Does neuroscience research change behaviour? A scoping review and case study in obesity neuroscience

The language employed by researchers to define and discuss diseases can itself be a determinant of health. Despite this, the framing of diseases in medical research literature is largely unexplored. This scoping review examines a prevalent medical issue with social determinants influenced by the framing of its pathogenesis: obesity. Specifically, we compare the currently dominant framing of obesity as an addiction to food with the emerging frame of obesity developing from neuroinflammation. We triangulate both corpus linguistic and bibliometric analysis of the top 200 most engaging neuroscience journal articles discussing obesity that were published open access in the past 10 years. The constructed Neurobesity Corpus is available for public use. The scoping review analysis confirmed that neuroinflammation is an emerging way for obesity to be framed in medical research. Importantly, the articles analysed that discussed neuroinflammation were less likely to use crisis terminology, such as referring to an obesity “ epidemic ” . We highlight a potential relationship between the adoption of addiction frames and the use of stigmatising language in medical research.


Introduction
Fatness has many interpretations and implications in the modern world.For example, corpulence can be conceptualised as a medical condition (obesity) correlated with chronic disease (Andolfi and Fisichella, 2018), a cause of excessive economic cost (Mayes, 2015), or as a visible invitation for stigmatisation (Nath, 2019).Human bodies across most of the world have gotten heavier in the last century (NCD Risk Factor Collaboration (NCD-RisC) 2016; OECD, 2022); chronic disease is becoming more prevalent, estimations of obesity costs are increasing, and accounts of weight stigma have expanded.Isolated intellectual stances on obesity have not successfully integrated to provide solutions to these troubling trends.Instead, research disciplines adopting different conceptualisations of obesity often disagree on proposed solutions (Campos et al., 2006).As a simple example of this, consider whether a medically obese person should or should not be offered gastric bypass surgery for weight loss.Proponents of a medical viewing of obesity would support this treatment option in an attempt to improve the patient's quality of life (Johari et al., 2020) and life expectancy (Adams et al., 2007;Carlsson et al., 2020).Proponents adopting an economic lens may suggest that the surgery is an unjust expense placed on taxpayers, given that the individual could potentially lose weight through inexpensive lifestyle changes (Mayes, 2015).Proponents of fat studies would suggest that the term "obese" unnecessarily pathologises the individual's body, which may not necessarily be unhealthy (Chrisler and Barney, 2017).They would instead advocate for a Health at Every Size approach to the person's medical care, where weight loss is not a primary goal of the individual's care (Tylka et al., 2014).There are diverse theoretical toolkits, disciplines and expertise underlying these diverging perspectives on corpulence.A common argument in fat studies is that medical perspectives of obesity stigmatise fat people.Recent studies have shown that fat stigma is a social determinant of health, and accounts for a proportion of the pathology that develops in obese individuals (Brown et al., 2022;Himmelstein et al., 2015).Therefore, combatting the widespread economic and social burden of chronic disease and weight stigma relies on a more coordinated response from traditionally polarised disciplines.
The neuroscience of obesity deeply intersects with both the social and physiological determinants of obesity-related illness.Recent advances in neuroscience have allowed for researchers to examine alterations in the brain resulting from the consumption of obesogenic foods, such as fat and sugar (Beecher et al., 2021(Beecher et al., , 2022;;Wang, 2023;Wang et al., 2023).Neuroscientists initially drew parallels between diet-induced neuroplasticity, and the neuroplasticity resulting from, and reinforcing, drug abuse.As a result, a new medical frame of obesity was constructed: obesity as an addiction to obesogenic foods (Foddy, 2011;Wang, 2023).Obesity as an addiction to food is a specific medical frame.The frame was first deployed in the 1950s due to the post-war popularity of psychiatry (Rasmussen, 2012(Rasmussen, , 2015)), and subsequently reinforced by the war on drugs.Of course, not all obese people are addicted to food, and some people addicted to food are not medically obese (Gearhardt et al., 2016;Meule, 2019;Saffari et al., 2022;Throsby, 2021;Yarnell-Mac Grory et al., 2021).Although not currently included in the DSM-V as an authoritative psychological disease, a quantitative food addiction scale has been developed and used in clinical practice (Saffari et al., 2022).The addiction frame has been widely accepted by the public in surveyed countries (Barry et al., 2009;Wilson et al., 2009); most recently 69% of US citizens and 74% of Australian citizens surveyed by Lee et al. (2013) agreed that obesity was caused by a food addiction.Their responses did not differ according to their Body Mass Index (BMI).In addition to its usage as a clinical and preclinical theoretical framework, the food addiction frame also has the capacity to enhance fat stigma (Cassin et al., 2019).It is therefore important to consider alternative neuroscientific framings of obesity which may provide both therapeutic discovery and promote the de-stigmatisation of obese people.
