High-dose naloxone: Effects by late administration on pain and hyperalgesia following a human heat injury model. A randomized, double-blind, placebo-controlled, crossover trial with an enriched enrollment design

Severe chronic postsurgical pain has a prevalence of 4–10% in the surgical population. The underlying nociceptive mechanisms have not been well characterized. Following the late resolution phase of an inflammatory injury, high-dose μ-opioid-receptor inverse agonists reinstate hypersensitivity to nociceptive stimuli. This unmasking of latent pain sensitization has been a consistent finding in rodents while only observed in a limited number of human volunteers. Latent sensitization could be a potential triggering venue in chronic postsurgical pain. The objective of the present trial was in detail to examine the association between injury-induced secondary hyperalgesia and naloxone-induced unmasking of latent sensitization. Healthy volunteers (n = 80) received a cutaneous heat injury (47°C, 420 s, 12.5 cm2). Baseline secondary hyperalgesia areas were assessed 1 h post-injury. Utilizing an enriched enrollment design, subjects with a magnitude of secondary hyperalgesia areas in the upper quartile (‘high-sensitizers’ [n = 20]) and the lower quartile (‘low-sensitizers’ [n = 20]) were selected for further study. In four consecutive experimental sessions (Sessions 1 to 4), the subjects at two sessions (Sessions 1 and 3) received a cutaneous heat injury followed 168 h later (Sessions 2 and 4) by a three-step target-controlled intravenous infusion of naloxone (3.25 mg/kg), or normal saline. Assessments of secondary hyperalgesia areas were made immediately before and stepwise during the infusions. Simple univariate statistics revealed no significant differences in secondary hyperalgesia areas between naloxone and placebo treatments (P = 0.215), or between ‘high-sensitizers’ and ‘low-sensitizers’ (P = 0.757). In a mixed-effects model, secondary hyperalgesia areas were significantly larger following naloxone as compared to placebo for ‘high-sensitizers’ (P < 0.001), but not ‘low-sensitizers’ (P = 0.651). Although we could not unequivocally demonstrate naloxone-induced reinstatement of heat injury-induced hyperalgesia, further studies in clinical postsurgical pain models are warranted.


Introduction
The endogenous opioid analgesia system can be impaired or altered in chronic pain conditions [1][2][3][4], playing a putative pathophysiological role in the transition from acute to chronic pain [5][6][7]. Naloxone and naltrexone are μ-opioid-receptor (MOR) inverse agonists [6] used in experimental research to determine the activity of the endogenous opioid analgesia system [8,9]. Naloxone produces either hypoalgesic or hyperalgesic responses to nociceptive stimulation, depending on the dose administered [10]. Studies in rodents indicate that endogenous MOR constitutive activity masks a process called latent sensitization [6], defined as an increased responsiveness of nociceptive neurons to afferent input induced by either injury, chronic opioid administration, or physiological stress. Latent sensitization can persist in the absence of behavioral signs of hypersensitivity and outlasts the duration of tissue healing. It can be revealed upon administration of an opioid receptor inverse agonist, leading to reinstatement of hyperalgesia in rodents [5,6,11,12].
In an initial randomized, controlled, crossover trial design, we reported that an intravenous dose of naloxone (21 microg/kg), delivered 72 h after a cutaneous heat injury (CHI), failed to reinstate hyperalgesia in healthy human subjects [13]. In a follow-up trial using a higher dose of naloxone (2 mg/kg), four out of twelve subjects demonstrated reinstatement of hyperalgesia [14]. Reinstatement of latent sensitization, if consistently present in humans may constitute one of the basic trigger mechanisms in development of chronic pain states, e.g. chronic postsurgical pain.
Furthermore, we noticed that the subjects developing reinstatement of hyperalgesia were subjects with larger initial areas of secondary hyperalgesia (SHA), leading us to hypothesize that latent sensitization occurs more often in 'high-sensitizers' as compared to 'low-sensitizers'. If the hypothesis is validated, the enriched design could be used in future trials to determine indicators of vulnerability to chronic postsurgical pain [14].
The objectives of the current trial were, first, to replicate [15] our previous latent sensitization trial applying a larger sample size. Second, using an enriched enrollment design, to examine whether 'high-sensitizers' express larger hyperalgesia areas after a naloxone challenge than 'low-sensitizers'.

