Efficacy of Essential Oils in Pain: A Systematic Review and Meta-Analysis of Preclinical Evidence

Background: The demand for essential oils (EOs) has been steadily growing over the years. This is mirrored by a substantial increase in research concerned with EOs also in the field of inflammatory and neuropathic pain. The purpose of this present systematic review and meta-analysis is to investigate the preclinical evidence in favor of the working hypothesis of the analgesic properties of EOs, elucidating whether there is a consistent rational basis for translation into clinical settings. Methods: A literature search has been conducted on databases relevant for medical scientific literature, i.e., PubMed/MEDLINE, Scopus, and Web of Science from database inception until November 2, 2020, following the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) criteria for systematic reviews and meta-analyses. Results: The search was conducted in order to answer the following PICOS (participants/population, interventions, comparisons, outcomes, and study design) question: are EOs efficacious in reducing acute nociceptive pain and/or neuropathic pain in mice experimental models? The search retrieved 2,491 records, leaving 954 studies to screen after the removal of duplicates. The title and abstract of all 954 studies were screened, which left 127 records to evaluate in full text. Of these, 30 articles were eligible for inclusion. Conclusion: Most studies (27) assessed the analgesic properties of EOs on acute nociceptive pain models, e.g. the acetic acid writhings test, the formalin test, and the hot plate test. Unfortunately, efficacy in neuropathic pain models, which are a more suitable model for human conditions of chronic pain, had fewer results (only three studies). Moreover, some methodologies raised concerns in terms of the risk of bias. Therefore, EOs with proven efficacy in both types of pain were corroborated by methodologically consistent studies, like the EO of bergamot, which should be studied in clinical trials to enhance the translational impact of preclinical modeling on clinical pain research.