Obesity and its associated conditions (for example metabolic disorder) are correlated with inflammatory signalling in the brain (Won et al., 2009).In the early 2010s, neuroinflammatory signalling in appetite-regulating brain regions following obesogenic diets began to be hypothesised as a potential cause and therapeutic target for diet-induced obesity (Cai and Liu, 2011).This neuroinflammation model of obesity pathogenesis is still currently used to guide certain obesity neuroscience research projects (Noronha et al., 2019;Wang and Beecher, 2021).As a model for obesity pathogenesis, neuroinflammation satisfies the similar epistemological and ontological alignments to an addiction model theorisation, yet potentially reduces the chances of stigmatisation.Unlike addiction, inflammation holds acute connotations in its everyday use.Additionally, a neuroinflammation model of obesity lacks ties to drug addiction models, which have been heavily stigmatised (Cassin et al., 2019).Hence, considering a neuroinflammation perspective for understanding obesity pathogenesis could offer a more effective approach to explore both its physiological and social dimensions.To the best of our knowledge, there is a lack of studies specifically investigating the framing of a neuroinflammation model for obesity.Nevertheless, it is possible to assess the influence of the neuroinflammation framework on obesity by conducting a direct review of existing research.
The framing of obesity in media texts has been extensively characterised.Several critical discourse analyses demonstrate the hyperbolic deployment of obesity metaphors fosters a hostile view of obese people and compound weight stigma (Boero, 2007;Brookes and Baker, 2021;Coltman-Patel, 2020;Saguy and Almeling, 2008).These studies have not yet reported on the use of neuroinflammation framing, likely due to the enduring popularity of addiction framing and the comparatively short history of neuroinflammation being linked with appetite (Cai and Liu, 2011;Wang and Beecher, 2021).An important tool used in news media is the use of scholarly material as an elite source to provide credibility to news stories (Calsamiglia and Lopez-Ferraro, 2003;Sundar, 1998;van Dijk, 2008).The language deployed in academic peer-reviewed literature can therefore heavily influence the press.For example, a respected cardiovascular health research institute published a report including the phrase "Australia's Future Fat Bomb" (Holland et al., 2011); the framing of obesity as a time-bomb was then extensively reproduced in Australian media alongside exaggerated and incorrect statements regarding obesity epidemiology (Holland et al., 2011).Research articles therefore constitute an authoritative source on obesity, and the framings within these texts can be easily reproduced in the volatile climate of press media.
Theoretical frameworks are rarely explicitly declared in traditional scientific writing given the supposed objectivity of quantitative empirical work (Kivunja, 2018;Semino et al., 2018;Simandan, 2017).However, the interpretations made from obesity neuroscience data, which are often widely cited in public media (Brookes and Baker, 2021) and policy documents (Cotter et al., 2021), rely on subjectively deploying socially constructed metaphors to summarise what are largely unknown and hugely complicated neuronal phenomena (Goffman, 1974).In other words, researchers selectively research and report on salient features of the neuroscience of appetite and metabolic regulation (Entman, 1993).Globally researchers, like the press media, are assessed on the level of engagement with their texts (Crotty, 2013;Millar et al., 2020;Oravec, 2017).This workplace environment encourages the selection of frames that are likely to attract press media engagement, successful grant applications and selection for publication in high-impact journals (Benner and Holmqvist, 2023;Feldman and Sandoval, 2018;Oravec, 2019).Such a process reinforces existing, widely-accepted disease frames regardless of their utility, which may explain why frames self-perpetuate (Aronowitz, 2008).
This scoping review aims to examine the use of neuroinflammation and addiction frames in highly engaging obesity neuroscience research.These frames are gradually disseminated to the public from obesity neurobehavioural research.Therefore, in order to examine the framing of obesity pathogenesis in neuroscience research, the language deployed by the authors of this research must be analysed (McNealy, 2021;Vicari, 2010).While the linguistic analysis of framing has traditionally been conducted through qualitative methods, more recent corpus linguistic methods have been utilised, allowing for the analysis of large numbers of journalistic texts (Brookes and Baker, 2021;McEnery and Brezina, 2022;Touri and Koteyko, 2015).Corpus linguistics is a group of linguistic methods that analyse the language used in texts through statistical approaches (See Box 1 for a more thorough introduction of corpus linguistics).However, quantitative corpus linguistic analysis has rarely been used to analyse academic writing, especially in the neurosciences (Oleksandr, 2016).This review will now characterise the prevalence of addiction and neuroinflammation obesity framing in neuroscience journal articles, and assess the ability of corpus linguistics as a tool to analyse theoretical frameworks in neuroscience texts.

