Molecular neuroscience at its “high”: bibliometric analysis of the most cited papers on endocannabinoid system, cannabis and cannabinoids

Background Cannabis, cannabinoids and endocannabinoids are heavily investigated topics with many articles published every year. We aimed to identify the 100 most cited manuscripts among the vast literature and analyze their contents. Methods Web of Science (WoS) Core Collection was searched to identify the 100 most cited relevant manuscripts, which were analyzed with reference to (1) authorship, (2) institution, (3) country, (4) document type, (5) journal, (6) publication year, (7) WoS category, and (8) citation count. Semantic content and citation data of the manuscripts were analyzed with VOSviewer. Results The most cited manuscripts were published between 1986 and 2016, with the majority being published in the 2000s (n = 51). The number of citations for the top 100 articles ranged from 469 to 3651, with a median citation count of 635.5. The most prolific authors were Vincenzo Di Marzo (n = 11) and Daniele Piomelli (n = 11). The major contributing countries were USA (n = 49), Italy (n = 22), UK (n = 19), and France (n = 11). The most prolific institutions were University of California (n = 14), National Research Council of Italy (n = 12) and National Institutes of Health USA (n = 12). The manuscripts consisted of original articles (n = 75), reviews (n = 24) and a note (n = 1). The most dominant journal was Nature (n = 15). The major WoS categories associated were Multidisciplinary sciences (n = 31), Neurosciences (n = 20), Pharmacology / Pharmacy (n = 16), and General / Internal Medicine (n = 11). Conclusions The top-ranked manuscripts among the 100 were concerning analgesia, weight loss, long-term potentiation, depolarization-induced suppression of inhibition, opiates and other topics. Cannabinoid type 1 (CB1) receptor was studied by more of the top 100 papers in comparison to cannabinoid type 2 (CB2) receptor. The most frequently mentioned chemicals in these publications were 2-arachidonoylglycerol, tetrahydrocannabinol, and anandamide. Together, these manuscripts comprise the most highly cited publications in the topic, literally the molecular neuroscience at its “high”.


Background
Cannabis sativa L. (cannabis) has been cultivated by humanity for more than 2000 years (Mercuri et al. 2002). It is estimated that 7.2% of the USA population abused cannabis during their lifetime (Stinson et al. 2006). The main reason for its recreational use is its ability to produce euphoria, or feeling of "high" (Ashton 2001). The euphoriant effect of cannabinoids derived from cannabis is attributed to their interactions with the mammalian endogenous cannabinoid system; stimulation of the endocannabinoid system may also lead to effects such as perceptual alterations, impaired psychomotor performance, and tachycardia (Ashton 2001). On the other hand, endocannabinoids are important lipid messengers that regulate synaptic transmission and neurotransmitter release in the brain (Piomelli 2003).
The endocannabinoid system consists of G proteincoupled cannabinoid receptors, cannabinoid type 1 (CB1) and type 2 (CB2) receptors, and endogenous lipid-based neurotransmitters, known as the endocannabinoids, that target these receptors in central and peripheral nervous systems (Pertwee 2015). The CB1 receptor was discovered by Bonner and his co-workers back in 1990 (Matsuda et al. 1990) and is mainly concentrated in the brain (Pacher et al. 2006), whereas the CB2 receptor was discovered by Munro et al. in 1993(Munro et al. 1993) and is mainly found in the immune system and also the gastrointestinal system (Pacher and Mechoulam 2011). Research findings have hinted that there could be additional CB receptors; one candidate of such is GPR55, which was discovered by O'Dowd and his co-workers in 1999 (Sawzdargo et al. 1999) and later found to be activated by various cannabinoids (Ryberg et al. 2007). Other candidates may also exist, though there is still to be a consensus (Rodriguez de Fonseca and Schneider 2008).
Due to large volume of literature on the topic, the application of bibliometric analysis can facilitate better understanding of the field. Bibliometric analysis encompasses for example surveying journal editorial practice (Yeung 2017), or assessing the publication and citation data of a specific research field (Yeung et al. 2017a).
A recent bibliometric analysis of cannabis-related literature investigated six topics involving genetics, biochemistry, and biology (Matielo et al. 2018). We aimed to evaluate the literature from another perspective, by identifying the most impactful manuscripts concerning endocannabinoid, cannabis and cannabinoid. Further in the manuscript we would relate to these manuscripts as cannabis and cannabinoid-related manuscripts. By analyzing the 100 most cited articles, we aimed to provide a quick guide on the most influential research in the field, which can serve as a starting point for fellow researchers to quickly identify the high impact topics, their contributors, and outlining possible future research directions and collaborations.
In the current manuscript we described the 100 most cited cannabis and cannabinoid-related manuscripts and identified the major contributors and research themes. We also analyzed relationships between the citation count of these manuscripts and various bibliometric parameters, such as author number, reference number, and journal impact factor.