Rationale
Essential oils (EOs) containing components in exact proportion contributing synergically to the whole plant effect, have been used in traditional medicine for centuries since The Divine Farmer's Materia Medica, the first text of Chinese Traditional Medicine, representing a form of combinatorial medicine (Li and Weng, 2017). The search for natural and green products is constantly increasing the use of essential oils and the demand for these products from developing countries. There has been a remarkable increase in the import of EOs by the European market from  and it is estimated that the demand for essential oils in the global market will grow by 7.5% from 2020 to 2027 (GVR, 2020). These data are mirrored by the steady increase of research on EOs that pave the way for the development of these products.
Identifying the year 1880 as this field emerged (Wood and Reichut, 1880), we found a remarkable increase in publications concerned with EOs up to 2020 ( Figures 1A,B) (see also (Scuteri et al., 2017a)). EOs have shown several beneficial properties, many of which concern the treatment of neurologic diseases, mood disturbances, and pain. Modulation of the γ-aminobutyric acid (GABA) neurotransmission and blockade of neuronal voltage-gated sodium channels (Na + channels) as well as activity on serotonergic neurotransmission are proposed as mechanisms involved in the action of EOs endowed with anxiolytic and anti-nociceptive properties like bergamot essential oil (BEO) Scuteri et al., 2018a;Scuteri et al., 2019a;Scuteri et al., 2019b;Rombolà et al., 2019;Rombola et al., 2020), lavender essential oil (LEO) (Lopez et al., 2017), and melissa (lemon balm) (Abuhamdah et al., 2008;Awad et al., 2009). The cholinergic system is targeted by extracts of plants as sage (Perry et al., 2000;Savelev et al., 2003;Savelev et al., 2004), ginkgo (Stein et al., 2015;Zhang et al., 2018), and lemon balm (Dastmalchi et al., 2009;Guginski et al., 2009), showing therapeutic potential for diseases like dementia. The gathered evidence shows the potential benefits of EOs in the treatment of pain in fragile patients for whom several drugs can be more harmful, e.g. in aging or chronic neurologic diseases such as dementia (Achterberg et al., 2020). Pain is associated with mood disturbances (Evans, 1987;Husebo et al., 2011) influenced by aging (Hamm and Knisely, 1985;Scuteri et al., 2020a) and neuropathology (Scherder et al., 2003) and its treatment represents a field of strong inappropriateness in patients suffering from Alzheimer's disease. (Scuteri et al., 2017b;Scuteri et al., 2018b;Achterberg et al., 2020;Scuteri et al., 2020f). Therefore, aromatherapy represents a fundamental tool for the safer handling of pain.
Despite a large amount of continuously growing research on EOs, a real translation of aromatherapy into clinical settings and the treatment of pain has not occurred. Research efforts have aimed to discover the mechanisms at the root of the analgesic activity of EOs, often focusing on the single components commonly present in different plant oils e.g., linalool, limonene, pinene, eugenol, and cinnamal. For instance, linalool, limonene, and pinene contribute to the anxiolytic and antidepressant properties of some EOs (see (Lizarraga-Valderrama, 2020)). In particular, some natural components of plants have been suggested as possible candidates for an analgesic action in neuropathic pain . However, the strongest effect of EOs is due to the whole phytocomplex made up of various plant components that need to be present in a precise ratio to exert the so called entourage effect (Ribeiro, 2018). Definite combination of the constituents of EOs is necessary, but further studies are needed to highlight the exact active composition for each EO. The EOs of the species Citrus contain volatile components (85-99%), most abundantly terpenoids, and a non-volatile fraction including coumarins i.e. bergapten inducing phototoxicity (Zaynoun et al., 1977). Thus, the EO of bergamot has been deprived of bergapten (Bagetta et al., 2010), but is still endowed FIGURE 1 | Research in the field of essential oils (EOs) over the years.(A,B) Increase of research in the field of essential oils (EOs). (A) A PubMed advanced search using the key word "essential oils" combined with the dates of publication from 1880 to present through the Boolean operator AND has retrieved an increase from 106 to 17,212 (date of last search November 19, 2020) of results. The first interval "1880-1950" is wider because no great amount of research in this field has been detected up until the 1950s. (B) Data are presented per year of publication based on search query "essential oils" (date of last search November 25, 2020). Modified from (Scuteri et al., 2017a). with its characteristics. The EO of bergamot can modulate the synaptic level of glutamate and this occurs when it is used as a bergapten-free fraction (Morrone et al., 2007). Hence, a mixture of monoterpene hydrocarbons within the volatile fraction may be responsible for bergamot analgesic activity since glutamate is significantly involved in the pain descending pathway due to metabotropic glutamate receptors mGluR7 and mGluR8 (Boccella et al., 2020). The novel phytocannabinoid cannabidihexol, with the terpenophenolic core of cannabidiol and Δ9-tetrahydrocannabinol, has proven to significantly reduce the late phase of the formalin test at low doses in C57BL/6J mice (Linciano et al., 2020). Cannabidiol oil has been demonstrated to reduce traumatic brain injury-induced allodynia (Belardo et al., 2019). Certain EOs have been proven to have enhanced efficacy if combined: e.g., peppermint and caraway oil are significantly effective on post-inflammatory visceral hyperalgesia only when used in combination (Adam et al., 2006). Likewise, the route of administration and the time of exposure can influence the effects of EOs (Scuteri et al., 2018a;Koyama and Heinbockel, 2020). Moreover, some EOs are efficacious in a preclinical setting (Sarmento-Neto et al., 2015), but often only in a definite model of pain, usually acute e.g. the acetic acid-induced writhings, that does not find a significant counterpart in clinic. Furthermore, EOs are often administered as gavage or for inhalation not always allowing an exact determination of the dose.
Clinical trials in aromatherapy are few, small and methodologically limited, hence it is not always possible to draw rigorous conclusions, particularly in dementia. As recently demonstrated in a Cochrane systematic review by Ball et al. (2020), the design, reporting and consistency of outcome measurement have been identified as the weakest points and need to be improved in the future. Thus, despite accumulating preclinical and clinical evidence for EOs (Scuteri et al., 2020d) and nutraceuticals (Scuteri et al., 2020e) in lots of forms and supplements, which have been studied in several neurodegenerative conditions, a sound rationale for their clinical use, especially in treating chronic pain (Lakhan et al., 2016), has not yet emerged.

