Pharmacological Interventions for Opioid-Induced Hyperalgesia: A Scoping Review of Preclinical Trials

Background: Opioid analgesics are the most effective pharmacological agents for moderate and severe pain. However, opioid use has several limitations such as opioid-induced hyperalgesia (OIH), which refers to the increased pain sensitivity that occurs once analgesia wears off after opioid administration. Several pharmacological interventions have been suggested for OIH, but the current literature does not provide guidelines on which interventions are the most effective and whether they differ depending on the opioid that induces hyperalgesia. This scoping review aimed to identify and describe all the preclinical trials investigating pharmacological interventions for OIH caused by remifentanil, fentanyl, or morphine as the first step towards evaluating whether the most effective OIH interventions are different for different opioids. Methods: Electronic database searches were carried out in Embase, PubMed, and Web of Science. Detailed data extraction was conducted on the eligible trials. Results: 72 trials were eligible for the review. Of these, 27 trials investigated remifentanil, 14 trials investigated fentanyl, and 31 trials investigated morphine. A total of 82 interventions were identified. The most studied interventions were ketamine (eight trials) and gabapentin (four trials). The majority of the interventions were studied in only one trial. The most common mechanism suggested for the interventions was inhibition of N-methyl-D-aspartate (NMDA) receptors. Conclusion: This scoping review identified plenty of preclinical trials investigating pharmacological interventions for OIH. Using the current literature, it is not possible to directly compare the effectiveness of the interventions. Hence, to identify the most effective interventions for each opioid, the interventions must be indirectly compared in a meta-analysis.


Introduction
Opioid analgesics are the most effective pharmacological agents for moderate and severe pain [1]. However, opioid use has several limitations such as the build-up of tolerance and a high risk for the development of addiction. One of the lesser-known limitations is opioid-induced hyperalgesia (OIH), which refers to the increased pain sensitivity that occurs once analgesia wears off after opioid administration [2]. OIH is commonly measured by quantitative sensory testing (QST), where mechanical or thermal stimuli are used to assess the subject's pain threshold [3]. OIH has been mostly studied with surgical patients that have been administered remifentanil [4]. A few trials have observed OIH in chronic pain patients and addiction patients, but it is still unclear whether OIH is clinically significant in non-surgical patient groups [5,6]. To improve pain management with opioids, plenty of research has been carried out to identify interventions for OIH. Most of the interventions are pharmacological, but non-pharmacological interventions have also been investigated such as exercise [7], polyamine-deficient diet [8], and electroacupuncture therapy [9]. Current reviews of the pharmacological OIH interventions are qualitative and lack statistical analysis of effectiveness [10,11]. Additionally, there are two unpublished reviews registered in PROSPERO that aim to rank the effectiveness of pharmacological interventions for remifentanil-induced hyperalgesia (RIH) in surgical patients [12,13]. According to our preliminary searches, not enough randomised controlled trials (RCTs) have been carried out in other patient groups or using other opioids to evaluate the effectiveness of different interventions in these perspectives. Therefore, the current literature has a significant knowledge gap, as it is not known whether OIH in different contexts or caused by different opioids is attenuated most effectively by the same interventions. According to Heinl et al. [14], fentanyl-induced hyperalgesia (FIH) and morphine-induced hyperalgesia (MIH) occur via a distinct mechanism to RIH, which suggests that their most effective interventions may differ. Preliminary searches have shown that several preclinical trials have been conducted to investigate pharmacological interventions for OIH caused by remifentanil, fentanyl, or morphine. Hence, this scoping review aimed to identify all the preclinical trials investigating pharmacological interventions for OIH caused by remifentanil, fentanyl, or morphine as the first step towards evaluating whether the most effective OIH interventions are different for different opioids.

Aims
The aim of this scoping review was to identify and describe all the preclinical trials investigating pharmacological interventions for OIH caused by remifentanil, fentanyl, or morphine. Analysis of these trials will be carried out as the first step towards evaluating whether the most effective OIH interventions are different for different opioids. Hence, this review will qualitatively review the trials to (1) map the existing literature, (2) describe the trial characteristics, (3) identify gaps and limitations in the current research, and (4) make recommendations for future trials.

Protocol
This scoping review was planned according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis Extension for Scoping Reviews (PRISMA-ScR) checklist developed by Tricco et al. [15]. See Appendix A for a summary of the checklist. A protocol to fulfil the requirements was planned as the first step of the review. The protocol included (1) preliminary searches into OIH intervention research to identify a knowledge gap and a suitable research question, (2) defining the eligibility criteria for including and excluding trials, (3) electronic database searches to gather potentially relevant research, (4) selecting eligible articles, (5) data extraction and description of the trials, and (6) summarising the findings and providing recommendations for future research. It should be noted that the protocol had a few minor deviations from the PRISMA-ScR checklist. The checklist includes a step for "review registration", but since scoping reviews cannot be registered in the PROSPERO database, this step was skipped. Similarly, the methods subsection "critical appraisal of individual sources of evidence" and the results subsection "critical appraisal within sources of evidence" were skipped, as the assessment of the methodological quality of the included articles is not typically included in scoping reviews [16].

Eligibility Criteria
For a trial to be eligible in the review, it had to (1) investigate pharmacological intervention(s) for OIH, (2) investigate OIH caused by remifentanil, fentanyl, or morphine, (3) use an in vivo animal model, (4) measure hyperalgesia via QST, (5) be an original full research paper, and (6) be written in English. Pharmacological interventions with all timings, frequencies, dosages and administration methods were included. No time period restrictions were set.