Trial management
The trial was approved by the Committee of Health Research Ethics of the Capital Region (H-15018869), the Danish Medicines Agency (2015-005426- 19), and the Data Inspection Authority of the Capital Region (RH-2015-284, I-suite no. 04296). Trial registrations were additionally in EUDRACT (2015-005426-19, registered on January 22, 2016) and ClinicalTrials.gov (NCT02684669, registered on February 10, 2016), with the principal investigator MUW. The trial protocol with detailed methodological information has been published [16] and the original approved protocol and the CONSORT checklist are available as supporting information (S1 Protocol; S1 Checklist).

Participants
Participants were recruited from a registry of previously completed experimental pain studies at the Neuroscience Center, Copenhagen University Hospitals and by an advertisement at the Danish website forsoegsperson.dk (Inclusion and exclusion criteria Table 1). Additionally, the advertisement was posted on facebook.com and hung up on bulletin boards at the University of Copenhagen. Following written and verbal information signed informed consents were obtained from all subjects prior to any assessments. The enrollment process was performed by the corresponding author (ADS). The CONSORT flow diagram shows the included subjects in each session of the trial (Fig 1). The complete date range for participant recruitment and follow-up was February 22, 2016 to October 1, 2016.
Females were not allowed to participate in the study, since it cannot be excluded that exposure to the supra-pharmacological dose (high-dose) of naloxone may cause teratogenic effects: pregnancy tests are not reliable indicators of a gestational age of less than five weeks. The pregnancy prevention measures, e.g. mechanical or hormonal, have an anti-conception success rate below 100%.

Laboratory environment
The experimental procedures took place at the Multidisciplinary Pain Center, Neuroscience Center, Copenhagen University Hospital, and were performed in a quiet, daylighted room (22-25˚C; relative humidity 20%-45%). Participants adopted a relaxed, recumbent position during sensory assessments. Sensory assessments were performed between 8:00 AM and 4:00 PM.

Trial design
A randomized, double-blinded, placebo-controlled, crossover trial with an enriched design. Enrichment. The trial included an enrichment selection session (Session 0) and four experimental sessions (Sessions 1 to 4; Fig 2). The enrichment enrollment selection [17,18] was based on the magnitude of SHA following a CHI, thus separating 'high-sensitizers' (upper quartile [Q4]) and 'low-sensitizers' (lower quartile [Q1]), from 'intermediate-sensitizers' (Q2/ Q3), who were excluded from further trial participation. Subjects in Q4 (n = 20) and Q1 (n = 20) continued to experimental Sessions 1 to 4. However, two subjects from Q4 were excluded before the allocation procedure in Session 1 for reasons unrelated to the study ( Fig  1).
Randomization procedure. After Session 0 a computer-generated random permutation of numbers 1 to 40 was applied to subjects in Q1 (n = 20) and Q4 (n = 20; randomization. com). Each subject from Q1 and Q4 then received a sequential rank order according to the magnitude of the SHA (1 to 20; 61 to 80). These sequential rank numbers were then consecutively paired with the list of random permutation of numbers, rendering randomization of the order of 'high-and low-sensitizers'. The subjects were then invited to participate in the experimental sessions in this randomized order. This randomization procedure was performed by the principal investigator (MUW; not participating in the assessments), and source data were locked away safely. The examiners were blinded to the subject´s 'sensitizer'-affiliation.
Computer-Generated sequence randomization, using blocks of four subjects (randomization.com), was performed by the hospital pharmacy (Skanderborg Pharmacy, Skanderborg, DENMARK), responsible for manufacturing, labeling and packaging of the drugs. Two sets of non-transparent, sealed envelopes, containing information on treatment A total of 170 individuals replied to the advertisement. Ninety-four subjects were assessed for general eligibility on an enrichment selection session (Session 0). Twelve subjects were excluded before the assessments, three tested positive for tetrahydrocannabinol (THC) in their urine, one used prescription drugs and seven had scars or tattoos in the assessment areas. Eighty-three subjects received a cutaneous heat injury (CHI) on Session 0. Subsequently, three subjects were excluded, two subjects did not develop secondary hyperalgesia areas (SHAs), and one subject had a vasovagal syncope following the CHI. Eighty subjects' SHAs were analyzed and ranked according to magnitude, assessed by planimetric measurements. Through an enrichment process, 40 subjects with SHAs in the two middle quartiles (Q2/Q3) were discontinued, while the 20 subjects belonging to the lower quartile (Q1) and the 20 subjects belonging to the upper quartile (Q4) were randomized and continued to the experimental sessions. Two subjects from Q4 were, however, excluded before the allocation procedure in Session 1 for reasons unrelated to the trial, one subject was unavailable, and another was diagnosed with acute lymphocytic leukemia (ALL). All other subjects completed perprotocol the trial sessions, and the final analysis, thus, included 38 subjects.
https://doi.org/10.1371/journal.pone.0242169.g001 allocation order for each participant, were prepared and stored securely while an additional envelope was kept in the examination room, to be opened in case of a medical emergency. Both participants and investigators were blinded to the treatment sequence throughout the trial.