Study section and Neurobesity Corpus design
Some large-scale corpora of neuroscience articles have been developed (Kenkel, 2019;Müller et al., 2008).However, these corpora either do not allow for refined corpus linguistic analysis (Müller et al., 2008), or do not include the body of the articles (Jiang and Hyland, 2023;Kenkel, 2019;Lo et al., 2020).Given that the frame utilised in a scientific article may be constructed throughout the text, including the body of the article in the corpus design is important.It is therefore necessary to construct a new corpus ab initio to address the aim of this study.Given that this study examines the potential stigmatising consequences of obesity discourse, the corpus was designed to include influential obesity neuroscience research.Academic articles with a high level of engagement from other parties that are both within and outside of the academic literature were included.Two predominate metrics assess this manner of engagement: PlumX and Altmetrics.While both metrics consider a plethora of engagement data, PlumX is more sensitive to academic citations (Lindsay, 2016), whereas Altmetrics measures more evenly a wider variety of public dissemination (Ortega, 2018;Thelwall, 2021) Box 1 Definitions and explanations of key terms used in the methods of this scoping review.

(A).
Corpus linguistic methods are methods that, unlike other traditional linguistic methods, analyse a large collection of texts (termed a corpus) using quantitative and statistical measurements (Lusta et al., 2023).Corpus linguistic methods encompass strategies to quantify the frequency, location and collocation of terms within texts, and allow for tests of statistically significant differences in these quantities between groups of texts within the corpus (Brezina, 2018).Corpus linguistic methods are therefore best suited to systematic, large-scale analyses of texts.

(B).
Alternative metrics (often shorted to Altmetrics) are metrics that serve to quantify a research article's impact in a broader sense than citation counts (Chavda and Patel, 2016).These metrics can include, but are not limited to: an article's view/download count, its level of discussion on social media, its citation in Wikipedia articles or scientific blogs as well as its citation in other formal academic publications.These measurements therefore consider not just the impact an article has within the academic community (commonly measured by citation count), but also the level of engagement the article has with the public.Given that our framework for this scoping review considers the impact of obesity neuroscience research on weight stigma amongst the public, articles with high alternative metrics are important to include in our Corpus.
A number of providers exist which synthesise and calculate an overall metric based on these individual factors; two of these providers are Altmetrics.com(also often shortened to Altmetrics) and Plum Analytics (shorted to PlumX).Each provider uses different mechanisms to find instances of engagement with an article online, and each therefore have stronger and weaker data extraction across the different individual alternative metrics.Overall, we chose to use Altmetrics.com and their Altmetric score as our overall measure of article impact as this provider has been shown to better extract data from journalistic sources such as blogs and news sites when compared to PlumX (Lindsay, 2016;Ortega, 2018;Thelwall, 2021).rendering it the more suitable metric for the scoping review.
The scope of the corpus was determined by searching for articles with "obesity" in the title or keywords and published between 2012 and 2022 in a journal within the neurosciences (subject area 1109) using Altmetric explorer on the 16 th of October 2022.Many journals that publish non-neuroscientific work may also fall under subject area 1109 in this screening (for example, the journal Gastroenterology).Therefore, each article was manually screened for inclusion if the article included neuroscience concepts.A year of publication range beginning from 2012 was chosen to align with the development of the neuroinflammation model of obesity during the early 2010's.Both review and research papers were included.While acknowledging the potential for duplication in review papers, we found that considering duplication in the context of language use provided valuable insights into the framing of obesity in neuroscience research, allowing for the capture of nuanced variations and trends.To avoid copyright infringement, only open access articles were marked for inclusion in the corpus.The 200 articles remaining after these exclusions with the highest Altmetric scores were used as the basis for the corpus (see Fig. 1 for article selection strategy).The sample size was chosen according to the frequency distribution of Altmetric scores retrieved from the search (Fig. 2), given that after these first 200 articles, the Altmetric scores do not vary and are likely too small to represent meaningful engagement with either the public or research community.

