Endogenous Opioid Antagonism in Physiological Experimental Pain Models: A Systematic Review

Opioid antagonists are pharmacological tools applied as an indirect measure to detect activation of the endogenous opioid system (EOS) in experimental pain models. The objective of this systematic review was to examine the effect of mu-opioid-receptor (MOR) antagonists in placebo-controlled, double-blind studies using ʻinhibitoryʼ or ʻsensitizingʼ, physiological test paradigms in healthy human subjects. The databases PubMed and Embase were searched according to predefined criteria. Out of a total of 2,142 records, 63 studies (1,477 subjects [male/female ratio = 1.5]) were considered relevant. Twenty-five studies utilized ʻinhibitoryʼ test paradigms (ITP) and 38 studies utilized ʻsensitizingʼ test paradigms (STP). The ITP-studies were characterized as conditioning modulation models (22 studies) and repetitive transcranial magnetic stimulation models (rTMS; 3 studies), and, the STP-studies as secondary hyperalgesia models (6 studies), ʻpainʼ models (25 studies), summation models (2 studies), nociceptive reflex models (3 studies) and miscellaneous models (2 studies). A consistent reversal of analgesia by a MOR-antagonist was demonstrated in 10 of the 25 ITP-studies, including stress-induced analgesia and rTMS. In the remaining 14 conditioning modulation studies either absence of effects or ambiguous effects by MOR-antagonists, were observed. In the STP-studies, no effect of the opioid-blockade could be demonstrated in 5 out of 6 secondary hyperalgesia studies. The direction of MOR-antagonist dependent effects upon pain ratings, threshold assessments and somatosensory evoked potentials (SSEP), did not appear consistent in 28 out of 32 ʻpainʼ model studies. In conclusion, only in 2 experimental human pain models, i.e., stress-induced analgesia and rTMS, administration of MOR-antagonist demonstrated a consistent effect, presumably mediated by an EOS-dependent mechanisms of analgesia and hyperalgesia.


Introduction
Human experimental pain models are essential in physiological and pharmacological research, testing hypothetical pain mechanisms, forward-translating observations from animal research or establishing evidence of analgesic drug efficacy.A number of receptor-specific agonists and antagonists are utilized as adjuncts investigating physiologic mechanisms behind pain inhibition and pain sensitization.Research has focused on various receptors, e.g., α 2 -receptors, 5-HT 1A -receptors, NMDA-receptors and TRPV1-receptors, but above all, major interest has been dedicated to the endogenous mu-opioid-receptor (MOR).Selective MOR-antagonists have been used in a large number of human experimental  and clinical studies [64].Early animal data demonstrated that MOR-antagonists increase nociceptive responding across various stimulation paradigms and species [61].Subsequent studies in monkeys and humans showed that microinjections of morphine [65] or electrical stimulation [66] of the periaqueductal grey area (PAG) produced marked analgesia, which could effectively be antagonized by systemic administration of naloxone [67].
In human experimental pain models the research involving MOR-antagonists has primarily focused on pain thresholds and tolerance to pain stimuli, conceptualizing the idea that activity of the EOS hypothetically could be responsible for an attenuation of the responses to pain [43].Consequently the administration of MOR-antagonist could indirectly substantiate or question the involvement of the EOS in acute experimental pain perception.Since results from the literature on the effect of MOR-antagonists on experimental pain seem ambiguous [57,61], the authors decided to undertake a systematic review separating the search data into studies utilizing 'inhibitory' test paradigms and 'sensitizing' test paradigms.The main objective was to examine if certain physiological stimulation paradigms, techniques or methods could be modulated by naloxone or naltrexone, which is considered presumptive evidence of activation of the EOS.The primary outcomes were direct measures of experimental pain perception (pain ratings, pain thresholds, pain tolerance, hyperalgesia) or indirect measures of nociception (neuroimaging responses [BOLD (blood-oxygen-level dependent) contrast imaging, fMRI, PET], nociceptive reflexes [NRF], somatosensory evoked potentials [SSEP]).The secondary outcomes were autonomic measures of pain and nociception (autonomic, hemodynamic and neuroendocrine responses).

Registration and Search Strategy
The review was registered in the PROSPERO international database (CRD42014013102; http:// www.crd.york.ac.uk/PROSPERO/DisplayPDF.php?ID=CRD42014013102).Only placebo-controlled, double-blind, experimental studies, including healthy human subjects, examining the effect of MOR-antagonists on pain inhibition and pain sensitization, were considered.It was required that the studies employed physiological stimuli, i.e., chemical, electrical, mechanical, pharmacological, thermal or a combination of stimuli.Psychological conditioning stimuli, often applied in placebo or behavioral studies, were not included in this review.Studies primarily concerning acupuncture, cardiovascular reactivity, clinical outcomes, endocrine functions, psychological or psychiatric outcomes and substance abuse, as well as, non-English studies, abstracts from scientific meetings and material from textbooks were not included.Studies with opioid-administration prior to administration of the MOR-antagonist were not included.
A literature search (LPHA, MPP, MUW) was performed in the databases PubMed and EMBASE (search completed August 8, 2014) using the following search terms: (pain OR pain measurement OR pain threshold OR pain perception OR pain sensitization OR pain inhibition OR pain summation OR pain conditioning OR pain habituation OR pain modulation OR secondary hyperalgesia OR hyperalgesia OR diffuse noxious inhibitory controls OR diffuse noxious inhibitory control OR DNIC) AND (levallorphan OR naloxone OR naltrexone OR methyl-naltrexone OR alvimopan OR diprenorphine OR meptazinol OR Receptors, Opioid, mu/antagonists and inhibitors OR mu-opioid receptor antagonist OR mu opiate receptor antagonist) AND (healthy OR subjects OR control group OR normal OR normals OR doubleblind placebo controlled OR double-blind method).Reference-lists from retrieved studies were searched for additional relevant material (MUW).No contact with study authors to identify additional studies was made.In case of uncertainty concerning relevance of an article, the subject was discussed between the authors and a final decision was taken by the senior author (MUW).From the 2,142 records 86 full-text articles were assessed for eligibility.Sixty-three relevant studies were included in the review (Fig 1 : PRISMA 2009 Flow Diagram).Assessing risk of bias was made by the Oxford quality scoring system [68] (MPP, MUW).Descriptive data and outcome data were extracted from these studies and accumulated in tables (MUW) and verified independently (MPP, LPHA).The PRISMA 2009 Checklist is in a supporting file (S1 PRISMA Checklist).