Data sources
Bibliometric data was extracted from Web of Science (WoS) Core Collection online database, a multidisciplinary database hosted by Clarivate Analytics. In November 2018, we queried WoS with string: TOPIC = ("endocannabinoid*" OR "cannabi*"). This query returned manuscripts that contain the words "endocannabinoid", "cannabis", "cannabinoid" or their derivatives in their title, abstract or keywords. WoS search engine does not distinguish uppercase and lowercase characters, so we did not include the same search words in uppercase. No additional restriction was placed on the search. The authors' WoS subscription included publications from the year 1956 forward. Therefore, publications published before that year could not be analyzed.
The manuscripts were sorted by descending citation count. Two authors (AWKY and AGA) assessed the titles and abstracts of the manuscripts to exclude irrelevant ones.

Bubble maps
The VOSviewer software was used to analyze the semantic content of titles, abstracts and keywords, relate them to citation data, and visualize the results as bubble maps (van Eck and Waltman 2009). Each bubble represents a word or phrase. The bubble size indicates the appearance frequency of the term (multiple appearances in one manuscript count as one). The bubble color indicates the mean citation count received by manuscripts containing the term. Two bubbles are closer to each other if the two terms co-appeared in the manuscripts more frequently. Only terms that appeared in at least 5 of the manuscripts were included in the figure.

Major contributors
The search resulted in 44,643 manuscripts. The 100 most-cited cannabis manuscripts were all in English. They were published between 1986 and 2016 (Fig. 1). The most prolific authors were Vincenzo Di Marzo (n = 11) and Daniele Piomelli (n = 11). The major contributing countries were USA (n = 49), Italy (n = 22), UK (n = 19), and France (n = 11). The most prolific institutions were University of California (n = 14), National Research Council of Italy (n = 12) and National Institutes of Health USA (n = 12). The manuscripts consisted of original articles (n = 75), reviews (n = 24) and a note (n = 1). The most dominant journal was Nature (n = 15). The major WoS categories associated with the manuscripts were Multidisciplinary sciences (n = 31), Neurosciences (n = 20), Pharmacology / Pharmacy (n = 16), and General / Internal Medicine (n = 11). Consistent to previous surveys on Cannabis literature, the majority of the publications focused on the biological/medicinal science instead of plant science (Matielo et al. 2018;Treister-Goltzman et al. 2018). Table 1 lists the top five most prolific authors, institutions, countries, and journals, in terms of their publication count and averaged citations per manuscript. Here, we observed that Italy was behind the USA as the second major contributor, and the Italian roots of the most prolific authors, Vincenzo Di Marzo and Daniele Piomelli. The Italian contribution in the cannabis and cannabinoids research is large compared to research in other related fields, such as ethnopharmacology (1%) (Yeung et al. 2018b), nutraceuticals (2%) (Yeung et al. 2018c), natural products in cancer studies (5.2%) (Yeung et al. 2018a), and neurosciences (5-6.5%) (Yeung 2018;Yeung et al. 2017a;Yeung et al. 2017b). In fact, pollen records has suggested that the central Italy has a long history of cultivating Cannabis for more than 2000 years (Mercuri et al. 2002). Moreover, it was estimated that 3.3-5.5 million people in Italy (with a total population of around 61 million) had used cannabis at least once (Farcomeni and Scacciatelli 2013). The abundance of C. sativa and the history of its cultivation may partly explain large Italian contribution to the cannabis-related research. It is likely that Italy's prominence in the field also results from political, regulatory or funding-related factors but analysis of such claims lies beyond scope of this work.
The additional search for identifying manuscripts that mentioned CB1, CB2 or marijuana while omitting terms endocannabinoid, cannabis or cannabinoid returned 11,436 manuscripts. Nineteen of them had citation count of over 469, the citation count of the last manuscript ranked 100th on our list. However, while we focused on molecular biology of the cannabis, most of the additionally identified 19 manuscripts were national surveys or epidemiology studies that targeted drug dependence / substance use / drug addiction, in which marijuana was only one of the many items recorded. The only relevant manuscript was a survey reporting that maternal marijuana use led to decreased infant birth weight and length (Zuckerman et al. 1989).