Objectives
The present systematic review and meta-analysis aimed to verify the working hypothesis that EOs have analgesic properties by investigating preclinical evidence in favor of the latter, to understand whether there is a consistent rational basis for clinical translation. For this purpose, the objective was to assess the efficacy of EOs in preclinical models of both nociceptive and neuropathic pain through the PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) Moher et al., 2009) criteria for systematic reviews and meta-analyses. The systematic review and meta-analysis focuses on the following PICOS (participants/ population, interventions, comparisons, outcomes, and study design) question: are EOs efficacious in reducing acute nociceptive pain and/or neuropathic pain in mice experimental models? In particular, this work aimed at evaluating: • analgesic effectiveness (outcome); • of EOs with a known composition (interventions), and not single components or extracts, administered intraperitoneally (i.p.) or subcutaneously (s.c.) to allow determination of the exact dose and reproducibility; • in male mice subjected to nociceptive or neuropathic pain models (participants/population); • with respect to providing a vehicle or other treatments (comparators); • in studies designed according to legislation to minimize the suffering of animals (study design).
To the best of our knowledge, this is the first meta-analysis of preclinical studies on the analgesic effects of EOs interventions in models of both nociceptive and neuropathic pain.

Protocol
The search strategy and extraction of data to answer to PICOS question followed the PRISMA Moher et al., 2009) criteria. Due to the nature of preclinical animal intervention systematic review and meta-analysis, the latter aims at investigating the consistency of the body of evidence for clinical translation without an outcome of clear human relevance. For this reason, it has not been registered in the International prospective register of systematic reviews PROSPERO. However, statistically analyzing basic research independent studies testing the same hypothesis with comparable parameters can: determine its consistency allowing to study that phenomenon in a larger sample surmounting the issues concerned with small sample sizes; correct confounders; improve reproducibility (Editorial, 2016). Thus, a systematic review and meta-analysis is fundamental to establish a real possible clinical translation of a preclinically studied effect since it can highlight whether it has been consistently proven with the most reliable human disease modelled approach. Two independent researchers ran the search in agreement with the previously established protocol and inclusion and exclusion criteria, including double-checking the retrieved results, and any conflicts found by them were resolved by a third author.

Inclusion Criteria
The analysis included studies assessing the antinociceptive effect of EOs, administered i. p. or s. c. to allow determination of the exact dose and reproducibility, with a known percentage of components on male mice subjected to nociceptive or neuropathic pain models. Compliance with animal welfare regulations was an inclusion criterion of the utmost importance. The studies included needed to be designed according to legislation to minimize animal suffering. Either acute nociceptive or neuropathic pain models are included. Independently of the model used, the outcome of the study had to be antinociception for eligibility.

Exclusion Criteria
Studies on species different from mice or any strains and female gender were not eligible. The use of different species and genders would not allow comparison and the number of papers examining pain in non-rodent species is very small. Papers in which extracts or single plant components are used were excluded. Studies that did not consider ethics were excluded. Narrative or systematic reviews and meta-analysis, in vitro studies, abstracts and congress communications, proceedings, editorials, book chapters, and studies not published in English and not available in full text were not eligible.

Information Sources
A literature search was performed on PubMed/MEDLINE, Scopus, and Web of Science. Embase could not be searched as it was not freely/institutionally available. No restriction of publication date was applied and databases were searched for records matching the search strings used from their inception. The date of the last search was November 2, 2020. After the elimination of duplicate records, the first screening evaluated the title and abstract, and then the full text was assessed to define inclusion in qualitative and/or in quantitative synthesis.

Data Collection Process
conflicts were raised. The selection process is illustrated in the PRISMA flow diagram (Figure 2).