Corpus construction
To synthesise the 200 included articles into a single corpus for analysis, each article's XML file had its main text copied into separate word documents.To achieve this task efficiently, Pubmed Central versions of the articles were copied as this allowed for consistency in formatting between articles.Included blocks of text were highlighted until either the end of each section (in an XML file) or page (in a PDF file if XML unavailable) to avoid metadata, or until an excluded text/image was reached.The highlighted text was then copied and pasted into a word doc, and the next included text was identified.This process was repeated, manually concatenating the copied portions and adding spaces if necessary to separate words and the beginning and end of pasted sections.Each article's included text was saved as separate word files to allow for their individual annotation in later analyses.These files were then concurrently uploaded into #lancsbox v. 6.0 (Brezina et al., 2015(Brezina et al., , 2020) ) under the corpus title "Neurobesity", alongside the 2021 Impact Factor of the publishing journal as metadata.
Decisions on which sections of text to include and exclude from the corpus were based on alignment with the research question.This study analyses obesity discourse using addiction and neuroinflammation frames.Therefore, the main body of the articles were included, beyond just the abstract as other existing research corpora have done (Harwood, 2005;Hsu, 2013;Martínez et al., 2009;Valipouri and Nassaji, 2013).Additionally, articles with non-introduction/method/results/discussion structures were not excluded as influential obesity discourse may occur in other paper genres such as reviews or editorials (Lei and Liu, 2016).
Since the aim of this scoping review is to analyse how obesity is framed in neuroscience research, both abstracts and review articles provide potentially rich locations for these framings to be constructed and critiqued.A detailed list of inclusion and exclusion criteria are available in Table 1.To verify the consistency of the corpus construction method, two of the authors followed the above guidelines in an attempt to duplicate corpus entries for 10 randomly selected research articles.Two duplicate corpora were produced of the 10 articles (replication test 1 JW and replication test 2 KB) and a few rudimentary analyses were performed to verify their identity with one another.

Keyword and collocation analyses
In corpus linguistics, a keyword refers to a term that is more frequent in a particular corpus of interest when compared to a reference corpus (Kilgarriff, 2012;Scott, 1997).The British National Corpus baby version was used as the reference corpus for this keyword analysis (Brezina et al., 2021).The simple maths parameter was calculated for each word type in the corpus (Brezina, 2018;Kilgarriff, 2012).The constant k was set to 100 to allow for analysis of more commonly occurring terms (Brezina, 2018;Kilgarriff, 2012).The top 50 keywords are reported.
To examine how obesity is discussed within the Neurobesity Corpus, collocation analysis was performed around the reference wildcard "obes*".The statistical parameters used for this collocation analysis were chosen to align with a different corpus linguistics study of obesity discourse in the British press (Brookes and Baker, 2021).The mutual information (MI) test was utilised as it is sensitive for infrequent terms (Gablasova et al., 2017).A span of 5 words left and right of the queried term, and a cut-off statistic of at least 3.0 and collocation frequency threshold of 30 instances were applied.The resulting collocants were analysed and any collocates referring specifically to an addiction or neuroinflammatory model of obesity are reported below.Any collocates that summon crisis metaphors are also reported, given that these metaphors have been shown to influence anti-fat stigma when used in press media (Saguy et al., 2014).Representative examples of important collocates were also reported as quotations to provide qualitative depth to the findings.To determine if the frame adopted influences the usage of crisis metaphors, the frequency of three key obesity crisis terms used in previous obesity discourse research ("epidemic", "war", "crisis"; Saguy and Almeling, 2008) were measured across the two pathogenic frames.

Assessing correlation between frame and crisis metaphors
Saguy and Almeling ( 2008) identified three single-word metaphors used to invoke an emotion of crisis in reporting on obesity: war, crisis and epidemic.Each of these metaphors work to reinforce weight bias (Cotter et al., 2021;Saguy et al., 2014).To determine if the usage of these metaphors differs between studies that use an addiction or neuroinflammation frame, studies were first grouped into those that include a neuroinflammation related term (neuroinflammation, neuroinflammatory, gliosis, microgliosis), the term addiction, both terms or neither term.Then, the number of articles that included each of the crisis terms were counted.A chi squared test was then performed to determine if the number of neuroinflammation/addiction framing articles that mention crisis terminology differ from their expected average proportions.