Definitions
Preliminary examination of the retrieved studies indicated that a classification of the studies into 'inhibitory' and 'sensitizing' test paradigms would facilitate the presentation and interpretation of data.
2.2.1 'Inhibitory' Test Paradigms (ITP).ITP-studies were characterized by implementation of a noxious or non-noxious inhibitory conditioning stimulus (Fig 2, upper panel; stressinduced analgesia [SIA], spatial summation induced conditioning, diffuse noxious inhibitory control [DNIC], heterotopic noxious conditioning stimulation, conditioned pain modulation [CPM], repetitive noxious stimulation, non-noxious frequency modulated peripheral conditioning and repetitive transcranial magnetic stimulation [rTMS]) [69].The test-stimulus (Fig 2 ) was applied heterotopically, at a site different from the site of the conditioning stimulus, or homotopically, at the same site as the conditioning stimulus, where the test stimulus became an integrated part of the conditioning stimulus [19].The response to the test-stimulus was evaluated by psychophysical measures, e.g., pain ratings, pain threshold and pain tolerance assessments, or physiological measures, e.g., the spinal nociceptive flexion reflex (RIII; Fig 2) [70].The conditioning inhibitory effect was evaluated by the associated decrease in the response to the test-stimulus: 4test-stimulus (Fig 2).MOR-antagonist was administered in order to indirectly uncover an EOS-dependent mechanism in the conditioning response: if the 4test-stimulus was attenuated by the MOR-antagonist, a role of the EOS was presumed.In all the studies the outcomes were evaluated against baseline conditions and placebo-controls.
2.2.2 'Sensitizing' Test Paradigms (STP).STP-studies were characterized by implementation of a pain stimulus leading to quantifiable, 'sensitizing', nociceptive responses, i.e., changes in behavioral measures (hyperalgesia, pain ratings, thresholds, pain tolerance), thresholds of nociceptive reflexes, SSEP, or, miscellaneous neuroimaging or neuroendocrine variables (Fig 2, lower panel).In a number of the STP-studies an additional conditioning stimulus was applied, e.g., a burn injury [31] or capsaicin [35,36], enhancing the nociceptive response.MOR-antagonists were administered in order to indirectly uncover an EOS-dependent mechanism in the 'sensitizing' nociceptive response: if the response was enhanced by the MOR-antagonist, an inhibitory role of the EOS was presumed.In all the studies the outcomes were evaluated against baseline conditions and placebo controls.
2.2.3Habituation and Sensitization.The phenomenon by which repeated identical stimuli elicit progressively decrements in responses has been operationally defined as habituation  The ITPstudies employed an inhibitory conditioning stimulus with evaluation of the associated change in the applied test-stimulus (4test-stimulus).The objective of the ITP-studies was to examine the effect of mu-opioid-receptor (MOR) antagonist on the magnitude of the 4test-stimulus, indicating an activation of the endogenous opioid system (EOS) responsible for the conditioning response leading to antinociception/hypoalgesia (the central rectangle [Opioid-dependent mechanism?]indicates a hypothetical augmentation of the conditioning response by the EOS).The STP-studies (lower panel) employed a pain stimulus leading to quantifiable 'sensitizing' CNS-responses, e.g., changes in behavioral measures (hyperalgesia, pain ratings, thresholds, tolerance), nociceptive [71].The phenomenon by which repeated identical stimuli elicit progressively increments in responses is here defined as sensitization.

MOR-antagonists
The MOR-antagonists used in human research are alvimopan, diprenorphine, methylnaltrexone, naloxone and naltrexone.In addition, MOR-antagonists, or MOR-antagonists with partial κ-agonist effects, levallorphan, meptazinol and nalorphine, have been used in opioid blocking research.In the retrieved ITP-and STP-studies only naloxone and naltrexone were used.
Naloxone and naltrexone are non-specific opioid-antagonists with high affinity for the MOR [72].Both drugs cross the blood-brain barrier and demonstrate central opioid-blocking effects, in contrast to the peripherally acting MOR-antagonists, e.g., alvimopan and methylnaltrexone.Due to low systemic bioavailability of naloxone after oral administration, i.e., 2-3% [73], naloxone is given parenterally, when systemic opioid-blocking effects are required.In adults the distribution half-life (T ½α ) is 40 to 70 seconds [74], and the elimination (T ½β ) halflife is 54 to 64 min [74,75].Naloxone, with a rapid onset and short duration of action, is suited for acute management of opioid-induced serious adverse effects [24] and is administered in IV doses of 0.04 mg to 0.4 mg [76].Interestingly, naloxone expresses a dose-dependent, biphasic response with low doses producing analgesia and high doses producing hyperalgesia, both in animal inflammatory models [77] and in clinical models [78][79][80].
Naltrexone has a systemic bioavailability after oral administration of 5% to 60% [81] and since its main use clinically is treatment of substance dependence, the oral route is preferred.The elimination half-life of naltrexone and its active metabolite 6-beta-naltrexol, after oral administration is 4 to 10 hours [82].Naltrexone is clinically given in daily doses of 50 to 100 mg.