Information Sources
The databases were selected for the search according to Bramer et al.'s [17] analysis of the "optimal database combinations for literature searches in systematic reviews". Hence, to identify potentially relevant trials, electronic database searches were carried out in Embase, PubMed, and Web of Science. Reference lists of the eligible articles were scanned to find trials that may have been missed in the search. All searches were carried out in January 2022.

Search
The search strategy was created according to Leenaars et al.'s [18] "step-by-step guide to systematically identify all relevant animal studies" that combines search items for the disease of interest, interventions, and animal trials. In this review, the disease of interest words included, for example, "opioid-induced hyperalgesia" and "remifentanil-induced hyperalgesia". The intervention search items included generic intervention words such as "drug therapy" and "pharmacological intervention", as well as specific intervention words identified during preliminary searches such as "ketamine". To identify all the available animal trials, the most recently developed animal filters were used [19]. The full search strategies for each database are shown in Appendix B.

Selection of Sources of Evidence
The articles obtained in the final search of each database were exported to EndNote. Duplicate articles were removed and the titles and abstracts were checked for eligibility. Full-text articles were evaluated in the next step if the eligibility was unclear. The process was repeated for the articles identified from the references of eligible articles.

Data Extraction
Data from eligible trials were charted onto a Microsoft Excel sheet. Data items extracted included the following categories: general, opioid, intervention, study design, QST, and animal model. The general information included the first author and year of publication. The type of opioid investigated was used to place the trial in the right category, and the dose, administration method, and regimen type were noted. The opioid regimen was categorised as "acute" if the opioid was given for less than 1 day and "chronic" if the opioid was given for several days. Intervention data items included the pharmacological agent used and its dose, administration method, administration time in comparison to opioid administration, and mechanism to attenuate OIH. In addition, intervention effectiveness was noted for each experimental group. Intervention in an experimental group was categorised as "effective" if statistical difference to the opioid-only group could be shown at any point with any type of QST. For the study design, the type and size of experimental and control groups were extracted, as well as the experimental injury model used and the length of the trial. In the trial lengths, catheterization procedures or baseline days were not taken into account. QST data items were the type of QST used, units of measurement, type of behaviour recorded (e.g., withdrawal or vocalisation), body area used for QST, and timing of measurements compared to opioid administration. Data items extracted about the animal model used included the species, strain, and sex of the animals.
For simplicity and relevance, not all of the study groups discussed in the articles were recorded. Groups exposed to the intervention without an opioid were excluded, since the analgesic property or the lack of it does not determine the intervention's antihyperalgesic property. Similarly, groups that were exposed to the injury model without an opioid were excluded, as the research question specifically focuses on OIH and not other types of hyperalgesia. Groups that only provided additional information on the mechanism of the intervention were not included in the study groups. However, the information was used in data extraction to classify the intervention under the right mechanism. If a trial used several control groups, only the most relevant was included. For example, if both saline only and vehicle only groups were used, the vehicle group was selected. Moreover, whether the opioid only group received saline or other control injections (compared to the opioid and intervention group) was not recorded. Likewise, additional control exposures in other groups were not recorded.

Synthesis of Results
The electronic database search results and the article screening process are summarised as a flow diagram in Figure 1. To be able to present the data extraction results without a supplementary materials section, three tables with only the main characteristics of each type of opioid trials were created (Tables 1-3). These included the first author, year of publication, intervention, study groups, intervention dose(s), administration method and injury model or opioid regimen used. The trial characteristics not included in these tables were described narratively. In addition, all the interventions investigated were listed separately for each opioid and are presented in Table 4. The interventions studied in more than one trial are summarised in Table 5. The mechanisms of the interventions are summarised in tables stating the intervention, its general class, the mechanism to attenuate OIH, and the mechanism group based on the shared pathway for OIH attenuation. Three intervention mechanism tables were created to separately represent the interventions studied for each type of opioid (Tables 6-8).

Selection of Sources of Evidence
Electronic database searches in Embase, PubMed, and Web of Science provided 904 potentially relevant records ( Figure 1). After the duplicates were removed, 472 records were screened for eligibility. Of these, 57 articles were deemed eligible and 14 further articles were identified by scanning the reference lists of eligible articles. A total of 71 articles were included in the review.