Citation counts
The citation counts of the identified 100 papers ranged from 469 to 3651 (mean ± SD: 845.1 ± 546.3, cumulative total citations = 84,510; Table 2). The adjusted citation count (i.e., citation count per year since publication) ranged from 16.6 to 245.5 (mean ± SD: 56.5 ± 38.4, Table 2). Devane et al. (Devane et al. 1992) published the top-ranked manuscript that reported the discovery of anandamide, a natural ligand for the cannabinoid receptors. The total citation count positively correlated with the number of authors (r = 0.232, p = 0.023), but not with journal impact factor (r = 0.170, p = 0.099), or number of references (r = − 0.044, p = 0.668). The adjusted citation count did not correlate with number of authors (r = 0.033, p = 0.753), impact factor (r = 0.128, p = 0.216) nor number of references (r = 0.038, p = 0.714). Meanwhile, the total citation and adjusted citation counts were positively correlated (r = 0.408, p < 0.001). These relationships were just partly consistent to the summary of citation analysis literature reported by Hanel and Haase (Hanel and Haase 2017), in which they reported that citation frequency was always positively correlated to impact factor and number of references, while the relationship was mixed for number of authors. The discrepancy here could be explained by the existence of possible research field-specific differences or by the fact that in the current study we analyzed only the 100 most cited manuscripts, so the results of this exploratory correlation analyses may not be the same for the literature body as a whole.