Literature Search
The search algorithm with the number of retrieved studies is presented in Fig 1 .A total of 2,142 records were retrieved, and after subtracting 554 repeat entries, 1,588 records were considered for analysis.From these 1,502 records were not considered relevant for the review and therefore excluded.Eighty-six full text articles were assessed and of these 27 were excluded.Four additional studies were retrieved from reference lists and from consultation with experts in the field giving a total of 63 studies considered relevant for this review .

Research Areas
For the sake of clarity, data for ITP and STP are presented separately, each in a subsection.
3.6.2'Sensitizing' Test Paradigms.Demographics are presented in Table 4.The total number of subjects in the STP-studies was 1,048, with a median (IQR) number in each study of 14.5 (11.3 to 23.8) subjects.The second largest (n = 158) [60] and the third largest (n = 151) [61]   Objectives related to the specific perspectives of the review.† ratio of placebo-treated vs. naloxone-treated was 0.5.
# study design is for remifentanil-placebo infusions.
3.9.2Mechanical Stimuli.'Inhibitory' Test Paradigms.One study used pin-prick stimulations by nylon filaments [19] with bending forces from 0.08 mN to 2,492 mN [19] for I study includes patients with bulimia nervosa (n = 10) and anorexia nervosa (n = 10; data not reported here).
J study includes patients with major depression (n = 20; data not reported here).
K study includes placebo-controlled treatment arm with ketamine (0.4 mg/kg; data not reported here).
L the total number of subjects were 20 (4 were excluded).
M the total number of subjects were 12 (2 were excluded).
N study includes treatment arms with paracetamol (1g i.v.) and paracetamol/naloxone (8 mg i.v.; data not reported here).SD standard deviation.For explanation of abbreviations, please, refer to legend Table 3.
3.10.2Mechanical Stimuli.'Inhibitory' Test Paradigms.Two studies [4,9] used the modified, ischemic submaximal effort tourniquet test, with assessment of hand grip strength, as conditioning stimulation (Table 5).One of these studies in addition used exercise (6.3 mile [10 km] run) at 85% of maximal aerobic capacity as a physiological conditioning stressor [9].Another study employed 20 min leg and arm conditioning exercises on ergometers [13].