Characteristics of Sources of Evidence
The main characteristics of 72 trials investigating pharmacological interventions for OIH are presented in three tables in Section 4.3. Table 1 summarises 27 remifentanil trials, while 14 fentanyl trials are summarised in Table 2 and 31 morphine trials are summarised in Table 3. The trial by Wei and Wei [20] appears in two of the tables because the trial studied interventions for both FIH and MIH. The main characteristics presented are the first author, year of publication, intervention, study groups, intervention dose, administration method and injury model or opioid regimen used. The types of injections used for all of the interventions and opioids are marked by abbreviations: i.v. = intravenous, s.c. = subcutaneous, i.t. = intrathecal, i.p. = intraperitoneal, i.g. = intragastric. In the table for morphine trials, opioid regimen (acute or chronic) is presented instead of the injury models since only one trial used an injury model. On the other hand, in remifentanil and fentanyl trials, no chronic administration was used, so the injury models used are presented. In 23 of the morphine trials, the morphine regimen lasted for several days, and in 8 trials, the regimen consisted of only one morphine injection. In 22 remifentanil trials, a plantar incision injury model was used, while 5 trials did not include any injury model. In fentanyl trials, the plantar incision model was used in some of the experimental groups by Richebé et al. [21] and all experimental groups in the trial by Richebé et al. [22]. Some of the experimental groups in Richebé et al. [21] used a carrageenan injection model, which was also used by Bessière et al. [23], Van Elstraete et al. [24], and Richebé et al. [25]. Le Roy et al. [26] used a non-nociceptive environmental stress (NNES) model, which was also included in some of the experimental groups by Bessière et al. [23]. In morphine trials, Doyle et al. [27] were the only trial with an injury model and they used chronic constriction injury (CCI). Overall, most of the experimental groups marked as "not statistically different to the opioid only group" (#) were the lowest doses of the intervention that was successful with higher doses. Further comments about the interventions are included in Section 4.4. The data extraction items not included in the main characteristic tables are described below.

Remifentanil Trials
In 23 of the remifentanil trials, male Sprague-Dawley rats were used as subjects. In addition, Aguado et al. [28] and Aguado et al. [29] used male Wistar rats, while Qi et al. [30] and Zhou et al. [31] used male Institute of Cancer Research (ICR) mice. The most common experimental group sizes were eight animals (12 trials) and six animals (5 trials). All of the trials used 6-12 animals per study group, apart from Cui et al. [32], who included 20 animals per study group. For QST, most trials used a mechanical and a thermal test. Cui et al. [33], Aguado et al. [28], and Aguado et al. [29] used only a mechanical test. Manual von Frey filaments were used in 13 of the trials, while another 13 trials used an electronic von Frey device. The Liu et al. [34] trial was the only one that used a dynamic plantar analgesiometer. The thermal tests used were a radiant heat paw withdrawal test (13 trials) and a hot plate test (11 trials). In all of the tests, the withdrawal of a hind paw was monitored. In 16 trials, the intervention was administered within a 30-min frame before remifentanil administration, and in 9 trials the intervention was coadministered with remifentanil. Lv et al. [35] administered betulinic acid for 7 days before remifentanil administration, while Li et al. [36] infused lithium chloride (LiCl) or thiadiazolidinone-8 (TDZD-8) for an hour before remifentanil infusion, and Liu et al. [34] administered N-acetyl-cysteine (NAC) 24 h after remifentanil. Most of the trials (16 trials) conducted measurements for 2 days. Cui et al. [33] conducted the shortest trial, which lasted 5 h, while Gu et al. [37] conducted the longest trial, lasting 21 days. The most common remifentanil administration regimens were 0.04 mg/kg subcutaneous infusion for 30 min (9 trials) or 1.0 µg/kg/min intravenous infusion for 60 min (6 trials).

Fentanyl Trials
Male Sprague-Dawley rats were used in all of the fentanyl trials except in the Mert et al. [38] trial that used female Wistar rats. The most common experimental group sizes were 8 animals (five trials) and 10 animals (four trials). All of the trials used 6-12 animals per study group apart from Richebé et al. [25] and Van Elstraete et al. [39], which used 15 and 18 animals per study group respectively. Most of the trials used only one type of QST. However, Richebé et al. [22] used von Frey filaments and the Basile analgesimeter, and Li et al. [40] used von Frey filaments and a radiant heat paw withdrawal test. In total, eight of the trials used the Basile analgesimeter, where the hind paw pain threshold was marked by a vocalisation. A similar technique was used by the Van Elstraete et al. [41] trial using an analgesimeter. Four trials used von Frey filaments and three trials used a radiant heat paw withdrawal test, both of which recorded hind paw withdrawal. In six trials, the intervention was administered within a 30-min frame before fentanyl, and in four trials, the intervention was coadministered with fentanyl. Van Elstraete et al. [24] and Wei and Wei [20]