Bibliometric analyses
A bibliometric analysis was performed to complement the aforementioned corpus linguistic analysis of the scoping review.This was done predominately so that article metadata could be tracked in an effort to trace the development of ideas within the Neurobesity Corpus (Pritchard, 1969).The analysis was performed through the R package Bibliometrix (Aria and Cuccurullo, 2017), a well-validated bibliometric tool with flexible functions (Moral-Muñoz et al., 2020).Bibliometrix was also chosen as it is open source, allowing for reproduction of this study by other members of the academic community.We constructed the dataset ab initio, matching the articles used in the Neurobesity Corpus.This was performed by searching the DOIs of each article on the Web of Science (search string provided in Supplementary Data 1).One article (Anjum et al., 2018) was not found in Scopus and was unavailable to download in another form compatible with the Bibliometrix package.One preprint was also excluded from the bibliometric analysis.The remaining 198 articles were included.The results were downloaded as a BibTeX document, and the Better BibTeX update (/https://retorque.re/zotero-better-bibtex/) for Zotero was installed prior to analysis to ensure all records maintained their citation information.The analysis was performed using the Biblioshiny interface.Specifically, the following analyses were conducted using default quantitative parameters on Biblioshiny: collaboration world map, countries' production over time and trend topics.

Results
The Neurobesity Corpus consists of articles published in neurosciencerelated journals in the past 10 years with the wildcard "obes*" in the article title.The Altmetric search delivered 1822 articles, 1511 of which were flagged as open access.The closed access articles excluded had a lower Altmetric score compared to the top open access articles included in the corpus (Table 2).These articles were then further screened until 200 articles suitable for inclusion were found.Overall, 224 articles were screened fully.An additional 24 of the articles screened were excluded due to either lacking a neuroscientific focus, not actually being open access despite being labelled by Altmetric explorer as such (n=23) or being retracted (n=1).Overall, the distribution of Altmetric scores of the included articles is positively skewed, with scores ranging from 16 to 1491 (Fig. 2, Table 2).The top 200 articles not excluded with the highest Altmetric scores ultimately comprised the Neurobesity Corpus.
The Neurobesity Corpus was successfully compiled in #lancsbox, consisting of 1129016 tokens, 49968 types and 47625 lemmas.A basic keyword analysis using the 2014 BNC-baby corpus confirms that the Neurobesity Corpus focusses on neuroscience research on obesity, with the types "obesity", "neurons", "obese", "brain", "hfd" (high-fat diet) and "weight" appearing in the top 15 types with the largest keyword statistic (Supplementary Data 2).The corpus draws heavily on literature  from the global north, predominately citing literature with collaborations between western, developed nations (Fig. 3).Over the past 10 years, the USA has been progressively more cited within Neurobesity Corpus articles (Fig. 4).Obesity discourse A collocation analysis of the wildcard "obes*" within the Neurobesity Corpus generated a number of noteworthy collocates within 5 words proximity (left or right) which are summarised in Table 3.Of note is the term "epidemic", appearing to the right of obesity in nearly all of its occurrences in the corpus (Table 3).This alarmist language is indicative of medical frames of fatness, and has been critiqued as an overexaggerated moral panic when actual epidemiological data is considered (Campos et al., 2006): "Obesity is reaching epidemic proportions globally and represents a major cause of comorbidities" (Guarino et al., 2017) and "We used diet manipulations to induce obesity because the global obesity epidemic is generally believed to be caused by excessive caloric intake" (Cope et al., 2018).
Collocates denoting an addiction pathophysiology were also present (Table 3).The prevalence of these two different frames of obesity differs over time (Fig. 5).Specifically, addiction-related terms such as "addictive" and "reward" were more common from 2010 to 2016.On the other hand, the term "inflammation" peaked in 2019 for this corpus.These data suggest that a neuroinflammation frame of obesity is more novel in obesity discourse, and that addiction frames are the traditional norm.To assess the potential relationship between the type of framing adopted and the use of crisis terminology, a Chi-square test of independence was performed (Fig. 6, Table 4).The test reached near-significance, suggesting that a greater sample size is needed to detect whether there is likely a true difference in the deployment of crisis terminology between addiction framing and neuroinflammation framing of obesity.