Synthesis, Risk of Bias, and Statistical Analysis
A systematic and narrative synthesis of the results, according to the Cochrane Consumers and Communication Review Group guidelines (Ryan, 2019http://cccrg.cochrane.org, March 13, 2019.) was carried out. The risk of bias (internal validity) and the quality of the studies was assessed by two independent researchers through tools specific to preclinical animal studies like the Systematic Review Center for Laboratory Animal Experimentation (SYRCLE's) risk of bias (RoB) tool (Hooijmans et al., 2014) and the Collaborative Approach to Meta-Analysis and Review of Animal Data from Experimental Studies (CAMARADES) checklist for study quality (Macleod et al., 2004). Any discrepancies were resolved through consensus or with the help of a third author.
Meta-analyses were conducted using Cochrane Review Manager 5.3 (RevMan5.3; Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration). A minimum of five articles per outcome measure was required according to the systematic review protocol for animal intervention studies by SYRCLE. When the tests included in articles were multiple and performed at different times and doses, only the most significant time point for pain development and progression in the specific model was considered for meta-analysis and only data related to the most efficacious dose were included. Studies expressing the analgesic outcome in a comparable way were included in the meta-analysis. Data available and comparable, but not expressed with the same measure of effect size as proportional reduction of outcome in treated animals relative to untreated controls were converted in mean and standard deviation to allow statistical comparison. Data not available and not extractable from graphs using digital ruler software, e.g., PlotDigitizer 2.6.9, were excluded from quantitative analysis. The Higgins I 2 value was calculated to assess the heterogeneity of studies (Higgins and Thompson, 2002). Differences were presented as risk ratios (RR) including 95% confidence intervals (CI), using a random effect model (DerSimonian and Kacker, 2007) to manage the eventual heterogeneity of the studies and to assess intra-and interstudy variation. Publication bias was assessed through Egger's linear regression test to measure funnel plot asymmetry, adjusted through the "trim and fill" method (Egger et al., 1997;Duval and Tweedie, 2000;Sterne and Egger, 2001).

Selection Process and Data Collection
The search retrieved 2,491 results from databases and there were no results from additional searches. The records were screened for duplicates, leaving 954 studies. Title and abstract screening led to an initial exclusion of narrative or systematic reviews and meta-analysis, in vitro studies, abstracts and congress communications, proceedings, editorials, book chapters. This left 127 records in full text. Among these, two had to be excluded because the text was in Chinese (Li et al., 1991;Chen et al., 2011) and one was excluded because it was written in Spanish (Do Nascimento Silva et al., 2018). After full text screening, 30 studies were included in qualitative analysis: 40 studies were not available in full text and 54 were excluded because they did not meet inclusion criteria because of species used, route of administration, composition, or lack of compliance with animal welfare regulations. For instance, the study by Ali et al. (2012) in which the EO of Nepeta pogonosperma Jamzad et Assadi was proven to have significant efficacy at different doses in the tail-flick and formalin test in Wistar rats was therefore not eligible. Among the records included in qualitative analysis, eight were included in quantitative synthesis, reporting comparable outcomes and the exact number of animals used. The process of literature search and screening was illustrated in the PRISMA flow diagram (Moher et al., 2009)

Qualitative Synthesis
The data obtained from the 46 studies included in the qualitative analysis were grouped according to Cochrane Consumers and Communication Review Group guidelines (Ryan, 2019 http:// cccrg.cochrane.org, March 13, 2019 (accessed DATE).). These groups were based on the experimental pain model used in 1) EOs showing analgesia in nociceptive models, and 2) EOs with analgesic properties in neuropathic pain. The majority (27/30) of the studies used an acute nociceptive model. Studies providing a range and not an exact number of animals per group were not considered eligible for quantitative analysis. Studies expressing the analgesic outcome in a not comparable manner to the majority were excluded from the meta-analysis. The main characteristics of the studies with reference to the PICOS question are reported in Tables 1, 2.