Primary Objective and Outcome
Objective and outcome are related to the perspectives of this review and does not necessarily imply that these also are the main objectives and outcomes of the reviewed studies.
Repetitive Transcranial Magnetic Stimulation Models.The primary objective of the dorsolateral prefrontal cortex (DLPFC) targeted rTMS-studies was to examine the effect of a MOR-antagonist on stimulation-evoked analgesia [23][24][25] (Table 1).The test stimuli were applied homotopically, i.e., at the right side when rTMS was targeted at the left hemisphere and vice versa.The outcomes were changes in pain perception, pain threshold and pain tolerance (Table 5).
3.11.2'Sensitizing' Test Paradigms.Secondary Hyperalgesia Models.The objectives were to examine if administration of a MOR-antagonist was associated with an increase in secondary hyperalgesia areas induced by a thermal injury [26,27,31], thermal suprathreshold stimulation [31] or noxious electrical high-intensity stimulation [28][29][30] (Table 2).Three studies examined either the effect of naloxone on ketamine-induced secondary hyperalgesia area [26] or opioid-induced hyperalgesia following remifentanil [28,30].In both of these studies only data pertaining naloxone vs. placebo administration were considered relevant for this review.Another study examined re-instatement of secondary hyperalgesia area 72 hours after the burn-injury [31].Outcomes were changes in secondary hyperalgesia areas, i.e. allodynia-or hyperalgesia-areas, evaluated by tactile or pin-prick stimuli, respectively (Table 6).Summation Models.The objectives of the 2 summation studies were to examine the effect of a MOR-antagonist in regard to spatially directed expectation of pain [32], or temporal summation of phasic heat-and cold-stimuli [33] (Table 2).In the former, largest study (n = 173) of the review [32], SC injections of capsaicin were administered, unilaterally in the hand or in the foot, and expectations of analgesia were induced by application of a placebo cream, told to contain a 'powerful local anesthetic substance'.This paradigm induced a placebo response in the treated body part, but not in untreated body parts.The objective of the study was to examine reversal effects of a MOR-antagonist on the placebo-response.In the latter study [33], repeated heat and cold stimuli, induced central sensitization, and the objective was to study the differential effect of naloxone on the first (single thermal stimuli [A-delta]) and second pain (repetitive thermal stimuli [C-fiber]) in the temporal summation process.The outcome parameters were in both studies pain ratings (Table 6).
Nociceptive Reflex Models.The common objective of these 3 studies [59][60][61] was to examine the effect of MOR-antagonists on the thresholds of nociceptive flexion reflexes (Table 2).In addition, the first study [59] investigated non-nociceptive spinal reflexes, while the 2 other studies [60,61] investigated pain thresholds and pain ratings.The outcome parameters were nociceptive sural nerve stimulation induced changes in the biceps femoris muscle EMG-RIII component (Table 6).
Miscellaneous.The objective in the 2 studies was to examine the effects of MOR-antagonist on opioid induced antihyperalgesia following a burn-injury [62,63].In both studies the MORantagonist was administered prior to opioid.The outcomes were pain ratings [62,63] and pain thresholds [63].
3.12 Secondary Objectives.Secondary objectives related to the administration of MOR-antagonists are presented in Tables 1 and 2.
In the spatial summation induced conditioning study [17] naloxone reversed the spatial summation induced activation of the endogenous pain inhibitory system.
In 2 [14,16] of the 4 DNIC-studies [4,14-16], a naloxone-dependent complete reversal of the DNIC-induced increase in nociceptive flexion reflex [14] and an increased cardiovascular reactivity to tonic noxious cold stimulus [16] were demonstrated.However, in the latter study [16] no effect of naloxone on the heat pain perception (DNIC-efficiency [86]), compared to placebo, was observed.In the two remaining studies the findings were ambiguous [4,15].In one of these studies [4] a significant naloxone-dependent effects were not demonstrated, though a likely reversal effect of naloxone on the conditioning-induced increase in heat pain thresholds was seen.In the other study [15] a trend (P = 0.07) towards a naloxone-dependent blocking effect of DNIC was observed and the authors attributed this to a type II error.The sample size in this cross-over study was 20 subjects (subgroup of extensive metabolizers of sparteine) indicating that the naloxone-dependent effect in this study, probably was rather weak.
In the heterotopic noxious conditioning stimulation study [21], naloxone generally did not affect pain ratings to phasic heat test stimuli during the cold-water immersion test (CWIT), however increased pain ratings to the tonic CWIT-conditioning stimulus were observed, compared with placebo treatment.The study also demonstrated an impaired correlation between the CWIT-pain ratings and the measure of endogenous analgesia (assessed by the phasic heat stimuli) in the naloxone-sessions compared to the placebo-sessions.
In the single CPM-study [22] (Table 5), it was demonstrated that naltrexone abolished a CPM-induced decrease in heat pain ratings, but only in subjects with low ratings on the pain catastrophizing scale (PCS).
The 3 studies [11,18,19] employing repetitive noxious stimulation, utilized very different methodological designs and reached different conclusions.One study observed that local administration of naloxone was associated with augmented sensitivity during repeated CWIT [18], while the other studies [11,19] were unable to demonstrate any naloxone-dependent effects on electrical pain thresholds [11] or on the magnitude of habituation in a complex model of repeated heat stimuli [19].
In 3 out of 5 studies [3,5,6,13,20] utilizing a non-noxious frequency modulated peripheral condition stimulation model, no effect of naloxone on the nociceptive component of the blink reflex [5] or dental electrical pain thresholds [3,6] was seen.In one study a paradoxically prolonged increase in the electrical dental pain threshold was observed after naloxone administration [13], indicating a hypoalgesic effect.A study using high-frequency transcutaneous electrical nerve stimulation induced thermal hypoalgesia was not affected by placebo or lowdose naloxone (0.04 mg/kg), but was blocked with high-dose naloxone (0.28 mg/kg) [20].
Summation Models.In the large placebo study [32] naloxone completely abolished the placebo response indicating that endogenous opioids, spatially modulate specific placebo responses.In the temporal summation study [33] using repeated phasic heat and cold stimuli, no effect of naloxone, compared to placebo, on thermal wind-up, "first" pain or "second" pain, was observed (Table 6).
Summarizing the results from the 25 studies, the direction of MOR-antagonist dependent effect on pain ratings, threshold assessments and SSEP appear quite ambiguous and inconsistent [42].Any evidence for stimulation modality specific changes in response to MOR-antagonists is lacking.However, the results on heat stimuli from the 2 neuroimaging studies seem consistent and promising [54,57].
Nociceptive Reflex Models.In the first study [59] naloxone facilitated, i.e. increased the amplitude, of the monosynaptic spinal reflex, but did not affect the tactile polysynaptic reflex (RII).In 2 [59,60] of the 3 threshold studies [59][60][61] no effect of naloxone on the threshold of nociceptive flexion reflex (Table 6) could be demonstrated.In the third study [61] the results were somewhat at odds with previous findings, indicating lowered nociceptive flexion reflex activity (defined as the rectified biceps femoris EMG measured at the 90-150 ms post-stimulation interval) after administration of naltrexone: a hypoalgesic effect corroborated by the naltrexone-associated findings of significantly decreased pain ratings at electrical pain and tolerance thresholds.However, both in the second [60] and third study [61] administration of naltrexone during noxious sural nerve-stimulation was associated with increased pain in female subjects, while in male subjects naloxone administration was associated with an increase in electrical pain thresholds.
Miscellaneous Models.In the 2 burn-injury studies [62,63] a naloxone-dependent reversal of opioid-induced anti-hyperalgesia was demonstrated primarily by an increase in heat sensitivity (Table 6).
3.14 Adverse Effects, Withdrawals and Outliers 3.14.1 'Inhibitory' Test Paradigms.Seven of the 24 studies described either drug-related adverse effects [1,22], withdrawals not related to administration of MOR-antagonist [13,21,24,25] or the occurrence of outliers [8].In one study [22] adverse effects, termed "mild side effects", like mental dulling, confusion, sedation and poor balance were reported.Unfortunately this study only described mean values of the adverse effects based on the group, while the absolute number of subjects experiencing the adverse effects was not given.In the study with the highest naloxone dosis, i.e. 6,000 microg/kg, the subjects were unable to tell the correct order (with a likelihood higher than chance) of active drug vs. placebo and no adverse effects were reported [16].
3.14.2'Sensitizing' Test Paradigms.Thirteen of the 38 studies described either occurrence of drug-related adverse effects [26,27,38,39,42,51,62], withdrawals not related to administration of MOR-antagonist [30,36,38,53,57,58] or the occurrence of outliers [55].Six subjects were reported to experience adverse effects: 1 subject had psychotropic effects due to ketamine [26], 1 subject developed a second degree burn injury [26] and 4 subjects experienced sensation of warmth, palpitations, drowsiness, nausea and vomiting [27,62]: in 3 of the subjects very likely related to administration of MOR-antagonist.In one study [39] the responses in the side effect questionnaires showed that tiredness, lightheadedness, nausea, abdominal "grumbling", and mood changes were reported slightly more often after naloxone than after placebo.In another study [51] a "drowsiness" scale demonstrated higher values for naloxone than for placebo.

