) crossover study. Arzneim.-Forsch. /Drug Res. 41 (4), 427–435. Schaffler, K., Wauschkuhn, C.H., Brunnauer, H., Rehn, D., 1992. Evaluation of the local anaesthetic activity of dimethindene maleate by means of laser algesimetry in healthy volunteers. Arzneim.-Forsch. /Drug Res. 42 (II) (Nr. 11), 1332–1335. Segerdahl, M., 2006. Multiple dose gabapentin attenuates cutaneous pain and central sensitisation but not muscle pain in healthy volunteers. Pain 125, 158–164.

Sycha, T., Gustorff, B., Lehr, S., Tanew, A., Eichler, H.G., Schmetterer, L., 2003. A simple pain model for the evaluation of analgesic effects of NSAIDs in healthy subjects. Br. J. Clin. Pharmacol. 56, 165–172. Sycha, T., Anzenhofer, S., Lehr, S., Schmetterer, L., Chizh, B., Eichler, H.G., Gustorff, B., 2005. Rofecoxib attenuates both primary and secondary inflammatory hyperalgesia: a randomized, double blinded, placebo controlled crossover trial in the UV-B pain model. Pain 113, 316–322. Thienel, U., Neto, W., Schwabe, S.K., Vijapurkar, U., 2004. Topiramate in painful diabetic polyneuropathy: findings from three double-blind placebo-controlled trials. Acta Neurol. Scand. 110, 221–231. Turner, E.H., Matthews, A.M., Linardatos, E., Tell, R.A., Rosenthal, R., 2008. Selective publication of antidepressant trials and its influence on apparent efficacy. N. Eng. J. Med. 358, 252–260. Wallace, M.S., Laitin, S., Licht, D., Yaksh, T.L., 1997. Concentration–effect relations for intravenous lidocaine infusions in human volunteers: effects on acute sensory thresholds and capsaicin-evoked hyperpathia. Anesthesiology 86, 1262–1272.