Essential Oils Endowed With Efficacy in Acute Nociceptive Models
Based on the obtained results, several EOs showed analgesic activity in acute nociceptive tests like the acetic acid writhings test, hot-plate test, and the formalin test, with the latter very useful since it includes features of both peripheral and central pain. In the study by Anaya-Eugenio et al. (2016) the EO of artemisia ludoviciana Nutt (Asteraceae) exerted dose-dependent antinociceptive activity in the hot-plate and the formalin test. It was less potent than the reference drug morphine and antagonism studies have revealed that it was inhibited by the non-selective opioid receptor antagonist naloxone. Inula britannica L (Asteraceae) has shown analgesia in the acetic acid writhings test, in the formalin test, in the tail-flick, and the glutamate test (Zarei et al., 2018). This effect is reversed by naloxone and potentiated by L-arginine, therefore all the studies performed with negative and positive controls highlighted the involvement of the opioid system and NO pathway (Zarei et al., 2018).
The EO of Myrcia pubiflora DC., Myrtaceae (Andrade et al., 2012) has demonstrated analgesic efficacy in the acetic acid writhings test and the formalin test, but not in the hot plate test. From the same family, the EO of Eugenia candolleana DC (Myrtaceae) reduced the number of writhings and licking behavior in the second phase of the formalin test in a dosedependent manner (only at the dose of 100 mg/kg in the first phase, but not the nociceptive reaction in the hot-plate test (Guimaraes et al., 2009)).
Clove bud oil (Eugenia caryophyllata, Myrtaceae) significantly reduced formalin-induced pain behavior but affected tail-flick response in a variable way (Halder et al., 2012). The study by Bae et al. (2020) considered basil for its i. p. administration and demonstrated analgesic properties linked to action on δand µ-opioid pathways. Moreover, it provides orofacial antinociception at high doses (Venâncio et al., 2011). Aristolochia trilobata L. demonstrated strong analgesia in the formalin test and was comparable to morphine in the acetic acid test (Quintans et al., 2017).
The EO of Croton conduplicatus Kunth (Euphorbiaceae) has shown efficacy (de Oliveira Júnior et al., 2017;de Oliveira et al., 2018): in the acetic acid test; on the formalin-induced nociceptive behavior at all the doses and in both phases, with effect antagonized by naloxone; on nociception in term of latency time at the highest dose (50 (deOliveira et al., 2018) and 100 (de Oliveira Júnior et al., 2017) mg/kg) in the hot-plate test. In particular, the mechanism of action of this EO has been proposed to be influenced by ATP-sensitive K+ channels, opioid and cholinergic systems (de Oliveira Júnior et al., 2017;de Oliveira et al., 2018).
Croton cordiifolius Baill (Euphorbiaceae) also had effective results in acetic acid, formalin, and glutamate but not the capsaicin test. This antinociceptive effect was independent on naloxone (Nogueira et al., 2015). Croton adamantinus Müll. Arg. showed a strongly effective comparison with morphine in reducing licking and was more efficacious than indomethacin in decreasing abdominal contortions (Ximenes et al., 2013). Of the study by Hajhashemi et al. (2009) only the experiments using the EO i. p. and on mice could be included in the analysis, showing the effectiveness of Heracleum persicum to be almost comparable to indomethacin in the reduction of the number of writhings.
In the study by Jahandar et al. (2018) only the experiments performed on mice were considered. Pycnocycla bashagardiana (Apiaceae) has not proven analgesic but anti-inflammatory properties. In another study by Ulku Karabay-Yavasoglu et al. (2006) only experiments with the formalin test in mice were considered. The EO of Satureja thymbra L (Lamiaceae) was demonstrated to have antinociceptive efficacy in both the early and late (also at a lower dose) phases of the formalin test (Ulku Karabay-Yavasoglu et al., 2006).
In the study by Katsuyama et al. (2015) the EO of bergamot (Citrus bergamia Risso) demonstrated significant dosedependent analgesia in both phases of the formalin test, only when administered in the ipsilateral hindpaw and antagonized by naloxone hydrochloride and methiodide (not able to cross the blood brain barrier), suggesting the involvement of peripheral opioid mechanisms. This was earlier observed in the capsaicin test in which it also enhanced morphine analgesia (Sakurada et al., 2011).
Neroli (Citrus aurantium L.) significantly increases reaction time (at 40 mg/kg) in the hot-plate test and significantly decreased the number of writhings in the study by Khodabakhsh et al. (2015), with the latter effect potentiated by L-nitro arginine methyl ester (L-NAME). In the study by Khalid et al. (2011) the EO of Zingiber zerumbet (L.) Smith, dosedependent and comparable to acetylsalicylic acid, inhibited the nociceptive response to capsaicin, acetic acid, glutamate, and phorbol 12-myristate 13-acetate (PMA). Eucalyptus EO has significantly reduced licking time in the second phase of the formalin test in the study by Lee et al. (2019), and this effect was mediated by the opioid system. It also reduced the number of writhings in a dose-dependent manner but did not display activity on thermal hyperalgesia (Lee et al., 2019). In the study by Lima et al. (2012) the EO of Chrysopogon zizanioides L (Roberty, Poaceae) produced antinociception similar to morphine in the acetic acid test, and this effect was partially reversed by naloxone. Moreover, it reduced the licking time in the second phase of the formalin test, but it did not demonstrate any effects in the Hargreaves' test.
A common trait is the presence of antiinflammatory analgesia devoid of thermal anti-hyperalgesic effect. The EO of Zhumeria majdae Rech. F. and Wendelbo (Lamiaceae) has displayed doserelated antinociceptive properties in the acetic acid and in the hot-plate test (Miraghazadeh et al., 2015). Chamaecyparis obtuse has also shown analgesia in the writhings and in the formalin, but not in the hot-plate test (Park et al., 2015). Furthermore, in the study by Mishra et al. (2010) Senecio rufinervis D.C. (Asteraceae) produced significant and dose-dependent inhibition of writhes and thermal hyperalgesia. In the study by Sharif et al. (2020) Tanacetum balsamita (Compositae) presented an antihyperalgesic effect. The antinociceptive properties exerted by Xylopia laevigata (Annonaceae) in the acetic acid and in the formalin test have not proven dependency on opioid pathways (Queiroz et al., 2014). The antinociceptive effect of Bunium persicum (Boiss.) is reversed by naloxone and attenuated by chlorpheniramine and cimetidine (Zendehdel et al., 2015), thus confirming the complex neuromodulation and the involvement of histamine in nociception (Hayashi et al., 2020). The main features of the studies on EOs analgesia in nociceptive models are summarized in Table 1.