Potential Clinical Implications
4.2.1 'Inhibitory' Test Paradigm.The descending conditioned pain modulation system (DNIC or CPM; cf.4.4.2) is considered an important factor regulating pain sensitivity in humans [87][88][89] and it has been suggested that pathological changes in the CPM-system are important for the development of chronic pain in chronic tension headache, fibromyalgia and persistent postoperative pain [33,88,[90][91][92][93][94].The CPM system is in part modulated by exogenous opioids: a sub-therapeutic dose of morphine may uncouple the conditioning system, deregulating the balance between pain sensitization and inhibition [14,95].Impairment of the descending inhibitory systems, e.g., the EOS and CPM, may contribute to the trajectory from acute to chronic pain.Research in blockade of the opioid system may improve our understanding of the underlying pathophysiological mechanisms and may lead to a reformulation of strategies for the prevention and management of chronic pain.
4.2.2 'Sensitizing' Test Paradigm.Obviously, a number of scientific issues are common for the complementary test paradigms, ITP and STP, however, injury or disease related nociceptive input to the CNS may trigger a sustained excitability and increased synaptic efficiency in central neurons [93], i.e., central sensitization (CS), a stimulus-response enhancing mode which may contribute to the development and maintenance of a chronic pain state [93,96].Animal data suggest that CS outlasts overt signs of hyperalgesia, in a silent form termed 'latent sensitization' (LS).The LS far outlasts the conventional duration of the injury assessed by behavioral measures, but can be unmasked by administration of a centrally acting MOR-antagonist leading to "rekindling" or reinstatement of hyperalgesia [92,93,97].Thus, post-injury pain remission is maintained in part by the EOS that masks the pro-nociceptive components of LS. 'Latent sensitization' could prime central nociceptive circuitry such that, when inhibitory systems fail, as upon exposure to excessive stress, a pain episode ensues [93,97].The latent predisposition to relapse, may explain the episodic nature and vulnerability to stressors that accompany chronic pain states in humans [93,97].

