학술

Tweeting with precision: A multiscale longitudinal digital ethnography of online cancer communities

PLOS ONE
CC BY
이 매체는 공공·자유 라이선스로 본문을 직접 표시합니다.

Figures
Abstract
This study offers a digital-ethnography of cancer communities on Twitter, focusing on discussions surrounding precision oncology and biomarkers such as ROS1, ALK or HER2. Leveraging a comprehensive Twitter dataset spanning more than ten years (2011–2022) and comprising over 8M tweets from a variety of stakeholders, the research captures both the macro-level structural dynamics and micro-level interaction patterns within cancer-related online discourse. The combination of semantic analysis, user classification, and analysis of interaction patterns through retweet networks enables the identification of the roles social-media users play in disseminating information about targeted therapies. Our findings demonstrate emerging interactions between patient advocates and professionals, prior to formal recognition of innovative treatments. These dynamics are exemplified by in-depth examination of exchanges surrounding the foundation of a ROS1-positive patient group. The qualitative approach highlights how advocates and oncologists turn Twitter into a discussion space where patients find answers while care organization and clinical trial regulation policies are questioned. Precision oncology, online communities, Patient Advocacy Groups, social network analysis, text as data.
Citation: Luneau A, Akrich M, Bourret P, Cambrosio A, Cointet J-P (2026) Tweeting with precision: A multiscale longitudinal digital ethnography of online cancer communities. PLoS One 21(11): e0362866. https://doi.org/10.1371/journal.pone.0362866
Editor: Aamna AlShehhi, Khalifa University, UNITED ARAB EMIRATES
Received: March 13, 2026; Accepted: November 1, 2026; Published: November 20, 2026
Copyright: © 2026 Luneau et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Institut National du Cancer under Grant SHSESP 19-044 and the Canadian Institutes of Health Research under Grant PJT-162252. This work was supported also by TIERED (Transforming Interdisciplinary Education and Research for Evolving Democracies), a French government grant managed by the Agence Nationale de la Recherche under the France 2030 program (ANR-22-EXES-0014). “The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript”.
Competing interests: The authors report there are no competing interests to declare.
Introduction
On July 4, 2011, Alicia Staley and Jodi Schoger published the first tweet containing the ‘#BCSM’ (breast cancer social media) hashtag [1]. Inspired by the ‘#hcsm’ (health care social media) chat, ‘#BCSM’ has been used to announce a weekly Twitter live chat, but also to tag breast cancer-related tweets. Hashtags echoing the ‘#BCSM’ model have been established for various types of cancers in the following years, leading to the development of an organ-of-origin based ‘cancer tag ontology’ consisting of approximately 30 hashtags [2]. For instance, in 2013, Janet Freeman-Daily, a lung cancer patient activist and cancer survivor together with two other advocates and two oncologists confounded the hashtag ‘#LCSM’ (lung cancer social media).
Patients and activists jointly use these hashtags to leverage social media in order to build hybrid epistemic and practice communities [3], raise public awareness about their conditions, enhance their expertise, and attempt to shift the balance of power between patients and health professionals. Janet Freeman-Daily [4] stated that Twitter had been ‘the one place where all stakeholders […] could meet and exchange ideas on something that had life-and-death implications for so many.’ Our paper delves into this Twitter cancer community to map its complex structure and inner dynamics by focusing on the targeted therapies at the core of the precision oncology turn. We talk about ‘Twitter,’ because the scope of our study ends in 2022, before the platform became ‘X.’ Beyond a new name, the mutation of ‘Twitter’ into ‘X’ has had an impact on data openness, platform policies and functionalities, and its algorithm. By using ‘Twitter’ instead of ‘X,’ we want to insist on the fact that our analyses are grounded in a digital era that, as Janet Freeman-Daily said [4], no longer exists.