Essential Oils Endowed With Efficacy in Neuropathic Models
Studies assessing the analgesic properties of EOs in neuropathic pain models are fundamental because these painful conditions are the most appropriate to model chronic neuropathic pain in humans. In the study by Komatsu et al. (2018) the EO of bergamot (Citrus bergamia Risso) was demonstrated to reduce partial sciatic nerve ligation (PSNL)-induced mechanical allodynia on the seventh post-operative day, in which it peaks (Kusunose et al., 2010). In the study by Kuwahata et al. (2013) the EO of bergamot increased mechanical thresholds dosedependently and significantly at a dose of 20 μg/paw (Kuwahata et al., 2013). Moreover, this anti-allodynic effect is stronger than that of comparable doses of     morphine, of which the EO of bergamot enhances the activity (Kuwahata et al., 2013), and it was reversed by naloxone methiodide, peripherally μ-opioid receptor preferring antagonist, β-funaltrexamine hydrochloride, selective μ-opioid receptor antagonist, and β-endorphin antiserum, but not by the non-selective δ-opioid receptor antagonist naltrindole and by the selective κ-opioid receptor antagonist nor-binaltorphimine. Importantly, the study by Hamamura et al. (2020) in which the EO of bergamot was administered s. c. with an osmotic pump to allow a continuous delivery devoid of smell during PSNL, demonstrated that the antiallodynic effect of this EO is systemic and does not depend on olfactory stimulation. In this study  the increase of planar activity during the light period induced by PSNL, with the maximum effect at the seventh postoperative day and like allodynia, was shown to be abolished by continuously administered EO. This effect is antagonized by naloxone hydrochloride. Observation lasting 14 days with a theoretical duration of the osmotic pump of one week can mimic administration during chronic pain. The main features of the studies on EOs anti-allodyinic properties are summarized in Table 2.