Dose-issue
First, while the orally administered naltrexone dose was rather uniform 50 mg [22,40,50,53,56,[60][61][62], the parenterally administered naloxone doses, ranged from 0.21 to 6,000 microg/kg [16,29]: a 29,000 fold difference in doses across studies (Tables 3 and 4)!This difference may obviously bias the study results, particularly considering the biphasic response pattern induced by MOR-antagonists (cf.2.3).The dose-response issue, was examined by 8 studies [20,27,29,37,39,43,46,47], but only 2 studies recognized a dose-response pattern [20,29].The ITP-study [20] comparing naloxone 0.04 mg/kg and 0.28 mg/kg, demonstrated a significant effect of the highest dose on blockade of the analgesic effect of high-frequency transcutaneous electrical nerve stimulation.However, there are some methodological objections to this study.The authors observed highly significant sequence effects: subjects randomized to receive placebo at their first session (naloxone administered at the second and third sessions) demonstrated a higher difference in pain scores compared to the naloxone sessions, than when placebo was administered at the third session (naloxone administered at the at the first and second sessions).This effect most likely was attributed to habituation generated by the conditioning electrical stimuli (cf.4.4.1).Since only data from first sessions with placebo administration could be used, the authors in order to compensate for the unintended reduction in statistical power, in a posthoc manner included 3 more subjects, most likely violating the study protocol.The STP-study [29] used successively, increasing naloxone doses of 0.21, 2.1 and 21.0 microg/kg, administered as target-controlled infusions.Significant dose-dependent increases in secondary hyperalgesia areas were demonstrated (in 2 out of 3 test sessions), and this study is the only, in a statistically correct way, to confirm a dose-response relationship for naloxone in the present review.
Third, the weighted mean dose of parenterally administered naloxone, was 158 microg/kg (11.0 mg for 70 kg BW).Human data, data based on a PET-study, demonstrated complete inhibition of [ 11 C]-carfentanil binding to opioid-receptors following 100 microg/kg naloxone [98].The authors are not aware of additional binding-studies and though the effective naloxone blocking dose may be lower, it is interesting that only in 19 out of 53 studies in the present review, with parenterally administered naloxone, a dose of 6 mg (86 microg/kg per 70 kg BW) was used.Considering the short half-life of naloxone it seems important to maintain steady effect compartment concentrations by target-controlled infusions.
Transcutaneous electrical stimulation is a ubiquitous method used in pain research, due to its ease of use, flexibility, and in particular, due to its versatility, regarding stimulation rates, intensity adjustments and generation of sequence-patterns.The method elicits pain and hyperalgesia by direct axonal stimulation, bypassing the sensory nerve endings [100].Electrical stimuli are thus probably more suitable for examination of central pain components, than 'physiological' thermal and mechanical stimuli, reflecting both peripheral and central components of the pain response [101].High-intensity, noxious electrical stimulation (for stimulation characteristics cf.Table 5) is associated with activation of the EOS and an apparent stimulation-dependent decrease in pain ratings [28,29,102].The degree of habituation, assessed as decrements in pain ratings, following 45-180 min continuous noxious electrical stimulation, is between 20-60%, using constant stimulation intensity [28,29,100,102,103].Using a test paradigm, adjusting the current strength to a constant level of pain perception, the increase in stimulation intensity during 45 min of continuous stimulation, is 260% [103]!Thus, in the SIA-studies the decrements in pain ratings [7,10,12], increases in threshold of the nociceptive flexion reflex (RIII) [1,2,10,12], thresholds of the monosynaptic spinal reflex [1,12], electrical pain thresholds [8], and, decreases in SSEP [7], observed in controls, could in part be explained by habituation [7] (Table 5).
Prolonged and intense, noxious electrical conditioning stimulation activates a naloxonesensitive (indicating recruitment of the EOS) and a naloxone-insensitive inhibitory system [29].Interestingly, these findings are not modality specific, but also apply to noxious contact heat [104,105], laser stimuli [106] and capsaicin application [36].
In heat-models it has been shown, that habituation involves the descending antinociceptive system and in addition comprises a naloxone-insensitive component [19].The fMRI-based studies demonstrated that part of the antinociceptive system, the rostral and subgenual anterior cingulate cortex (rACC/sgACC)) and the periaqueductal grey area, were involved in habituation, indicating a contribution of central pathways to the phenomenon of habituation [104,106].
Experimental factors that potentially may influence habituation and its central correlates are the stimulation modality, stimulation rate, the timeline of the stimulation, i.e. short-term or long-term habituation, and, the use of phasic or tonic stimulation patterns [19,103,104] [88].In the present review the terms were used explicitly as stated in the studies.
The 4 DNIC-studies [4,[14][15][16] presented similar findings in regard to the 'classical' DNICparadigm: a high intensity noxious stimulus decreased the response or increased the threshold to a heterotopically applied pain-stimulus [87].However, naloxone-induced reversal of the DNIC-effect was only unambiguously demonstrated in one of the studies [14], an effect that would have been anticipated since previous studies have indicated opioid-sensitive components of DNIC [14,87,107].Among the 3 DNIC-studies [4,15,16] with ambiguous findings, two of the studies were low-powered [4,16] and the authors of one of the studies stated that the negative findings could likely be attributed to a lack of statistical power (n = 6) [16].Interestingly, one of the studies [15], seemingly adequately powered, indicated a weak naloxone-dependent effect (cf. 3.13.1).
4.4.3Non-noxious Frequency Modulated Peripheral Condition Stimulation.The 5 studies on non-noxious frequency modulated peripheral conditioning did not show any consistent sign of involvement of the EOS [3,5,6,13,20].
4.4.4Repetitive Transcranial Magnetic Stimulation.The methodological qualities of these studies were among the highest in the present review.An advantage was that the studies used sham-controlled and placebo-controlled procedures.Two of the studies [24,25] observed significant attenuating effects of naloxone on DLPFC-targeted rTMS, while the third study [23] only observed this for M1-targeted rTMS.The studies utilized thermal test paradigms, although they differed in regard to pretreatment with capsaicin that was used in two of the studies [24,25], i.e., application of capsaicin may lead to more intense pain perception [108].The studies also differed in regard to left [24,25] and right [23] hemisphere targeted stimulation.
4.5 'Sensitizing' Test Paradigms 4.5.1 Secondary Hyperalgesia Models.One study [29] demonstrated a dose-dependent naloxone response, with increasing magnitudes of secondary hyperalgesia areas induced by noxious, high intensity, intradermal electrical stimulation.Interestingly, in a preceding study by the same research group [28], using a nearly identical set-up, naloxone-infusion 10 microg/ kg, was associated with a highly significant increase in secondary hyperalgesia area (140%) compared to pre-infusion values (P < 0.01).However, since the baseline values successively increased during the 30 min infusion period, the increase in secondary hyperalgesia areas compared to controls, did not reach significance (P = 0.16; deviations in baseline assessments in noxious electrical stimulation, cf.4.4.1).The study abstract correctly indicated that naloxone resulted in increased pain ratings ('antianalgesia'; P < 0.001), but, erroneously indicated that naloxone resulted in 'mechanical hyperalgesia (P < 0.01)' [28].The third secondary hyperalgesia study using high intensity intradermal electrical stimulation, on remifentanil-induced opioid hyperalgesia (OIH) [30] was not able to demonstrate any effect of naloxone on secondary hyperalgesia.Though the study generally had a double-blind, controlled and randomized design, the naloxone part of the investigation did not include a placebo arm per se, but naloxone was administered blinded to the remifentanil and the placebo groups.
Consequently, in 5 of 6 secondary hyperalgesia studies obvious signs of EOS involvement could not be demonstrated, however, naloxone-dependent hyperalgesic responses during high intensity noxious electricity stimulation cannot be excluded.Recent data, examining latent sensitization [109] using a burn injury model [31] indicated that administration of a high dose of naloxone 2 mg/kg, 1 week after the injury, in 4 out of 12 subjects seem to be associated with late re-instatement of secondary hyperalgesia areas (submitted: Pereira MP et al. 'Endogenous opioid-masked latent pain sensitization: studies from mouse to human.').