The cancer tag ontology embodies the fact that tumors were and still are traditionally categorized by organ of origin, as is more broadly the division of labor within oncology. In recent years, however, following the advent of genomics and molecular profiling a growing array of oncogene mutations began offering an alternative, biomarker-based way of classifying tumors, breaking down organ-based categories, with which they coexist, into several different subgroups. Indeed, some oncologists speak of distinct diseases that should be treated by using biomarker-driven ‘targeted therapies’ [5]. Given the rarity of some biomarkers, patients but also oncologists, physicians and health professionals are scattered all over the world. Social media have been seen as a means to develop an ‘evidence-based activism’ [6] at the national and international levels. Justifying why she’s sharing her ‘cancer journey on public and private social media platforms’ despite data security risks, Janet Freeman-Daily [4] says that ‘it’s one of the most accurate ways to reach other people who are living with lung cancer, especially for rarer subtypes.’
The rise of ‘targeted therapies’ in the Twitter cancer community captured in our dataset gives us the opportunity to examine the role of this medium in oncology: does the view expressed above by Janet Freeman-Daily correspond to observable reality? How do patients, activists, oncologists and other healthcare professionals interact on Twitter? Do these exchanges bring these different categories of actors closer together? What level of symmetry and reciprocity can be observed in their relationships? To what extent do these exchanges contribute to changes in healthcare practices and, more generally, in the organization and regulation of the healthcare system? How do cancer patients and their advocates contribute to discussions and dissemination of information about biomedical innovation in oncology through links they have established on social media with oncologists, health professionals, researchers, or industries?
To answer these questions, we use both a semantic and network-based analysis of its leading actors and we produce a digital ethnography [7,8] of the way innovative therapies circulate and are translated over time. This digital exploration offers four important methodological contributions that challenge classical research using digital data.
A long-term perspective
Despite a few studies, such as the one by Katz et al. [2] which spans over 6 years, most existing studies of health-related communities on Twitter adopt a relatively short-term perspective. Pellino et al. [9] offer a review of empirical research based on data that span from one day to one year. McLaughlin et al. [10] analyze the circulation of information about preexposure prophylaxis for HIV through Twitter by collecting tweets posted between November 5 and December 7, 2012. Ye Sun [11] examines the generative process underlying the conversational network structure of a tweet chat community by analyzing Twitter network data over a two-years period. Platt and Brady [12] used data obtained for a period of 473 days. Analyzing Penile Cancer content shared on Social Media Platform, Ortiz-Guerra et al. [13] collected tweets from July 1st, 2021, through June 30th, 2022.
Our focus is different as it centers on understanding the fabrics of interactions through which innovative therapies become matters of concern and put into action. We capture a longitudinal appraisal of the structure of the conversation over more than 10 years. Our extensive analysis of the most active core of the community, moreover, is unique as it allows us to depict the general structure of the community from the creation of the first hashtags in 2011 to the most recent years.
A field-based categorisation of actors
Another common characteristic of papers exploring how people on social media deal with health topics is the use of Symplur’s tools and categories. While Symplur’s categories of actors have been defined from a generic perspective [14], we craft an original ontology specifically grounded in our case study that we designed from a manual and careful examination of the diversity of actors in the most active core of the community. We identify 8 categories of actors according to the way users self-define their identity in their Twitter description. Our categorization procedure was complemented by interviews with several key actors from the cancer advocacy community.
A multiscale analysis
Our data and methods allow for the appraisal of both the macro-structure emerging from the daily interactions in-between distinct profiles and social roles in the community. They also allow us to observe micro-events pertaining to the growing presence of specific biomarkers following their discussion during, for instance, scientific conferences. Besides, our approach differs from other contributions that centered on a single kind of biomarker such as ‘#HERLow’ [15] or ‘#BPDCN’ [16] -- Blastic Plasmacytoid Dendritic Cell Neoplasm.
A mixed method research
We combine quantitative and qualitative approaches. The quantitative approach enables us to describe all the interactions between the various categories of actors, identify the themes around which these interactions develop and characterize each category’s level of attachment to the identified major themes.
However, to understand what is at stake in these exchanges, we need to adopt a different perspective and examine them in detail. The qualitative approach reveals that, despite the limited nature of Twitter messages, both oncologists and advocates have turned it into a discussion space capable of providing concrete answers to patients’ questions and developing policy proposals concerning care organization and clinical trial regulation.