Risk of Bias Assessment
The studies included in the qualitative analysis were assessed for methodological quality according to the SYRCLE's RoB tool (Hooijmans et al., 2014) and the CAMARADES checklist (Macleod et al., 2004;Hooijmans et al., 2014;Suokas et al., 2014), based on the Cochrane RoB (Sterne et al., 2019). These items comprise all the possible forms of bias. 1) Selection bias-sequence generation (allocation sequence able to produce comparable groups). 2) Selection bias-baseline characteristics (comparable and not adjusted for confounders in the analysis).
3) Selection bias-allocation concealment (during the enrollment). 4) Performance bias-random housing and randomization during the study. 5) Performance bias-blinding of investigators during the study. 6) Detection bias-random outcome assessment. 7) Detection bias-blinding of outcome assessors. 8) Attrition bias (animals eventually excluded from outcome assessment). 9) Reporting bias-reports free of selective outcome reporting. Finally, 10) other sources of bias: lack of evidence of induced pain using the selected behavioral outcome measure before EO administration and examination (i.e., sham procedure), clear description of methods with number of animals used, attention to circadian regulation for behavioral studies, use of the same observer for behavioral tests, use of control and positive and negative control drugs, sample size calculation, statement of conflict of interest, statement of compliance with animal welfare regulations and attention to ethics. In terms of the two items regarding selection bias, no study reported the method of allocation and, even though they conducted baseline measures, none of the studies describe how experimental groups were composed to ensure homogeneity and consistency. Only the study by Lima and collaborators (Lima et al., 2012) in which mice with baseline latencies of more than 10 s, and studies by de Oliveira Júnior and colleagues (de Oliveira Júnior et al., 2017;de Oliveira et al., 2018) of more than 20 s, at the hot-plate were excluded from the experiments.
As reported in Table 1, five studies (Guimaraes et al., 2009;Andrade et al., 2012;Lima et al., 2012;de Oliveira Júnior et al., 2017;de Oliveira et al., 2018) adopted random housing of mice. The paper by Khodabakhsh et al. reported no randomization of mice but only of rats, which are not included in this systematic review and meta-analysis (Khodabakhsh et al., 2015). Mice were tested in a randomized order in studies by collaborators andKatsuyama et al., 2015 (Sakurada et al., 2011;Katsuyama et al., 2015). In the study by Bae et al. (2020) mice were randomly assigned to groups. The study by Khalid et al. (2011) used a blind, randomized design. Mice were randomly assigned to groups and experiments were performed in a blind manner in the study by Quintans and coworkers (Quintans et al., 2017). Moreover, in the study by Ximenes et al. (2013), the observation was conducted by a blind observer, but the number of animals used for behavioral testing was not reported, only for histological assays. Otherwise, the number of animals per group was reported, but studies that provided a range and not an exact number were not considered eligible for quantitative analysis. Attrition and reporting biases cannot be assessed from the full text of the included studies. Importantly, sham procedure and the certainty of exact execution of the pain model is present only in studies on allodynia, i.e., the studies by Hamamura et al. (2020), Komatsu et al. (2018a), and Kuwahata et al. (2013). Attention to the circadian rhythm in behavioral testing was reported by the following studies: (Andrade et al., 2012;Quintans et al., 2017;Guimaraes et al., 2009;Halder et al., 2012;Katsuyama et al., 2015;Khodabakhsh et al., 2015;Miraghazadeh et al., 2015;Sakurada et al., 2011;Sharif et al., 2020;Zarei et al., 2018;Zendehdel et al., 2015;Komatsu et al., 2018b;Kuwahata et al., 2013). All the studies used control and positive and negative modulators. Importantly, multiple controls were used in the following studies (Katsuyama et al., 2015;Komatsu et al., 2018b;Kuwahata et al., 2013;Sakurada et al., 2011). Behavioral testing was conducted by the same observers in the following studies (de Oliveira Júnior et al., 2017;de Oliveira et al., 2018;Venãncio et al., 2011). Sample size calculation was not reported and the conflict of interest statement is present only in eight studies (Queiroz et al., 2014;Nogueira et al., 2015;de Oliveira Júnior et al., 2017;de Oliveira et al., 2018;Jahandar et al., 2018;Zarei et al., 2018;Lee et al., 2019;Hamamura et al., 2020). This could be due to the lack of requirement of these aspects in journals in the last few years. A statement of compliance with animal welfare regulations is reported in all the studies since it is an inclusion criterion. Moreover, six studies (Ulku Karabay-Yavasoglu et al., 2006;Venâncio et al., 2011;Queiroz et al., 2014;Khodabakhsh et al., 2015;Miraghazadeh et al., 2015;Sharif et al., 2020) also stated that they used each mouse only once, thus proving particular attention to animal welfare. Importantly, only the study by Hamamura et al. (2020) reported acclimatization to lighting conditions for one week and that an observation period of 14 days can model examination during chronic pain.

Meta-Analysis
This meta-analysis comprises eight studies for a total of 140 mice. The studies were considered comparable when the analgesic outcome was expressed as mean ± standard error of the mean (SEM) since these measures could be converted for meta-analysis in mean and standard deviation (SD). Moreover, only studies reporting the exact number of animals per group were included in quantitative analysis. Studies investigating the same pain model were considered. The formalin test pain model was chosen since it provides a biphasic nociceptive response. Due to the sensitization processes occurring during the second phase, the study on mechanical allodynia expressed has been included (Komatsu et al., 2018b). The results of the forest plot favor the analgesic efficacy of EO (Mean difference MD −59.77; 95% CI (−93.32) -(−26.22); I 2 94%; p < 0.00001; Figure 3), but need to be carefully examined because of the extremely high heterogeneity, which is also confirmed by the asymmetry of the funnel plot analysis standing for high risk of publication bias, small studies and high differences in study precision.