Drug administration
In Sessions 2 and 4, target-controlled infusions (TCI) of naloxone or placebo were administered 168 h after induction of the CHI (Fig 3). The TCI-algorithm, based on previously reported population pharmacokinetic data [21], was calculated by the software NONMEM (7.3 ICON Development Solutions, Manchester, U.K. [property of UCSF, CA]), using computer simulations based on a population-kinetic model with 2,000 simulated administrations distributed on ten subjects. The estimated mean (10% and 90% percentiles) plasma concentrations of naloxone at each of the three TCI-steps are illustrated in Fig 4 [22] with each TCI-step containing a 1 min bolus and a 24 min continuous infusion. A total dose of naloxone 3.25 mg/ kg (4 mg/mL) vis-á-vis normal saline 0.81 mL/kg was administered ( Table 2). During the last 10 min of each 25 min step, plasma concentrations were considered to be stable, and therefore sensory assessments were performed (Fig 4). Monitoring. Vital sign monitoring in TCI-sessions (Sessions 2 and 4) was with continuous three-lead electrocardiogram, heart rate, pulse oximetry, noninvasive arterial blood pressure, and respiratory rate.

Primary outcome
Secondary hyperalgesia areas. Areas of secondary hyperalgesia were assessed by a weighted-pin instrument (128 mN; MRC Systems, Heidelberg, GERMANY) [20] delivering punctate stimuli (stimulus area: 0.049 mm 2 ) in Sessions 0, 1, and 3: before, and 1 and 2 h after the CHI (  Test-algorithm Session 2 and 4 for subjects with superimposed naloxone plasma-concentration curves. Median plasma-concentration (red) with 10% and 90% percentiles ranges (dashed black lines) during a three-step target-controlled infusion (TCI). Naloxone was administered at step 1 (ST1: 15 min to 25 min) with 0.25 mg/kg, step 2 (ST2: 39 min to 49 min) 0.75 mg/kg, and step 3 (ST3: 65 min to 75 min) 2.25 mg/kg. Magenta columns represent ratings with Clinical Opiate Withdrawal Scale (COWS) and blue columns indicate sensory testing (BL and ST1 to ST3: secondary hyperalgesia areas; BL and ST3: pin-prick pain thresholds). BL: baseline assessments [22]. CHI. The participant reported when the punctate sensation changed from an innocuous pinprick to a stinging, smarting, or unpleasant sensation. The corners of the octagon were marked on the skin and transferred to a clear acetate sheet. The SHA was determined by planimetrics using a vector-based drawing program (Canvas 12.0, ACD Systems International, BC, CANADA).