Discussion
Weight-based stigma perpetuates obesity and its health consequences.Language is the vector in which stigmatising ideologies can be spread.Given the cultural capital held by neuroscience researchers, the language deployed in obesity neuroscience research has the potential to shape public attitudes towards obesity.This scoping review aimed to examine the use of neuroinflammation and addiction frames in highly engaging obesity neuroscience research.The research aim was achieved by utilising a novel methodology for constructing corpora of obesity neuroscience research articles.Granular detail of the corpus construction was provided to address the complexity and heterogeneity of research article structure.The resulting Neurobesity Corpus is available online via the Open Science Framework (https://osf.io/tbxg2/?view_only=450cd3c4cc2b41f6a3f15bdc0d850295), allowing for further analyses to be conducted on obesity research in the neurosciences.The methods developed in this study may therefore be built upon to develop corpus linguistics as a meta research tool.
Considering the finding of the review itself, the analysis demonstrated that both neuroinflammation and addiction frames are extensively present in obesity neuroscience literature.Interestingly, more articles in the corpus discussed neuroinflammation than those discussing addiction (Fig. 6).To the best of the authors' knowledge, this analysis is the first empirical demonstration of the extensive prevalence of neuroinflammation across obesity neuroscience research.Additionally, while statistical significance was not reached, there is a trend demonstrating that articles adopting a neuroinflammation frame are less likely to deploy crisis metaphors in their descriptions of obesity.Such a finding may indicate that other texts discussing this research (such as news journalism or blog posts) will also be less likely to utilise terminology that can exacerbate moral panic and weight stigma (Saguy and Almeling, 2008).Therefore, this finding reinforces the hypothesis that framing obesity pathogenesis as either addiction or neuroinflammation has social consequences in addition to therapeutic consequences.
When interpreting the major findings of this study, it is important to consider the over-representation of Western researchers in this corpus (Fig. 3).The dominance of the United States, Europe and Australia in this corpus is not surprising, given the systemic exclusion of the Global South from academic research and publishing (Bojanic and Tan, 2021;Mekonnen et al., 2022).The Global South is likely to have different framings of obesity, which has been exemplified in African contexts (Fayemi, 2018).These frames are not captured by the analysis of the review.The Neurobesity Corpus therefore likely does not capture all framings of obesity in neuroscience research, and is instead designed to capture the dominant frames.
The framing of research problems in scientific articles has extensive implications.The media acts as an intermediate in the dissemination of research findings to the public.The direct accession of journal articles by journalists was not commonly reported in the early 2000s (Attfield and Dowell, 2003).However, following digitisation and the open access movement (Jurchen, 2020), a disintermediated network for knowledge dissemination has developed, and journalists more readily consult with academic literature directly (Hertzum, 2022).The discourse used in academic research matters, as journalists seeking news story ideas from academic literature are extremely time-poor and unlikely to hold the   domain-specific knowledge required to evaluate an argument's merits (Hertzum, 2022;Patterson, 2013, p. 76).Instead, they often seek the most salient features in the examined literature early in their investigation in order to establish an overarching angle for their story (Adler et al., 1998;Attfield and Dowell, 2003;Entman, 1993;Gilbert et al., 2022).Generally speaking, while journalists generate news stories through interactions with academic research (Showkat and Baumer, 2021), they are restricted to reporting on stories that align with what Galtung and Ruge coined as "news values" (Galtung and Ruge, 1965;Harcup and O'Neill, 2017).Given the accelerated neoliberal nature of academic work, researchers are also systematically rewarded for conducting and disseminating research that aligns with news values.
In addition to frames disseminated through traditional journalism, open-access neuroscience research can influence public perceptions of obesity more directly.A growing body of literature in science communication argues that academics should acknowledge their responsibility to engage more directly with the public and intervene more deeply in the research dissemination ecosystem (Eagleman, 2013;Hoffman, 2016Hoffman, , 2021;;Leyden and Menter, 2018;Peters, 2013).While this idealistic view is not a universally accepted (Fecher and Hebing, 2021;Martinez-Conde, 2016), a majority of academics hope to proactively enact new routes of research dissemination (Besley et al., 2018).Digitisation has provided many social media platforms for researchers to engage directly with the public (Brossard, 2013).The public are therefore exposed to the language of obesity research through journalism, and through directly engaging with the researchers' alternative dissemination forms.The results of this review therefore also serve as a reminder to researchers to consider the ethics of the subjective linguistic choices made when disseminating their research.