Morphine Trials
In morphine trials, the animal models used were more heterogeneous than in remifentanil and fentanyl trials. Nine of the trials used male Sprague-Dawley rats. Female Sprague-Dawley rats were included in a few study groups in the trial by Doyle et al. [27]. Other animal models used in the morphine trials were: male Wistar rats, male ICR mice, male Swiss-Webster mice, male Swiss albino mice, male C57BL/6J mice, male CD-1 mice, and male Fischer 344 rats. The most common experimental group sizes were six animals (10 trials) and eight animals (7 trials). The majority of the trials used 6-12 animals per study group. Yet, Dunbar et al. [44] used 16-24 animals per study group and Chen et al. [45], Milne et al. [46], and Doyle et al. [27] had less than 4-5 animals in some study groups. For QST, 15 trials used only one type of test and 14 trials used two different tests. Raghavendra et al. [47], Dogrul et al. [48], and Tumati et al. [49] used three different tests. The most used tests were a radiant heat paw withdrawal test (14 trials), von Frey filaments (hind paw withdrawal, 11 trials), and warm water tail-flick test (8 trials). Sanna et al. [50] and Haleem and Nawas [51] recorded the licking of paws during a hot plate test. In addition, three trials monitored paw withdrawal using the hot plate test. Three further trials used a radiant heat tail-flick test. Dunbar et al. [52] and Corder et al. [53] monitored several types of withdrawal behaviours. Dunbar et al. [52] monitored head shaking, teeth chattering, squeaking, jumping, and urination in response to von Frey filaments while Corder et al. [53] monitored paw flinching, paw guarding, paw attending, and escape jumps on a hot plate. Ferrini et al. [54] recorded the number of vocalisations to subcutaneous morphine injections. In 20 trials the intervention was coadministered with morphine, and in 6 trials the intervention was administered within a 60-min frame before morphine. In 5 trials, the intervention was administered 1-4 days after morphine administration. In the trials, Tumati et al. [55], Tumati et al. [49], and Tumati et al. [56] small interfering RNA for rapidly accelerated fibrosarcoma 1 (Raf-1) or cyclic adenosine-monophosphate (cAMP)-dependent protein kinase A (PKA) was administered for 3 days before morphine. Chen et al. [45] administered ceftriaxone during the 3 days before morphine and during 4 days of morphine administration. Hua et al. [57] administered mesenchymal stem cells (MSC) either 1 day before, 7 days before, or 14 days after morphine administration. Lin et al. [58] administered heat shock protein 70 (HSP70) using an adenovirus 24 h before, 24 h after, and 72 h after the first morphine injection during a twice a day morphine injection regimen that lasted 6 days. The trial lengths of morphine trials were very variable. The most common lengths were 9 days (six trials) and 6 days (five trials). The shortest trial was conducted by Sanna et al. [50] in which they studied the effect of an µ opioid antagonist CTOP within 55 min. The longest trial was conducted by Doyle et al. [27] in which they studied the effect of fingolimod in the presence of CCI for 8 weeks. Acute morphine administration subcutaneously ranged from 0.1 µg/kg to 2.5 mg/kg and intrathecally from 0.05 ng to 10 mg/kg. Two of the acute regimen trials used 1 µg/kg intraperitoneal administration. Chronic morphine regimen ranged from 4 days to 28 days. The most common regimen lengths were 6 days (seven trials) and 7 days (six trials). The most common way to administer morphine was to give two 10 mg/kg subcutaneous injections in a day with 12-h intervals.

Results of Individual Sources of Evidence
Details of the 72 trials reviewed are shown in the tables below. Table 1 is a summary of the remifentanil trials, Table 2 is a summary of the fentanyl trials, and Table 3 is a summary of the morphine trials.

Synthesis of Results
From the 72 trials eligible for the review, 82 different interventions were identified. Remifentanil trials investigated 27 interventions, fentanyl trials investigated 9 interventions, and morphine trials investigated 52 interventions. Table 4 lists all the interventions studied for each type of opioid. If an intervention was different from the control in at least one of the investigated doses it is listed as "effective". The number of trials an intervention was studied in is marked by an X and a number following the intervention. If an intervention was studied for several opioids, it is marked by an asterisk (*). The interventions that were studied most are placed at the top of the lists, while the ineffective interventions are placed at the bottom. Combination interventions were counted as separate interventions from their single interventions and were placed on the list before single interventions. Otherwise, the interventions are presented in arbitrary order.
Ketamine was the most studied intervention in the remifentanil and fentanyl trials. In morphine trials, the most studied interventions were melatonin and Raf-1 selective siRNA. In the fentanyl trials, the only combination intervention studied was ketamine and gabapentin (Van Elstraete et al. [39]) while no combinations were studied in the morphine trials. In the remifentanil trials, three combinations were studied: hydrogen-rich saline and Ro 25-6981 (Zhang et al. [66]), PHA-543613 and PNU-120596 (Zhang et al. [70]), and ketamine and KN93 (Qi et al. [30]). In the fentanyl trials, all interventions demonstrated potential in attenuating OIH. In the remifentanil trials, amitriptyline, minocycline, and maropitant studied by Aguado et al. [29] were the only interventions that were not found effective at all. In the morphine trials, six interventions were found unsuccessful: Rg1 ginsenoside, Rb1 ginsenoside, GAT211, JTE-013, CAY10444, and SEW2871. Table 4. Interventions studied for each type of opioid. Remifentanil trials investigated 27 interventions, fentanyl trials investigated 9 interventions, and morphine trials investigated 52 interventions. An intervention is listed as "ineffective" if none of the tested doses were different from the control. *: the intervention is studied for the other opioids too.

Remifentanil (27)
Fentanyl ( Interventions studied in more than one trial are presented in Table 5. Only 14 of the 82 interventions (17%) were studied more than once. Overall, ketamine was the most-investigated intervention, as it was included in eight trials. The next-most-studied intervention was gabapentin, which was included in four trials. Table 5. Interventions investigated overall in more than one trial. More than one trial was devoted to studying 14 of 82 interventions. Ketamine was the most-studied intervention. The mechanisms of the interventions used in the remifentanil trials are summarised in Table 6, while Tables 7 and 8 summarise the mechanisms of the interventions used in the fentanyl and morphine trials, respectively. These tables display each intervention's general class, the suggested mechanism for OIH attenuation, and the mechanism group based on the shared pathway for OIH attenuation. The interventions with similar mechanisms have been placed close to each other, but otherwise the content is presented in an arbitrary order. Compartments in the table are merged when interventions or mechanisms belong to the same category. Combination interventions are separated into their individual interventions, and their possible mechanisms for additive or synergic effects were not included in the analysis. Table 6 presents the 24 interventions studied in remifentanil trials. The largest mechanism group is N-methyl-D-aspartate receptor (NMDAR) inhibition that includes NMDAR block, NMDAR antagonism, NMDAR expression inhibition, NMDAR phosphorylation inhibition, NMDAR trafficking inhibition as well as calcium/calmodulin-dependent protein kinase II (CaMKII) inhibition and CaMKII phosphorylation inhibition, which influence NMDARs. Six other interventions include NMDAR inhibition in their mechanism group but also influence OIH via other mechanisms. Other most common mechanisms include proinflammatory cytokine reduction (five trials), α-amino-3-hydroxy-5-methyl-4isoxazolepropionic acid receptor (AMPAR) inhibition (three interventions), and inhibition of glial cells (three trials). Table 6. Mechanisms of the interventions in remifentanil trials. General class of 24 interventions, OIH attenuation mechanism, and mechanism group based on the shared pathway for OIH attenuation.