Other outcomes
Mechanical pain thresholds. Pin-prick pain thresholds (PPTs) were assessed in the primary and secondary injury area by punctate stimulators (8,16,32,64,128,256, and 512 mN) using a modified Dixon procedure (Fig 3) [23,24]. Pin-prick pain thresholds were assessed in Sessions 0, 1, and 3 (baseline, 1, and 2 h post-injury) and Sessions 2 and 4 (before the infusion 165 h 0 min and during Step 3 of the TCI 168 h 25 min). The punctate stimulators were applied perpendicularly to the skin, each five times. The lowest nominal value of the punctate stimulator eliciting � 3 stinging, smarting, or unpleasant sensations indicated the PPT. The median value of the four PPTs was used for further analysis. The PPT data was converted to corresponding ordinal values (1 to 8).
Clinical opiate withdrawal scale. Clinical signs of endogenous opiate withdrawal during the naloxone infusion were assessed by the examiner using the Clinical Opiate Withdrawal Scale (COWS) [25] on Session 0, and during the TCI-infusion on Sessions 2 and 4 (Fig 4).
Psychometric evaluations. Since anxiety and depression are well-known psychological parameters intimately associated with pain, participants completed the Hospital Anxiety and Depression Scale (HADS-A/D) [26,27], and the Pain Catastrophizing Scale (PCS) [28,29] at Session 0 before the sensory assessments.
Online reaction time test. Reaction time was tested since it is a measure of vigilance, important during sensory testing. Further, the possible sedating effects of high-dose naloxone was evaluated. Online reaction time was assessed using a computer application showing a redgreen traffic light (https://faculty.washington.edu/chudler/java/redgreen.html; Fig 3) [30]. Participants were asked to press a button immediately when the light switched from red to green. The median of three measurements indicated the reaction time.
Step TCI Time (min) Dose (mg/kg) Step The main analysis used maximal SHA values (SHA MAX ) during the TCI (independent of TCI-step).
The primary outcome SHA was analyzed according to the protocol (S1 Protocol) [16], by simple univariate statistics (Wilcoxon signed-rank test, Mann-Whitney, paired or unpaired t test) and by an advanced mixed-effects model. The mixed-effects model was with a random effect for subject, and fixed effects for the variables 'sensitizers' ('high-sensitizers'/'low-sensitizers'), intervention (naloxone/placebo), TCI-step (step 1, 2, 3), time (Session 2, 4 [added post hoc, see Results-Protocol violations for explanation]), HADS-scores, and PCS-scores for the primary outcome measure SHA. The starting model included all interactions. Main effects and interaction effects were examined. Non-significant factors (P > 0.05), beginning with interactions, were excluded until all included factors attained significance. Carry-over effects were assessed using SHA data from the CHI-sessions (Sessions 1 and 3).
Other outcomes were analyzed by univariate statistics (PPTs, pain during the CHI), ICCs (pain during the CHI), and two-way repeated-measures ANOVA (online reaction time). A Pvalue < 0.01 was considered statistically significant. Data are given as mean (95% CI) or median (95% CI), as appropriate.

Volunteers and trial chronology
A total of 94 subjects were assessed for eligibility (CONSORT flow diagram Fig 1), and 83 subjects were enrolled in Session 0 assessments. Three of these subjects were excluded, and the data from the remaining 80 subjects were analyzed per-protocol [16] (data available as supporting information: S1 Table). Forty subjects, 'high-sensitizers' (Q4; n = 20) and 'low-sensitizers' (Q1; n = 20), continued on to Sessions 1 to 4. Two Q4-subjects were excluded for reasons not related to the trial (Fig 1; specified in S1 Adverse events). Per-protocol data from the remaining 38 subjects were analyzed. Demographic and anthropometric data are presented in Table 3.