Materials and methods
Data collection
We hypothesize that the nearly 30 hashtags listed in Fig 1 offer a significant perspective on how cancer-related issues are discussed on Twitter. We resorted to this ontology as a standardized, easily replicable starting point for defining the cancer community on Twitter. Leveraging the capabilities of the academic Twitter API, we used the Minet tool [17] to build a first dataset gathering all the tweets published between January 1, 2010 and December 31, 2022 which contain one or more of these 30 hashtags. We made the collect between 01/07/2022 and 10/01/2023.
We did, however, realize that the broader conversation about cancer on Twitter extends beyond the tweets containing these specific hashtags. To capture the full scope of this conversation, we implemented an account-based strategy centered on selecting the most active users across the ‘#xxSM’ communities. Fig 1 illustrates the share of tweets corresponding to each hashtag from our original query, #BCSM being the first and largest (in terms of number of tweets) community of Twitter users. Specifically, we identified users who had posted more than 200 tweets containing the relevant hashtags. This criterion led us to select 834 users, representing less than 3% for the total users (28524) who used ‘#xxSM,’ yet their tweets constitute more than 80% of the total number of tweets containing these tags. Our second and final dataset is then made of the entire tweet history of the 834 users (whether or not they contain hashtags).
We again used the Minet tool [17] to query the Twitter API and retrieve the complete set of tweets and retweets published by the 834 selected users since the inception of their accounts. This second dataset was collected from 01/03/2023 to 30/04/2023. In the subsequent sections, we will delve deeper into this dataset, which encompasses a total of 8,614,284 publications. This includes 4,389,619 tweets and 3,031,281 retweets. This approach enables us to examine interactions among different categories of users and to analyze how discussions surrounding cancer biomarkers have influenced these social dynamics.
Data processing
We enriched this dataset through three distinct methodologies. Firstly, we categorized the social roles of users based on the information provided in their Twitter descriptions, which often serve as brief biographies. Contrary to classic studies leveraging web data and computational tools, where actors suffer from a lack of sociological depth, we aim at distinguishing the role structure of Twitter users at a finer grain and observing their interactions over a long period. This is why we first assign each user an ‘identity’ according to their own Twitter description. We adopted a grounded-theory approach to develop our classification scheme. Through iterative examination of Twitter bios from a diverse sample of accounts, we inductively identified eight categories that capture the range of stakeholders active in cancer-related discussions. Four researchers then manually classified all 834 accounts based on users’ handles and self-descriptions in their Twitter bios. Ambiguous cases, such as an oncologist who is also a cancer survivor, or accounts with empty descriptions, were discussed collectively until consensus was reached. When profiles spanned multiple roles, we assigned the category most prominently featured in the user’s self-description. The result is a system of eight categories: Patients, Advocates, Oncologists, Health Professionals, Health Centers, Media, Non-Profit Organizations, and Industries. While this categorization is similar to those used in the literature [1], it includes one unique category: ‘advocates,’ which is central to the literature on contemporary forms of patient engagement [3,6,18].
The ‘patient’ category includes individual accounts that mention having cancer in their profile descriptions. ‘NPOs’ refer to collective accounts linked to research foundations, academic societies, or (patient) organizations dedicated to raising public awareness about cancer and supporting patients such as the LUNGevity Foundation which is ‘dedicated to funding lung cancer research and providing support to ALL people affected by lung cancer.’ By contrast, the ‘advocate’ category encompasses activists and groups engaged in what Rabeharisoa and Doganova [19] call a ‘war on diseases’, i.e., who advocate for active participation in biomedical research, therapeutic development, and health policy. These activists and groups prioritize fighting illnesses by highlighting ‘zones of undone science’ [20] and addressing challenges within healthcare systems, rather than solely providing mutual support or advocating for patient rights [19]. Based on their own experiences, they engage in an ‘evidence-based activism’ [6], which consists of formatting patients’ experiential knowledge and confronting it to credentialed expertise to spotlight issues that matter to patients. In our case, EGFR Resisters (Epithelial Growth Factor Receptor) is an example of an advocacy group in ‘war on disease’. It defines itself as ‘a group of EGFR+ lung cancer patients who are resistant to targeted treatment. Sharing support & research news, accelerating research, improving outcomes.’