Intervention
General   Table 7). 6 of these share the NMDAR inhibition mechanism group that consists of NMDAR block, NMDAR antagonism and CaMKII inhibition. Other mechanisms for OIH attenuation include cyclooxygenase (COX) inhibition by ketorolac and voltage-gated calcium channel (VGCC) inhibition by gabapentin. The mechanisms of 52 interventions studied in morphine trials are summarised in Table 8. The largest mechanism group is NMDAR inhibition, which consists of seven interventions with NMDAR inhibition mechanism alone and four further interventions with NMDAR inhibition and other potential mechanisms to attenuate OIH. The second-largest mechanism group is sphingosine-1-phosphate receptor (S1PR) inhibition (eight interventions), including antagonists for different S1PR subtypes and S1pr1 silencer. Yet, all of the S1PR inhibition interventions were studied only by Doyle et al. [27]. Other large mechanism groups are opioid receptor inhibition (four interventions), endoplasmic reticulum (ER) stress suppression (four interventions), inhibition of glial cells and proinflammatory cytokine reduction (three interventions), and α2-adrenergic receptor inhibition (three interventions). The general class of olive leaf extract and Withania somnifera root extract is marked as "unclear" since they are a mixture of several compounds. However, for both interventions, potential OIH attenuation mechanisms have been identified. Olive leaf extract is thought to work by blocking calcium channels and reducing proinflammatory cytokine levels. Withania somnifera root extract is suggested to work via γ-aminobutyric acid (GABA) type A and B receptor agonism, NMDAR antagonism, and DOR antagonism. The mechanisms of Re ginsenoside, Rg1 ginsenoside, and Rb1 ginsenoside were marked as "not known" since their potential mechanisms for OIH attenuation are not yet understood. Table 8. Mechanisms of the interventions in morphine trials. General class of 52 interventions, OIH attenuation mechanism, and mechanism group based on the shared pathway for OIH attenuation.