This was a novel methodology on a small sample of literature.There are therefore many limitations to consider when attempting to extrapolate the findings from this analysis.Firstly, the articles excluded from the corpus are likely also highly influential texts whose framings shape obesity discourse.Altmetrics is a crude measure of engagement both within and outside of the academy.However, there are potentially influential, highly cited obesity neuroscience articles that did not achieve significant public engagement.Despite this, such articles may still introduce or reinforce obesity pathogenesis frames.Additionally, Altmetrics can be gamed by researchers and may be artificially inflated without representing true engagement with the public (Oravec, 2019;Strielkowski and Chigisheva, 2019).An alternative approach to constructing a corpus of influential articles could be to combine both alternative metrics and crude citation data as inclusion criteria.Additionally, while acknowledging the potential for duplication in review papers, we found that considering duplication in the context of language use offered valuable insights into the framing of obesity in neuroscience research, allowing us to capture nuanced variations and trends.Our research considers the prevalence of particular frames, rather than any particular knowledge trajectory, therefore duplicating ideas are still meaningful from this analysis as it still demonstrates the cumulative spread of a particular obesity frame.
The decision to configure the corpus around Altmetrics created additional limitations.Altmetric explorer only allows for simple search terms and only searches article titles.Therefore, obesity neuroscience articles that do not include the term 'obesity' were missed in the screening process chosen.This may occur when obesity is only mentioned in the abstract of the article and not the title.Alternatively, the medicalisation of obesity is criticised by some scholars that are concerned with weight stigma (Pausé, 2021).In the authors' experience, these scholars generally lie outside of neuroscience disciplines.However, the screening procedure used would not detect articles using alternative, non-medical terms for obesity.Articles missed due to this situation would potentially have unique linguistic characteristics that were not detected in the existing study.
Most existing meta research corpora, including in the neurosciences, only consider the title and/or abstract of the paper, excluding the body of the paper from analysis (Jiang and Hyland, 2023;Kenkel, 2019).The use of most of the papers' text in this article presents unique copyright issues that must be navigated.As a result, only open access articles were examined.In theory, this also aligns with the scoping review research question as open access articles are reported to receive more engagement (Araujo et al., 2021;Gargouri et al., 2010;Haneef et al., 2017).However, this correlation is not strong within the Neurobesity Corpus, and many highly influential articles are published as closed access (Table 2).Given that this is one of few studies that has used corpus linguistic methods to examine peer-reviewed literature, the copyright implications and boundaries of analysing closed-access article full-texts have not yet been thoroughly analysed by appropriate experts.Ultimately, the authors' considered this a necessary and unavoidable limitation in order to include full-text data in the Neurobesity Corpus while providing full records of our analysis to ensure reproducibility and further analyses by other researchers in the spirit of open science.The progressive appreciation of corpus linguistics as a meta research tool may therefore constitute an additional argument for the open access movement (Else, 2021).
Beyond the construction of the corpus itself, limitations exist within the data analysis strategies chosen.Firstly, frame analysis can be conducted either inductively or deductively (Touri and Koteyko, 2015).Given the authors' interest in comparing addiction and neuroinflammation frames of obesity, a deductive approach was chosen for this review.While this approach allows for a more systematic analysis pipeline, additional frames not originally hypothesised may be missed.Of the 200 articles included in the corpus, 124 articles do not mention either neuroinflammation or addiction.There are therefore likely a plethora of alternative frames beyond food addiction and neuroinflammation.The subsequent trend topics bibliometric analysis demonstrates that the gut-brain axis is emerging as a widely discussed aspect of obesity pathogenesis.Obesity may therefore be increasingly framed as a disorder of the gut-brain-axis, which is supported by recent biomedical research (Bastings et al., 2023).Future studies on obesity  framing may therefore benefit from adopting an inductive approach to the dataset developed here in order to identify potential alternative neuroscientific frames more comprehensively.
There are a number of important further areas of inquiry related to how neuroinflammation framing of obesity is or is not reproduced in the public sphere.While press media rely on scholarly material as authoritative sources and can reproduce the framing deployed in scholarly articles (Holland et al., 2011), the frames adopted do not always correlate.For example, in a study of press releases, Saguy and Almeling (2008) demonstrated a positive, yet loose correlation between the proportion of press releases adopting crisis terms, and the proportion of citing news articles adopting those terms.The press media is also reliant on adhering to news values (Galtung and Ruge, 1965;Harcup and O'Neill, 2017); scientific literature is only one factor guiding their writing decisions.It is therefore unknown if the differences detected between frames in the Neurobesity Corpus will be reflected in the press media.This research field would therefore benefit from similar studies analysing press releases of obesity neuroscience articles, as well as deductive examinations of the two aforementioned obesity frames in press media.
This scoping review aimed to examine the use of neuroinflammation and addiction frames in highly engaging obesity neuroscience research.The resulting analysis provides evidence that neuroinflammation is an increasingly discussed paradigm in relation to obesity, reflecting its contemporary preclinical importance.Additionally, the results likely demonstrate that a neuroinflammation obesity frame relies less on crisis framing which may influence public perceptions towards obesity.This preliminary evidence supports the hypothesis that neuroinflammation and addiction frames have differential impacts on weight stigma.Ultimately, behavioural neuroscience research itself can influence our behaviour by shifting public perceptions of diseases and impacting social determinants of health.