Bubble maps
There were 113 terms that appeared in the titles and abstracts of at least 5 of the 100 manuscripts (Fig. 2). The bubble map showed that manuscripts concerning analgesia seemed to have more citations than those concerning weight loss. Moreover, tetrahydrocannabinol (THC), which also has analgesic effects, also received many citations (n = 16, citations per manuscript = 972.4) (Munro et al. 1993). We further examined the dataset to look for some notable terms that might appear in fewer than 5 of the manuscripts, and found that opiate was mentioned in 2 manuscripts (citations per manuscript = 748.5), whereas the non-psychoactive compounds cannabidiol (CBD) and cannabinol (CBN) were mentioned in 2 manuscripts (citations per manuscript = 598.5) and 1 manuscript (citations = 568), respectively. There were 30 keywords that appeared in at least 5 of the 100 manuscripts (Fig. 3). The bubble map of keywords showed that the rat and mouse models were frequently used (keywords: "rat brain" (n = 23) and "mice" (n = 9)). "Long-term potentiation" (n = 9), "depolarization-induced suppression" (n = 6), and "food intake" (n = 5) were frequently mentioned concepts. Long-term potentiation is considered as one of the most crucial mechanisms underlying learning and memory (Bliss and Collingridge 1993). Rat study has revealed that long-term cannabinoid treatment can undermine reference and working memory performance, and impair long-term potentiation in the hippocampus (Hill et al. 2004). In addition, endocannabinoids may have a role in modulation of anxiety and conditioned fear via long-term potentiation (Marsch et al. 2007;   (1992). Isolation and structure of a brain constituent that binds to the cannabinoid receptor. Science, 258 (5090) Marsicano et al. 2002). Meanwhile, depolarization-induced suppression of inhibition mediated via endocannabinoids / cannabinoid type 1 (CB1) receptor is a primary cortical process that enables neurons to communicates backwards across synapses to modulate their inputs, and thus contributes to multiple forms of cortical plasticity and synaptic strengthening (Kreitzer and Regehr 2001;Ohno-Shosaku et al. 2001;Wilson and Nicoll 2001). With respect of food intake, endocannabinoids in the hypothalamus might conditionally activate CB1 receptors to maintain and regulate food intake together with leptin (Di Marzo et al. 2001).
Findings from clinical trial have demonstrated that the intake of CB1 receptor blocker, rimonabant, together with hypocaloric diet could lead to significant decrease in body weight and risk of having cardiovascular disease (Van Gaal et al. 2005). However, rimonabant was withdrawn for treatment due to its adverse effects such as causing mood swings and suicide (Christensen et al. 2007). Research is still ongoing for food intake regulation and it is proposed that peripherally restricted CB1 receptor blockers may be therapeutic in the future (Simon and Cota 2017). Moreover, the CB1 receptor was frequently the focus of research, with frequent recurrence of keywords: "CB1" (n = 6), "cannabinoid CB1 receptor" (n = 5), and anandamide (n = 17, citations per manuscript = 823.6), its famous agonist reported by the manuscript with the highest citation count (Devane et al. 1992). When appearances in titles, abstracts and keywords are considered together, CB1 appears in 23 manuscripts, whereas CB2 emerged in 10 manuscripts. One important function of CB2 receptors is modulation of mast cell activation and thus inflammation via agonist binding (Facci et al. 1995). Thus, it seems that the CB1 receptor 87 Facci, L., Dal Toso, R., Romanello, S., Buriani, A., Skaper, S. D., & Leon, A. (1995). Mast cells express a peripheral cannabinoid receptor with differential sensitivity to anandamide and palmitoylethanolamide. Proceedings of the National Academy of Sciences, 92 (8)  was studied by more of the top 100 papers in comparison to the CB2 receptor. Another frequently mentioned chemical was 2-arachidonoylglycerol (n = 9, citations per manuscript = 747.1), an endocannabinoid that mediates analgesia, hypothermia, hypomotility and modulates longterm potentiation (Hanuš et al. 2001;Long et al. 2009;Stella et al. 1997). The structures of these frequently mentioned chemicals are listed in Fig. 4.

General thoughts on existing evidence
The literature analysis has revealed that there were many rat and mice studies. The primary limitation of murine studies is they do not accurately mirror the wide spectrum of variability in the treatment response of the most human subjects tested with cannabis-based medicine. Reactions and effects may vary significantly due to for example genetic predispositions causing varying amounts of endocannabinoid receptors. Moreover, a complex multi-molecule mentality may be more suitable for understanding the cannabinoid compound class than a small molecule mentality, due to the limited number of over 10 active compounds that work in unison to create successful treatment. For example, without some amounts of THC, cannabidiol (CBD) becomes relatively inert as an anti-inflammatory agent (Aso et al. 2015). THC and CBD can also work in synergism for other therapeutic purposes such as neuroprotection and managing neoplasms (Burstein 2015;Russo and Guy 2006). This is why traditional pharmaceuticals-mentality sometimes may struggle to comprehend cannabis-based (as well as in general plant medicine-based) effects because isolating specific compounds from medicinal plants often results in poorer results or requiring significantly higher doses.