Summary of Evidence
This review identified 72 trials that in total investigated 82 different pharmacological interventions for OIH. There were 27 trials on RIH, 14 trials on FIH, and 31 trials on MIH. In the remifentanil trials, 27 different interventions were investigated while the fentanyl trials investigated 9 interventions and the morphine trials investigated 52 interventions. The majority of the interventions (68 out of 82, 83%) were studied in only one trial. Overall, ketamine and gabapentin were the most-investigated interventions. Only four combination interventions were investigated: ketamine and gabapentin, hydrogen-rich saline and Ro 25-6981, PHA-543613 and PNU-120596, and ketamine and KN93. Furthermore, only 17 trials investigated more than one intervention, and only Doyle et al. [27] and Lin et al. [58] investigated more than three interventions. These 17 trials can provide direct comparisons of interventions, although some of the trials slightly varied their experimental methodology between the interventions they tested. Therefore, to compare the effectiveness of most of the interventions in the literature, a network meta-analysis is required. Only 9 of the 82 interventions were found ineffective in attenuating OIH by all tested doses. The Aguado et al. [29] trial was the only remifentanil trial that studied ineffective interventions. These were amitriptyline, minocycline, and maropitant. In morphine trials, ineffective interventions included Rg1 ginsenoside and Rb1 ginsenoside studied by Li et al. [86]; GAT211 studied by Datta et al. [89]; and JTE-013, CAY10444, and SEW2871 studied by Doyle et al. [27]. In fentanyl trials, no ineffective interventions were reported. Hence, 89% of the tested interventions were reported effective, and this high percentage raises a concern for the presence of positive result publication bias in the literature.
The characteristics of each type of opioid trials reflect the opioid's clinical use and pharmacological properties. Remifentanil is a short-acting opioid with an elimination half-time of 10-20 min and is often used in general anaesthesia [90]. This was mirrored in the preclinical trials, as most of the trials used a plantar incision injury model to mimic surgery, the most common remifentanil administration regimens were 30-min and 60-min infusions, and most of the trials lasted only 2 days. Fentanyl is also mostly used in surgical settings, but may be used to treat chronic pain patients or renal failure patients with severe pain [91]. This is reflected in the preclinical trials as a mixed-use of plantar incision and carrageenan injection injury models as well as a lack of injury model. Although nonnociceptive environmental stress (NNES) is not technically a physiological injury model, it was included in the injury model category of two trials where it was used as stress experienced by patients that may influence the extent to which they develop OIH [26]. Nevertheless, the majority of fentanyl trials (nine trials) did not use any injury model, and hence it is recommended that future fentanyl trials would use a plantar incision injury model to mirror fentanyl's perioperative use. Moreover, in most of the fentanyl trials, fentanyl was administered via four subcutaneous injections within an hour. This is because fentanyl distributes rapidly from plasma to highly vascular tissues such as the heart, lungs, and the brain, and then redistributes rapidly to muscle and fat-whereas several injections force fentanyl to accumulate to its site of action in the brain [92]. Muscle and fat act as storages for fentanyl and slowly release it back to plasma for elimination, which gives fentanyl a long elimination half-time of 3-8 h. This may explain why most of the fentanyl trials had longer trial lengths (5-8 days) than the remifentanil trials. In all of the fentanyl trials, fentanyl was administered within an hour, but since fentanyl may be used for chronic pain and preclinical trials have demonstrated hyperalgesia after chronic fentanyl administration [93], future intervention trials should include chronic fentanyl administration. Morphine is used for moderate to severe acute and chronic pain and is used, for example, in palliative care and for cancer pain and arthritis [94]. This may explain why the majority of the morphine trials investigated OIH after chronic morphine administration, had long trial lengths (most trials ranged between 5-20 days), and lacked injury models.
Overall, the most common intervention mechanism group was found to be NMDAR inhibition. NMDAR inhibition was the mechanism group of 20 interventions and was one of the mechanisms of 10 other interventions. NMDAR inhibition was also the most common mechanism in each type of opioid trial. This suggests that OIH caused by different opioids can be treated with the same interventions. However, the abundance of trials supporting NMDAR inhibition cannot be used to conclude that the mechanism is necessarily the most effective for OIH caused by one or all of the opioids. The effectiveness of interventions and mechanism groups can only be concluded after a meta-analysis. Other large mechanism groups included S1PR inhibition (eight morphine trials), ER stress suppression (six morphine trials), opioid receptor inhibition (five morphine trials), and proinflammatory cytokine reduction (five remifentanil trials, two morphine trials). It should be noted that when opioid receptor inhibition is used for OIH attenuation, the interventions are not meant to alter opioid analgesia. For example, naltrexone can be given in ultra-low doses [76], and methylnaltrexone bromide is peripherally restricted [53], which allows the analgesia to be preserved.
The findings by Heinl et al. [14] show that RIH occurs via long-term potentiation of synaptic strength in the spinal cord dorsal horn C-fibres via activation of spinal MORs and NMDARs. On the other hand, FIH and MIH occur via an enhancement of synaptic transmission at spinal cord dorsal horn C-fibres via activation of spinal MORs and NM-DARs as well as via descending facilitation of C-fibre-evoked field potentials by activation of 5-HT3Rs and extraspinal MORs. These findings suggest that interventions producing NMDAR inhibition are likely to attenuate OIH caused by all opioids, but FIH and MIH may be most effectively targeted with interventions including 5-HT3R inhibition. None of the fentanyl or morphine trials identified investigated combinations of interventions producing NMDAR inhibition and 5-HT3R inhibition. Yet, the Liang et al. [82] morphine trial studied ondansetron, which is a 5-HT3R antagonist, and demonstrated that MIH can be attenuated with a 1 µg intrathecal injection or 2 mg/kg subcutaneous injection of ondansetron. Hence, future fentanyl and morphine trials should investigate whether interventions producing 5-HT3R inhibition may have an additive or synergic effect with interventions producing NMDAR inhibition. Moreover, Roeckel et al. [95] presented an extensive review of the cellular and molecular mechanisms of OIH, which demonstrates that the overall mechanism is a complex process involving multiple pathways. Many of the interventions identified in this scoping review can be connected to the mechanisms they presented such as glial cell activation, production of proinflammatory cytokines, sphingolipid ceramide upregulation, diminished KCC2 action, and neurokinin-1 receptor antagonism. Nevertheless, the mechanism for OIH is still not fully elucidated and the OIH mechanisms of different opioids are not much discussed [11]. The lack of full understanding of OIH mechanisms is also a challenge for OIH intervention research as the interventions are generally planned according to the research on the mechanisms. Furthermore, it has been shown that the choice of experimental animal model affects the extent to which OIH