Research funding
None declared.

2-4
Objectives 4 Provide an explicit statement of the questions and objectives being addressed with reference to their key elements (e.g., population or participants, concepts, and context) or other relevant key elements used to conceptualize the review questions and/or objectives.

Protocol and registration 5
Indicate whether a review protocol exists; state if and where it can be accessed (e.g., a Web address); and if available, provide registration information, including the registration number.

5
Eligibility criteria 6 Specify characteristics of the sources of evidence used as eligibility criteria (e.g., years considered, language, and publication status), and provide a rationale.

5-7
Information sources 7 Describe all information sources in the search (e.g., databases with dates of coverage and contact with authors to identify additional sources), as well as the date the most recent search was executed.

5-7
Search 8 Present the full electronic search strategy for at least 1 database, including any limits used, such that it could be repeated.

6
Selection of sources of evidence 9 State the process for selecting sources of evidence (i.e., screening and eligibility) included in the scoping review.7 Data charting process 10 Describe the methods of charting data from the included sources of evidence (e.g., calibrated forms or forms that have been tested by the team before their use, and whether data charting was done independently or in duplicate) and any processes for obtaining and confirming data from investigators.

6-9
Data items 11 List and define all variables for which data were sought and any assumptions and simplifications made.5-9 Critical appraisal of individual sources of evidence 12 If done, provide a rationale for conducting a critical appraisal of included sources of evidence; describe the methods used and how this information was used in any data synthesis (if appropriate).
N/A Synthesis of results 13 Describe the methods of handling and summarizing the data that were charted.8-9 RESULTS Selection of sources of evidence 14 Give numbers of sources of evidence screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally using a flow diagram.

6, 10
Characteristics of sources of evidence 15 For each source of evidence, present characteristics for which data were charted and provide the citations.

10-12
Critical appraisal within sources of evidence 16 If done, present data on critical appraisal of included sources of evidence (see item 12).N/A

Results of individual sources of evidence 17
For each included source of evidence, present the relevant data that were charted that relate to the review questions and objectives.

10-14
Synthesis of results 18 Summarize and/or present the charting results as they relate to the review questions and objectives.10-14 DISCUSSION (continued on next page) J. Wang et al.

Fig. 1 .
Fig. 1.Article selection strategy for the Neurobesity Corpus.Figure based on the design of the templates used in Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021).

Fig. 2 .
Fig. 2. Distribution of Altmetric scores of open access articles (n=1151) returned prior to screening for additional inclusion criteria.

Fig. 3 .
Fig. 3. Distribution of Neurobesity Corpus article author collaborations between countries produced using the Bibliometrix collaboration package in the Biblioshiny interface with default presentation parameters.The number of author collaborations in each country is signified by the country's shading with dark blue indicating the most and light blue indicating the least.The red lines between countries indicate collaborations with co-authorship.Thicker lines between countries indicate a greater total number of collaborations.Raw data used to generate this Figure is available through the Open Science Framework (https://osf.io/tbxg2/?view_only=4 50cd3c4cc2b41f6a3f15bdc0d850295).

Fig. 4 .
Fig. 4. Longitudinal comparison of corresponding author nation in the references cited by the Neurobesity Corpus articles.

Fig. 5 .
Fig. 5. Frequency of terms used in keywords and titles of Neurobesity Corpus articles between 2012 and 2022.Figure produced using the "trend topics" workflow in Biblioshiny.

Fig. 6 .
Fig. 6.Number of articles in Neurobesity Corpus adopting an addiction and/or neuroinflammation frame.
, Data curation, Formal Analysis, Investigation, Methodology, Writingoriginal draft, Writingreview & editing; FC: Supervision, Writingreview & editing; HM: Supervision, Writingreview & editing; KB: Data curation, Supervision, Writingreview & editing.for the review in the context of what is already known.Explain why the review questions/ objectives lend themselves to a scoping review approach.

Table 1
Inclusion and exclusion criteria for articles and their associated text in Neurobesity Corpus.

Table 2
Overview of articles returned from Altmetric explorer query.

Table 3
Collocation analysis of the wildcard "obes*" in the Neurobesity Corpus.

Table 4
(Fisher, 1922)isis terminology in Neurobesity Corpus articles depending on adoption of addiction and/or neuroinflammation frame.A Chi-squared test of independence(Fisher, 1922)indicated a near-significant relationship between frame choice and crisis terminology presence (X 2 =16.53301, df=9, p=0.056551).