Methodology
Changes induced by the cutaneous heat injury. The CHI induced significant SHAs in both Sessions 1 and 3 (Friedman tests: P < 0.001) and all subjects had measurable SHAs in both sessions. Furthermore, the CHI significantly reduced PPTs in both the primary and secondary areas of hyperalgesia in Sessions 1 and 3 (Friedman tests: P < 0.001).
Regarding the estimates of fixed effects, SHA following naloxone compared to placebo was 35.4 cm 2 (P < 0.001; Table 6) for 'high-sensitizers'. However, this difference was not seen in 'low-sensitizers' (P = 0.651). Main effects for TCI-step, HADS-scores, and PCS-scores were not significantly different (  Clinical opiate withdrawal scale. One subject (#AH39) had a COWS rating at step 2 during the naloxone infusion of 5 (1 at baseline), corresponding to mild opiate withdrawal symptoms. Symptoms included yawning, nausea, restlessness, and anxiousness. No other subjects experienced significant opiate withdrawal symptoms during the infusions. Online reaction time test. A two-way repeated-measures ANOVA showed no significant interaction between the effect of intervention and TCI-step on the reaction time (P = 0.615).

Adverse events
For a detailed description of adverse events (AEs), see S1 Adverse events.
Events related to the trial drug. Adverse events of mild intensity including tiredness, nausea, vomiting, dizziness, headache, itching, and restlessness were experienced by 22/38 subjects receiving naloxone compared to 4/38 subjects following placebo (Chi 2 test: P < 0.001). One subject (#AH39) experienced moderate AEs, including anxiety, nausea, and dizziness. One subject (#AX25) experienced anxiety, dizziness, unilateral paresthesia of the arm and leg, and perioral numbness during the naloxone infusion. In both subjects the  infusion was discontinued prematurely, but both agreed to further trial participation and were not excluded from the analysis. Events unrelated to the trial drug. One subject (#92) experienced a vasovagal syncope immediately following the CHI during Session 0 and was excluded from further participation in the trial (Fig 1). No blistering was observed following the CHIs.

Protocol violations
Trial drug package #31 was opened unintentionally, and therefore the package was discarded.
One subject (#43) had a body mass index of 30.7 kg/m 2 exceeding the predetermined upper margin of 30.0 kg/m 2 .
The statistical analysis plan concerning the a priori mixed-effects model with random effects for subject was changed before the un-blinding of data. The trial statistician recommended adding the variable time (Session 2, 4) to the other fixed effects variables (cf. Materials and methods-Statistical analysis: 'sensitizers'; intervention; TCI-step; HADS-scores, and PCS-scores), since time-dependent changes affecting the analyses were likely to occur.

Missing values
One subject (#AH39) did not finish the online reaction time test in the last TCI-step due to naloxone-induced side effects. This subject was therefore excluded from the online reaction time analysis.

Short summary
In the present randomized, controlled, crossover trial, an enriched design was essential in separating sensory phenotypes based on a conditioning heat injury. While the late post-injury naloxone challenge, analyzed by univariate statistics, could not demonstrate reinstatement of latent sensitization, a multivariate mixed-effects model inferred that this could be the case in 'high-sensitizer' subjects.

Cutaneous heat injury model
Previous studies. The CHI model is a reference model in human experimental pain, similarly to the ultraviolet-B (UVB) and capsaicin models [32]. In pharmacodynamical research, the CHI model has demonstrated moderate analgesic efficacy of ketamine [33-36] and opioids [34,37], but only limited efficacy of anti-inflammatory drugs [38][39][40], a paradox since the model is inflammatory in nature. The model, however, provides long-lasting hyperalgesia [41].
Secondary hyperalgesia. Assessments of SHAs are regularly used as a quantitative outcome measure of central sensitization and has been used in physiological studies [42,43], clinical predictive studies [44], and pharmacodynamical studies [19,35,45]. Cutaneous heat injury-induced secondary hyperalgesia is a consistent and reproducible measure with low intra-subject variability [20,46]. Delineation of SHAs is made by punctate (pin-prick) stimulations, conventionally using polyamide monofilaments. However, a comparative trial between the weighted-pin instrument and monofilaments has indicated more reliable measurements, with less variance and larger magnitudes of SHAs with the former method [20]. The development of large SHAs has been revealed as a risk factor for developing chronic pain [47][48][49]. Further, physiological brain similarities exist between subjects with large SHAs following a CHI and chronic pain patients [18]. On the other hand, an association between SHAs and acute postoperative pain is not readily apparent [50][51][52]. Although subjects clearly show phenotypical differences in the development of SHAs, the importance of SHAs could not be definitively established in the present proof of concept trial, in terms of opioid receptormasked latent sensitization.