may be observed, which can also have an effect on the conclusion of OIH intervention trials. It has been acknowledged that there are sex-related differences in the experience of pain [96] and this has been also documented in OIH. For example, in Holtman and Wala's [97] experiment, MIH was more pronounced in female Sprague-Dawley rats compared to male rats. Hence, it is important to note that in this scoping review the majority of the trials were conducted with male rats or mice and only two trials included female rats. These trials were the Mert et al. trial [38] that investigated the effect of magnesium on FIH with female Wistar rats, and the Doyle et al. trial [27] that included female Sprague-Dawley rats in their MIH and fingolimod experiment. Therefore, when the efficacy of OIH interventions is evaluated in RCTs, it is recommended that a subgroup analysis is conducted to compare the results in women and men. Moreover, the genetic background of the animal model has also been shown to influence the development of OIH. Liang et al. [98] analysed 15 strains of mice and quantified the reduction in mechanical pain threshold after 4 days of morphine administrations. They observed the largest reduction in MRL/MpJ mice (89%) and the smallest reduction in 129/SvlmJ mice (28.5%), which demonstrates how significantly the choice of animal model may impact a trial. The characteristics of the animal models have also been shown to interact with other trial characteristics. For instance, the experiments by Juni et al. [99] provide evidence that the extent of OIH is influenced by the interaction of the sex of the animal and the opioid dose used. They demonstrated that 1.6 mg/kg daily doses of morphine caused MIH in male and female CD-1 mice, but the male's pain threshold returned to baseline on day 6 while the female's pain threshold was still significantly different from the baseline on day 12. Yet, when male and female CD-1 mice were given 40 mg/kg daily doses of morphine, both groups' thresholds returned to baseline on day 12. After all, since it has been well demonstrated that the choice of animal model significantly impacts the extent to which OIH is observed, it raises a question which of the models is most similar to humans and whether any of them is adequately similar. In addition, it could be argued that the perspective of individualised pharmacology may be important in finding the most effective OIH interventions since the sex and genetic background of humans can influence the extent to which one experiences OIH and responds to OIH interventions.
Although the issue of sample size in preclinical trials has not been recognised to be as important as in clinical trials, the number of animals in each study group should be selected using the best available statistical models [100]. If a sample size is too small, a trial may not be able to represent the phenomena it is investigating in its true effect, while choosing an unnecessarily large sample size is a waste of resources and is ethically less justified. The majority of the trials included in this review used 6-12 animals per each study group. In six trials, more than 12 animals were used. Dunbar et al. [44] used the highest number of animals per study group (up to 24). In the trials by Chen et al. [45], Milne et al. [46], and Doyle et al. [27], less than 4-5 animals were included in some of the study groups. Moreover, the size of the study groups was not mentioned in the trials by Dogrul et al. [48] and Li et al. [86]. In the end, most of the trials did not provide an explanation of how a specific number of animals per study group was chosen, which would be good practice in the future.
The choice of QST could also affect the results of a trial. Kang et al. [42] demonstrated that the extent to which OIH is observed may depend on the QST used. They compared von Frey filaments, a radiant heat paw withdrawal test, and an analgesimeter (paw pressure withdrawal test) in male Sprague-Dawley rats that received a 320 µg subcutaneous fentanyl injection. The rats were monitored for 4 days. Using the radiant heat paw withdrawal test, a slight reduction in the pain threshold was observed through the 4-day period but the results were not statistically significant. In contrast, when the pain threshold was monitored using von Frey filaments, a statistically significant reduction was observed in each day of the measurements. On the other hand, pain threshold monitoring with the analgesimeter exhibited a statistically significant reduction on days 2 and 3, while on the 4th day, the pain threshold returned to baseline. Differing results based on the choice of QST have also been observed in intervention experiments. Dunbar et al. [52] tested the effect of ibuprofen on MIH in male Sprague-Dawley rats. The rats received morphine administrations for 5 days and a 10 µL spinal catheter bolus of 1.36 nM, 13.6 nM, or 136 nM ibuprofen on the 5th day. Their pain thresholds were monitored for an hour on the 5th day by recording paw withdrawal in response to radiant heat as well as monitoring several withdrawal behaviours (head shaking, teeth chattering, spontaneous squeaking, jumping, urination, and squeaking) in response to von Frey filaments. Using the radiant heat test, the results demonstrated that the lowest dose was not effective in attenuating OIH, while the results of the highest dose group were similar to the morphine free group. In contrast, none of the intervention groups were different from the placebo group when withdrawal behaviours in response to von Frey filaments were monitored. Furthermore, Cui et al. [33] demonstrated that the body area used for QST can influence the results of a trial. They conducted a plantar incision surgery in male Sprague-Dawley rats under remifentanil anaesthesia and investigated the effect of lidocaine on RIH. Hyperalgesia was quantified using von Frey filaments in the hind paw with the incision (ipsilateral paw) and in the contralateral hind paw. The results indicated that RIH was significantly more pronounced in all of the study groups in the ipsilateral paw compared to the contralateral paw. Yet, several of the trials included in this review that used an injury model failed to mention whether their QST was conducted on the ipsilateral or contralateral paw, which could affect the interventions' indirect efficacy comparisons. Despite the varying observations from different QSTs and QST methodologies, no reviews were found to evaluate the use of QSTs in preclinical trials.
While this review aimed to analyse trials investigating OIH caused by remifentanil, fentanyl, or morphine, in the screening process it was noted that the search would have identified only two trials conducted with other opioids. These trials were Minville et al. [101] that studied sufentanil-induced hyperalgesia (SIH) and Abreu et al. [102] that studied tramadol-induced hyperalgesia (TIH). Minville et al. [101] investigated the effect of a subcutaneous ketamine injection (1, 10, or 50 mg/kg) on SIH with C57BL/6 male mice that underwent tibial closed fracture surgery that was meant to mimic orthopaedic surgery. The mice were monitored for 7 days and their pain thresholds were measured using von Frey filaments and a hot plate test. The results demonstrated that none of the tested doses of ketamine attenuated SIH when measured using von Frey filaments, but all the tested doses attenuated SIH when measured with a hot plate test. Abreu et al. [102] monitored the effect of 10 mg/kg subcutaneous ketamine injection on TIH in male Wistar rats for 21 days. The rats' pain thresholds were measured using von Frey filaments and a digital Randall-Selitto device. Ketamine was found to attenuate TIH in both tests. In conclusion, these trials support the argument that OIH caused by all types of opioids can be attenuated via NMDAR inhibition.