Study limitations
Our study was designed such that manuscripts were collated from a single database only, the WoS Core Collection, so that we might have missed some relevant manuscripts in the literature. This was decided because citation numbers for each manuscript differ across different databases, and merging data from multiple databases is a complicated yet controversial challenge. We chose to focus on WoS Core Collection for our study because it is hosted by Clarivate Analytics, the same company that releases the journal impact factors and thus represents the most established authority in the area of research evaluation. This also allowed a better exploration of relationship between citation counts and journal impact factors. Finally, it should be noticed that 100 analyzed publications constitute only a very small percentage of the entire literature on the topic. Thus, the publications might not represent the complete literature but rather reflect the most influential research of this scientific area. Fig. 2 Bubble map showing words from titles and abstracts of the 100 most-cited cannabis manuscripts. Words from titles and abstracts were parsed, analyzed and visualized by VOSviewer. There were 113 terms that appeared in 5 or more manuscripts and hence were included in the map. Each bubble represents a word or phrase. The bubble size indicates its appearance frequency. The bubble color indicates the averaged citation count received by manuscripts containing the term. If two terms co-appeared in more manuscripts, the two bubbles are closer to each other. The lines indicate the 500 strongest co-appearance links between the terms Cannabis-and cannabinoids-related treatments: what is new?
More than two decades have passed until enormous changes have occurred in legal situation and social policies in regards to the use of cannabis for medical and other purposes (Corroon Jr et al. 2017). Today, there are around 30 countries worldwide that have legalized medical cannabis or cannabis-derived products for certain uses, including the United States (33 states so far plus the District of Columbia), Australia, Argentina, Canada, Chile, Israel, Mexico, Switzerland, Turkey, Uruguay, 15 European Union countries, and others. However, there are very strict guidelines regulating the medical use of cannabis-derived pharmaceuticals. It should be noticed, that some EU countries such as the Netherlands, Poland, Norway, Germany, Italy, and others are leaders in cannabis legalization so far, i.e., legalized access for the needs of patients with medical conditions. In other EU countries such as Spain, France, and Slovenia, the use of cannabis-derived drugs for some afflictions is permitted. Fig. 3 Bubble map showing keywords listed by the 100 most-cited cannabis manuscripts. Keywords added by the authors and by Web of Science (KeyWords Plus) were analyzed and visualized by VOSviewer. There were 30 keywords that appeared in 5 or more manuscripts and hence included in the map. Each bubble represents a keyword. The bubble size indicates its appearance frequency. The bubble color indicates the averaged citation count received by manuscripts containing the keyword. If two keywords co-appeared in more manuscripts, the two bubbles are closer to each other. The lines indicate the 500 strongest links between the keywords Meanwhile, Uruguay and Canada are the only countries that legalized the sale and consumption of cannabis for recreational use. In the United States, 14 states have more restrictive laws limiting THC content, while ten states and D. C. have legalized the recreational use of cannabis. In the United States, the cannabidiol (CBD) containing medicine Epidiolex was recently approved by the FDA as the first prescription drug available for the treatment of rare diseases such as Dravet syndrome (also known as severe myoclonic epilepsy of infancy, SMEI) or Lennox-Gastaut syndrome (LGS) (Rubin 2018). In addition, the FDA-approved drugs Marinol, Syndros and Cesamet represent synthetic cannabinoids, which structures are similar to that of THC (Rubin 2018). All these medicines are used to treat nausea and vomiting caused by cancer chemotherapy, whereas the first two are also used as appetizer in the treatment of patients with AIDS (Rubin 2018).

Conclusions
A bibliometric analysis was conducted to identify the 100 most cited endocannabinoid, cannabis and cannabinoid-related manuscripts. The top-ranked manuscripts among the 100 concerned analgesia, weight loss, long-term potentiation, depolarization-induced suppression of inhibition, and other topics. The CB1 receptor was studied by more of the top 100 papers in comparison to the CB2 receptor. The most frequently mentioned chemicals in these publications were 2-arachidonoylglycerol, tetrahydrocannabinol, and anandamide. Together, these manuscripts comprise the most highly cited literature body on the topic, literally the molecular neuroscience at its "high".