Study Bias
4.6.1 General Issues.The bibliographic age and the exploratory nature of the studies should be taken into consideration when discussing study bias, i.e., the risk that the true intervention effect will be overestimated or underestimated (http://handbook.cochrane.org/[accessed 07.24.2014]).The present review using the simple Oxford quality scoring system from 1996 [68] demonstrated a high likelihood of bias, due to inaccurate reporting of randomization and blinding procedures.The Oxford quality scoring system was chosen in respect of the bibliographic age of the studies since it presents a more lenient evaluation paradigm than more sophisticated methods like the Cochrane Collaboration's tool for assessing risk of bias.Approximately 25% and 50% of the studies were published before 1984 and before 2000, respectively, while the early report on the Consolidated Standards of Reporting Trials (CONSORT) was published in 1996 [110], with revised versions 2001 [111] and 2010 [112].Compliance to these standards has been a requirement for randomized controlled trials (RCTs) in a number of clinical journals for more than 17 years [110].The CONSORT statement thus contains guidelines applicable to clinical RCTs, but is it relevant for experimental RCTs?
The editorial accompanying the first CONSORT publication stated: "It seems reasonable to hope that, in addition to improved reporting, the wide adoption of this new publication standard will improve the conduct of future research by increasing awareness of the requirements for a good trial.Such success might lead to similar initiatives for other types of research" [110].Requirements for clinical RCTs as outlined in CONSORT and SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) [113], could also be considered essential for experimental research in order to heighten validity, reliability and reproducibility of data, facilitating accurate reporting, evaluation and interpretation of study-data.Guidelines for reporting animal research data, ARRIVE (Animals in Research: Reporting In Vivo Experiments) [114], based on the CONSORT criteria, have been published and even extrapolated for use in a review including human experimental research [115].
4.6.2Statistical Issues.The heterogeneity of the statistical methods used was considerable.Various ANOVA-methods, some rather advanced [24,42], were used in 41/63 studies, while a priori and post-hoc sample size estimations were only employed in 4/63 and 3/63 studies, respectively.Furthermore, in 49/63 studies, statistical methods aimed at reducing the risk of type I error (α [false positive]) associated with multiple comparisons, were not applied.It can readily be calculated that the likelihood by chance of achieving one or more false positive results, presenting as significant values (P < 0.05), during 5 and 10 uncorrected pair-wise comparisons, which was a normal procedure in the studies, are 23% and 40%, respectively, giving a high likelihood for falsely rejecting the null hypothesis [116,117].A simple measure to attenuate the risk is to decrease the significance level to 1% which would give corresponding likelihoods of less than 4.9% and 9.6%, respectively.Conversely, the risk of committing a type II error (β [false negative]), i.e., indicating the power of the study, was generally not reported in the studies, although the limited sample size was discussed in 10/63 studies.
However, it could be argued that all the reviewed studies are experimental, and as such exploratory and hypothesis-generating in nature.In studies of healthy individuals, inferences from qualitative aspects, within-subject variances and fixed-effect models are often more important than inferences from quantitative aspects, between-subject variances and randomeffect models, the latter being preferred in clinical research examining groups of patients [118].The number of subjects needed is obviously much smaller for experimental research compared to clinical studies, mainly due to a much larger inherent biological variance in patients compared to healthy subjects.However, the exploratory nature of an experimental study is sometimes used as a an excuse for not adhering strictly to common statistical requirements [117].For all research, decisions on the null hypothesis, primary and secondary outcomes, and, estimations of outcome variability (from pilot-data if not available from literature), minimal relevant differences, sample size and effect size calculations should be stated even in studies of exploratory nature.Otherwise, there is an obvious risk of wasting valuable research time and efforts, leading to ethical, economical or scientific dilemmas, which might impede future research [116,117,119].
4.6.3Methodological Issues.Lack of standardization across the studies and the stimulation methods are evident (Tables 5 and 6).Guidelines for sensory testing procedures have been presented [120][121][122][123], but standardized protocols, like the German Research Network on Neuropathic Pain (DFNS) [122][123][124] were not used in any of the studies.Aspects of data reproducibility and validity were only discussed in few studies [29,31,125].

Conclusion
'The consistent failure to find an effect of naloxone on experimental pain in humans suggests that endorphin release did not occur during these procedures' [37].This systematic review on endogenous opioid antagonism in physiological experimental pain models concludes that naloxone appears to have a demonstrable and relatively reliable effect in stress-induced analgesia (in all 7 studies) and repetitive transcranial magnetic stimulation (in all 3 studies).In all other pain models, both naloxone and naltrexone demonstrate a variable and unreliable effect.