Limitations
The focus of this review was to identify pharmacological interventions for OIH, but this perspective itself has limitations. Theoretically, it would be more ideal to avoid opioid use in the first place to avoid the development of OIH. In surgical settings, opioid use is common, but according to Lavand'homme [103], opioid-free anaesthesia can be recommended. In addition, novel multi-functional peptides with MOR agonism and neuropeptide FF receptor antagonism have been developed by Drieu La Rochelle et al. [104] and Zhang et al. [105]. These peptides could provide potent analgesia without hyperalgesia. Furthermore, Comelon et al. [106] and Richebé et al. [107] have shown that adjustments to opioid administration such as more gradual withdrawal or target-controlled infusion could be used to prevent OIH. Recommending non-pharmacological interventions such as exer-cise or patient education [11] would also have benefits over pharmacological interventions, as using any medication will always have a risk for side effects. Nevertheless, opioids are still used frequently and much more research is needed in each of the OIH prevention and intervention perspectives before they can be taken into clinical practice. Moreover, a review of the numerous preclinical trials conducted on pharmacological OIH interventions is vital for avoiding repetition of similar work and for planning better preclinical trials and RCTs in the future.
The methodology of this scoping review was planned in line with the best available guidelines, yet a few points for improvement can be identified. After the article selection was completed, a couple of potential search items were identified that were not used in the searches. These included for example "pain sensitisation" and "pretreatment". However, the lack of these search items should not have much effect on missing potential articles, since the references of all eligible trials included were scanned. Since the study selection and data extraction processes were carried out by only one researcher, the possibility for errors could have been reduced by a second reviewer. As discussed, a high likelihood for publication bias in the reviewed literature was identified, and so steps could have been taken to conduct an even wider search. Searches in the grey literature, for example through Google Scholar, could have been carried out. In addition, the analysis could have included relevant abstracts of trials that did not provide a full article and the abstracts of non-English articles.
The results of some of the data extraction groups had to be simplified in order to represent the data in a sensible qualitative way. This mostly concerned the interventions' effectiveness in each experimental group, opioid regimens, and intervention mechanisms. For example, intervention in an experimental group was categorised as "effective" if a statistical difference to the opioid-only group could be shown at any point with any type of QST. Yet, categorising experimental groups as "effective" or "not effective" oversimplifies the real spectrum of efficacy. This categorisation hides findings such as if an intervention was not found effective with all types of QST that were used, or if an intervention was only minimally effective at a certain time point. Ideally, a meta-analysis of the trials should be conducted, as statistical comparisons between the interventions would allow a more realistic representation of the efficacies. Moreover, opioid regimens were categorised as "acute" if an opioid was administered for less than one day and "chronic" if an opioid was administered for several days. However, the separation between acute and chronic administration is slightly arbitrary and was mostly used to emphasise the different lengths of the regimens. The majority of the trials with an acute regimen administered the opioid within one 1 h and only a few administered the opioid up to 3 h, whereas all of the chronic regimen trials administered the opioid for at least 24 h. When quantifying the efficacies of the interventions, it should be analysed whether the length of opioid use has any effect on the intervention ranking or doses required for OIH attenuation. Furthermore, the OIH attenuation mechanism and the mechanism group of each intervention were categorised based on the authors' findings and conclusions presented in the eligible articles and no further studies in the literature were explored. Some of the articles investigated the potential mechanism via in vitro experiments or using transgenic animal models, whereas some cited others' research. In a few articles, the mechanism suggested was based on an educated guess. Hence, the mechanism classification presented may provide a simplified view of the actual mechanisms, and alternative suggestions could be found in additional research.
The long-term objective of reviewing the preclinical trials investigating pharmacological interventions for OIH is to identify the most effective agents that could be used for humans. However, generalising the findings from animals to humans is not straightforward. According to the Leenaars et al. [108] review, preclinical to clinical translational success varies considerably, and the translational success tends to be unpredictable. Additionally, preclinical trials do not incorporate factors such as culture or human psychological phenomena, which is why RCTs should be used to evaluate if the most effective OIH interventions vary in different contexts. Nevertheless, preclinical trials have the strength of having fewer uncontrolled variables compared to trials with humans, which can be beneficial in generalising results across different situations.

Conclusions
In summary, this scoping review aimed to identify and describe all the preclinical trials investigating pharmacological interventions for OIH caused by remifentanil, fentanyl, or morphine as the first step towards evaluating whether the most effective OIH interventions are different for different opioids. Plenty of preclinical trials and interventions were identified. However, many of the interventions were studied in only one trial, which means that the evidence behind each intervention is not very strong. Furthermore, since only a few of the trials can provide direct comparisons of effectiveness, the interventions could be indirectly compared in a meta-analysis to identify the most effective ones for each opioid. In addition, conducting more preclinical trials comparing different interventions is recommended. Additionally, in the current literature, very few combination interventions were investigated, which could be addressed in future trials. NMDAR inhibition was found to be the most-studied mechanism for OIH attenuation for all of the opioids. Yet, the abundance of evidence alone cannot be used to conclude that the mechanism is necessarily the most effective for OIH caused by one or all of the opioids. In the end, these preclinical trials have provided several successful OIH interventions whose effectiveness for clinical use could be tested in RCTs.     (2), interventions (3), and their combinations (4).