Naloxone
Comparative dose aspects. The present trial used a naloxone dose of 3.25 mg/kg, which is 600 to 6,000 times higher than the recommended clinical dose used in the treatment of a severe opioid overdose [53]. Animal studies have unmasked latent sensitization with 0.3-10 mg/kg of naloxone [5,14] or 3 mg/kg of naltrexone [11,12]. Uncorroborated data from a positron emission tomography (PET) trial indicated that administration of 0.1 mg/kg of naloxone completely inhibited the binding of [ 11 C]-carfentanil to MOR receptors [54]. Given that the naloxone dose in the present trial by far exceeds 0.1 mg/kg, an unknown 'off-target' effect of the drug is not unlikely.
Adverse events. Previous studies have shown that systemic doses up to 6.0 mg/kg have been tolerated well in healthy participants [55][56][57][58][59][60], and even with higher doses in patients [61][62][63][64][65][66] with none or only mild to moderate AEs. However, AEs have not been systematically examined in any of these trials. In our previous trial, six out of 15 subjects reported mild AEs, including tiredness, epigastric pain, frontal headache, and photophobia, but no serious adverse events were observed [14]. Similar AEs were observed in the present 38 subjects (cf. Results-Adverse events). Based on these findings, no safety issues regarding the current naloxone dosing have been demonstrated.
Pharmacokinetics. We recently evaluated our three-stage stepwise infusion-algorithm in an exploratory pharmacokinetic study in healthy volunteers (n = 8), where we showed that the naloxone plasma concentrations estimated from the population-kinetic modelling were approximately 3-fold higher than the observed [67]. However, the observed peak plasma concentrations of naloxone and naloxone-3-glucuronide in the eight volunteers [67] were more than 30 times higher than seen in previously published studies [21,68,69] and were of the same magnitude (1000 mg/mL) as in the study by Pereira et al. [22].

Data interpretation
For statistical and methodological reasons, we could not unequivocally demonstrate a statistically significant unmasking effect of naloxone on latent sensitization in the CHI model.
First, univariate and multivariate statistical methods yielded different outcomes. While univariate comparisons did not demonstrate a significant difference between infusion treatments or groups, the post hoc mixed-effects model (including the variable time [cf. Materials and methods-Statistical analysis]) yielded a highly significant effect of naloxone on SHAs in the 'high-sensitizers'. No outlier analysis was stipulated in the protocol, but simple visual inspection of the data reveals high data heterogeneity that may have inferred statistical errors (Fig 6).
However, the assigned P-value 0.01 was targeted at mitigating type 1 errors. In the mixedeffect model, a highly significant interaction between 'sensitizers' � time was observed, indicating that not only sensitizer affiliation, but also the time of injection (Session 2 or 4) seemed to affect the SHA outcome. We have no definitive explanation for this possible time effect, but likely it is caused by data outliers.
Second, the current and previous human trials investigating the phenomenon of latent sensitization have been restricted to the use of MOR inverse agonists. By contrast, pre-clinical research suggests that several pain inhibitory G-protein coupled receptors, including μ-δ-and κ-opioid receptors, α 2A -adrenergic receptors, and neuropeptide Y1 and Y2 receptors all contribute to the masking of latent sensitization [5,11,12,[70][71][72][73]. This raises the question as to whether the mechanisms involved in latent sensitization differ between species, or whether blockade of just opioid receptors is not sufficient to reveal latent sensitization in humans. Indeed, receptor interactions include a 100-fold synergy between endogenous μ-opioid and neuropeptide Y1 receptors [74], leading us to speculate that perhaps a balanced mixture of drugs will be required to reveal latent sensitization in humans.
Third, the conditioning tissue injury used in the present trial, putatively leading to latent sensitization, only covered a fraction of the body surface area (< 0.1%), and only involved a relatively weak inflammatory response, that may have been insufficient to trigger the development of latent sensitization. The CHI model has been back translated to the mouse hind paw (52˚C, 40 s). Three weeks later, naloxone dose-dependently reinstated hyperalgesia in these mice. However, the injury was applied to weight-bearing glabrous skin covering a much larger fractional area [75] than in humans. The issue of the limited injury in the present trial could be overcome by studying surgical models with higher severity of tissue injury, e.g., impacted mandibular third molar extraction or groin hernia repair.
Fourth, an issue in previous mouse studies could be a selection bias. The inbred C57BL/6J mouse strain, used in earlier latent sensitization research [11,12], exhibits markedly enhanced pain sensitivity and thus can be considered belonging to a 'high-sensitizer' clone [76][77][78]. However, latent sensitization is readily observed in the outbred Sprague-Dawley rat [12].
Addressing some of the above-mentioned issues may accommodate future trials. Nevertheless, the pathophysiological role of latent sensitization in development of chronic pain is speculative. Albeit, if the unmasking of latent sensitization proves to be an essential element of the postsurgical chronification process, it may improve our understanding of the mechanisms involved significantly. This may lead to targeted presurgical, preventive interventions, but also to novel management strategies in established postsurgical pain states.