Fig 2 .
Fig 2. Schematic illustration of the 'inhibitory' test paradigms (ITP, upper panel) and the 'sensitizing' test paradigms (STP, lower panel).The ITPstudies employed an inhibitory conditioning stimulus with evaluation of the associated change in the applied test-stimulus (4test-stimulus).The objective of the ITP-studies was to examine the effect of mu-opioid-receptor (MOR) antagonist on the magnitude of the 4test-stimulus, indicating an activation of the endogenous opioid system (EOS) responsible for the conditioning response leading to antinociception/hypoalgesia (the central rectangle [Opioid-dependent mechanism?]indicates a hypothetical augmentation of the conditioning response by the EOS).The STP-studies (lower panel) employed a pain stimulus leading to quantifiable 'sensitizing' CNS-responses, e.g., changes in behavioral measures (hyperalgesia, pain ratings, thresholds, tolerance), nociceptive , enhancing the nociceptive responses.The objective of the STP-studies was to examine the effect of MOR-antagonist on the magnitude of elicited responses, indirectly either supporting or contradicting an effect mediated by the EOS (the central rectangle [Opioid-dependent mechanism?]indicates a hypothetical attenuation of the response by the EOS).FM Peripheral Conditioning = non-noxious Frequency Modulated Peripheral Conditioning; rTMS = repetitive Transcranial Magnetic Stimulation.doi:10.1371/journal.pone.0125887.g0023.3 Study Design 3.3.1 'Inhibitory' Test Paradigms.Study designs are presented in Table

3 2 [ 53 ] 2 [ 2 [stimuli 2 [ 58 ] 3 [ 62 ] 2 [ 63 ]
Effect of an 8 hr Nx-infusion on pain induced by the tourniquet test Effect of 8 hr Nx-infusion on cortisol, Posner J 1985 DB, R, CB, PC, 6-SX Effect of Nx on pain induced during the tourniquet test NRR Mechanical: [49] Schobel HP 1998 DB, R, PC, 2-WX Effect of Nx on pain ratings to pinching stimuli Effects of Nx on hemodynamic and sympathetic responses to pain 2 [50] Cook DB 2000 DB, R, CB, PC, 3-WX Effect of NTx on pain induced by dynamic hand grip fatiguing exercise Effect of NTx on sympathetic nerve activity during Effect of Nx on heat and cold pain thresholds, and vibratory thresholds NRR Al'Absi M 2004 DB, CB, PC, 2-WX Effect of NTx on pain induced by heat and CPTT NRR 54] Borras MC 2004 DB, R, PC, 2-WX Effect of Nx on pain and CNS-responses (fMRI) to suprathreshold heat stimuli 55] Kern D 2008 DB, R, PC, 2x2-WX, Effect of Nx on paradoxical pain induced by the "thermal grill" Effect of Nx on thermal thresholds 4 [56] Kotlyar M 2008 DB, R, PC, 2-WX Effect of NTx on pain induced by CPTT Effect of NTx on sympathetic responses induced by CPTT 3 [57] Schoell ED 2010 DB, CB, PC, 2-WX Effect of Nx on pain ratings and CNS-responses (BOLD) to suprathreshold heat Pickering G 2013 DB, R, PC, 4-WX Effect of Nx on pain induced by repeated heat stimuli Effect of Nx on SSEP induced by France CR 2005 DB, R, PC, 2-WX Effect of NTx on pain ratings, NFR thresholds and EPT assessments.NRR 61] France CR 2007 DB, R, PC, 2-WC Effect of NTx on pain thresholds, pain tolerance and NFR recordings.Eissenberg T 2000 DB, PC, 4-WX Effect of NTx on reversal of oxycodone induced antihyperalgesia in UV-exposed skin NRR Robertson LJ 2007 DB, R, PC, X Local effect of Nx on opioid induced antihyperalgesia following a burn NRR 2 4

4. 4 . 2
Diffuse Noxious Inhibitory Control and Conditioned Pain Modulation.DNIC and CPM are synonymous terms, and it has recently been recommended that DNIC should be reserved for animal and CPM for human research
[9] Janal M 1984 DB, R, PC, CB Effect of Nx on thermal and ischemic responses after exercise NRR [10] Willer JC 1986 DB, R, PC, 4-WX Effect of Nx on stress-induced changes in nociceptive flexion reflex threshold NRR [11] Ernst M 1986 DB, PC, 2-WX Effect of Nx on habituation to repeated noxious

Table 4 .
(Continued) § not interfering with the MOR-antagonist assessments (drugs without administration route stated are i.v.).# 1mg: cold water challenge; 2mg: ischemic pain challenge.A study includes fibromyalgia patients (n = 15, data not reported here).B study includes fibromyalgia patients (n = 10, data not reported here).C study included patients with chronic low back pain (n = 37; data not reported here) and 2 healthy subjects on antidepressant medication.D study includes chronic low back pain patients (n = 45, data not reported here).E study includes treatment arms of combinations of tilidine (100 mg) and naloxone (8-32 mg; data not reported here).F study includes treatment arms with codeine (60 mg p.o.) and codeine/naloxone (2 mg i.v.; data not reported here).G study includes subjects with borderline hypertension (n = 21, data not reported here).H study includes treatment arm with codeine (60 mg p.o.; data not reported here).