DISCUSSION
Interest in the use of EOs and aromatherapy has been continuously growing during the last few decades in parallel with preclinical research. However, in spite of all this effort of preclinical research, it is necessary to establish whether there is a strong rationale for the clinical use of EOs. This issue is even more controversial in the field of pain relief since the use of aromatherapy could reduce the dose of painkillers endowed with serious side effects, particularly in under studied areas of neuropathic pain, like opioids Scuteri et al., 2020b). Alternative pain treatments could increase time in treatment before the loss of efficacy. This is relevant to fragile populations, e.g., patients suffering from dementia, who are often undertreated compared to cognitively intact counterparts, more so during the Sars-CoV2 pandemic (Scuteri et al., 2020c). This systematic review and meta-analysis assesses the efficacy of EOs in preclinical models of acute inflammatory nociception and neuropathic pain to understand if there is a rational basis for clinical translation. Several EOs from multiple families were found to be efficacious, in particular, croton and bergamot EOs have been extensively studied. It is noteworthy that 27 out of the 30 studies included in the qualitative analysis were only performed on acute pain models like writhings and the hot plate test. These tests are very useful since they are easy to conduct and provide fast results, but they do not resemble clinical pain conditions. Taking this into account, the quantitative analysis only includes studies on formalin test, which is more similar to clinical conditions due to its biphasic nature, and the only study on mechanical allodynia that could compare to the other seven included.
All these studies included in this review have a different experimental design and most of them present serious concerns in terms of selection, performance, and detection biases. Most studies do not adhere to the guidelines for Animal Research: Reporting In Vivo Experiments (ARRIVE), which are fundamental for accurate in vivo preclinical research (Rice et al., 2013). Another methodological aspect responsible for bias in the meta-analysis is that control groups were often used in more than one experiment, and studies including multiple comparisons can introduce errors. Thus, this systematic review and meta-analysis points to the importance of appropriate in vivo modeling to enhance the translational impact of pain research. Future research is necessary to improve the methodological quality and homogeneity of studies.
The results of the meta-analysis highlighted the efficacy of EOs in preclinical pain, but these data are downgraded due to the high heterogeneity of the studies. In particular, the analyzed EOs present the analgesic efficacy required by the recommendations of the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) (Turk et al., 2003), according to which a decrease in pain is defined as clinically meaningful if it accounts for a 30% to 36% reduction. However, this is referred to chronic pain and this systematic review and meta-analysis have found that only the EO of bergamot had proven efficacy both in nociceptive and in neuropathic pain models. Moreover, it was also studied for 14 days, an experimental setting suitable for modeling chronic pain . Another important issue is that the consolidated data come from hypothesis-generating completely original preclinical studies and that they are then confirmed by hypothesis-driven studies (Mikolajewicz and Komarova, 2019). In this case, the EO of bergamot was confirmed to have strong analgesic properties in some of the most used and reliable models of inflammatory pain, i.e., formalin and capsaicin test in different experiments, sharing with most EO mechanisms involving opioid neurotransmission, and also in the PSNL. To the best of our knowledge, this is the first meta-analysis of preclinical studies on the analgesic effects of EOs and its working hypothesis was verified for bergamot EO, which could represent an important pharmacological tool for pain management in clinical settings. Along with clinical translations, more efforts are required to standardize in vivo preclinical studies in the field of pain research to allow for consistent research able to elucidate the mechanisms responsible for the analgesic properties of EOs.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article.

AUTHOR CONTRIBUTIONS
DS, GB, TS, SS, and MTC. conceived the study. All Authors participated in preparation and read and approved the final manuscript.

FUNDING
DS is a post-doc recipient of a research grant salary as part of the research project (Tutor: GB) "Pharmacoepidemiology of drugs used in the treatment of neuropsychiatric symptoms and pain in people aged (over 65) with dementia" funded by Calabria Region (POR Calabria FESR-FSE 2014/2020-Linea B) Azione 10.5.12.