Limitations
Data variability. The sample size estimates were based on a previous high-dose naloxone trial involving 12 subjects rendering three subjects in each quartile of 'high-sensitizers' and 'low-sensitizers' (cf. Materials and methods-Sample size calculation) [14]. Data from this low-powered trial obviously lead to fragile estimations of sample size. Other important statistical aspects are discussed above.
Effects during placebo treatment. The assessments during the placebo infusion yielded substantial SHAs corroborating findings from a previous naloxone trial [13]. Two explanations are possible. First, residual sensitizing effects of the preceding CHI 168 h earlier, enhanced by repeated pin-prick stimuli [41], may occur in susceptible subjects. Second, a 'classic' placebo response may potentially alter the magnitude of secondary hyperalgesia [79]: including subjects' knowledge of the expected dose-dependent effect, and an auditory awareness of changes in the pump infusion rate.

Strengths
Trial sample. Only healthy young male subjects were included in the trial, providing a very homogenous sample group. Only two subjects were excluded after Session 0, and missing data were negligible.
Enrichment design. A moderate to substantial agreement between the CHI-sessions was observed regarding the SHAs similar to previous observations [17,46]. However, there was a significant difference in SHAs between Session 0 and Session 1 for the 'low-sensitizers', which may have been caused by the use of different sites for the CHI across the two sessions (thigh vs. medial calf).

Methods
Two proficient investigators (ASD, EKJ) carried out all assessments. To reduce carry-over effects between CHI-sessions, contralateral mirror sites were used, including a 'recovery' period � 41 days between the sessions. The sensory equipment used for the CHIs were regularly calibrated.

Conclusion
The trial examined naloxone-induced reinstatement of latent sensitization in a cutaneous heat injury model in healthy males. The results could not unequivocally establish the phenomenon of latent sensitization in humans. While latent sensitization may occur in humans, it is not as prevalent a finding as in rodents. Further trials, including postsurgical pain models are needed to prove the clinical significance of latent sensitization and its opposing endogenous analgesia.
Supporting information S1 Checklist. CONSORT 2010 checklist. (DOC) S1 Table. Trial data. Table with individual data on demographics, sensitizer allocation, psychometrics (HADS and PCS), secondary hyperalgesia areas in relation to the cutaneous heat injuries and during the target-controlled infusions, VAS pain scores, mechanical pain thresholds, and online reaction times. (XLSX) S1 Protocol. Protocol. (PDF) S1 Adverse events. Adverse events during the trial.