The ‘oncologist’ category labels physicians and researchers specializing in oncology. Other physicians (surgeons, cardiologists, obstetricians,..), nurses, pharmacists and healthcare providers are grouped under the ‘health professional’ category. Lastly, the dataset contains tweets published by ‘health centers’, ‘industries’ such as pharmaceutical companies, and media focusing on news in oncology.
As Fig 2 shows, advocates make up a significant proportion of the population: they are the second largest group in terms of number of accounts, behind oncologists, who outnumber them by more than three to one.
The number of accounts corresponding to each category corresponds to the orange histogram. The number of tweets they produce is colored in green.
Proportionally, activists play a more significant role in discussions, tweeting twice as much as oncologists, who presumably use Twitter in a more passive way, primarily to keep up to date with the latest news in oncology. It is clear, therefore, that activists invest heavily in Twitter.
Secondly, we classified the content of tweets and retweets by employing a topic-modeling approach. We applied the BERTopic algorithm [21] which generates document embeddings using transformer-based language models (here we specifically used Twitter4SSE). The algorithm clusters these documents after reducing their embeddings to a lower-dimensional space. We reduced the document embeddings to five dimensions with UMAP [22] and clustered our data with HDBSCAN [23,24]. Key parameters used for the UMAP and HDBSCAN algorithms are given in Table 1, As it is classical with HDBSCAN, many data points remain unclassified. To accommodate for the high proportion of outliers, we assigned them in the BERTopic model to their closest topic in the semantic embedding and reduced their proportion to 40%. Lastly, we labeled each topic based on its most prototypical words and a qualitative analysis of its top 50 most representative tweets. Representative tweets were identified using BERTopic’s default selection mechanism, which ranks documents by their c-TF-IDF similarity to the overall topic representation.
Third, we conducted a systematic search for biomarkers and their associated drugs to identify users engaged in discussions about targeted therapy.
Within this dataset, we then specifically focused on tweets discussing precision oncology. Using regular expressions, we indexed the main biomarkers along with their associated therapies that have received FDA approval [25]. Consistent with our qualitative inquiry [18], we concentrated on biomarkers for which ‘oncogene-driven’ Patient Advocacy Groups are active, including ROS1 (ROS proto-oncogene 1), ALK (Anaplastic Lymphoma Kinase), EXON20, EGFR, KRAS, NTRK (Neurotrophic Receptor Tyrosine Kinase), BRAF, MET, RET, and HER2 (Human Epidermal growth factor Receptor 2).
Corpus selection for qualitative analysis
To conduct an in-depth analysis of Twitter exchanges about oncogenes, we drew on the results of our team’s previous qualitative work, based among other things on interviews with individuals involved in creating oncogene-focused patient advocacy groups [18]. We chose to focus on a key figure of this movement, Janet Freeman-Daily (a person who has ROS1-positive cancer). Admittedly, other advocates might have been selected who played a similar key role in the development of oncogene-driven patient advocacy groups, but an additional reason for focusing on Freeman-Daily was her role as one of the most active users of Twitter among advocates, who also happened to be closely involved in the establishment of Twitter chats (see below). We extracted from our Twitter database the exchanges in which Freeman-Daily was involved, consisting of 7,670 tweets organized into approximately 3,000 conversations. Three team members then read these exchanges and consensually selected the most relevant ones for providing a granular understanding of the content and dynamics of cancer social media discussions.
We have chosen to anonymize all of the other contributors to the analyzed discussions, with the exception of Dr Howard (Jack) West. As we shall see, this oncologist played an important role on Twitter with regard to the ROS1 oncogene that interests Janet Freeman-Daily.
Results and discussion
Discussions about biomarkers occur within a pre-existing ecosystem of interactions and information exchange that we will initially describe. Our primary focus is on the content of these tweets. We aim to uncover the nature of conversations being sustained online. Are these tweets predominantly about scientific breakthroughs and healthcare policy, or does Twitter serve as a platform for more interpersonal support exchanges? As expected, the reality is a mixture of both. Our topic model reveals a wide range of topics being discussed, encompassing cancer-related issues and beyond. We will explore this diversity and show how different types of users distinguish themselves in the topics they address.
Additionally, we examine the retweet network, which provides a unique perspective on the circulation pathways within this ecosystem. Users do not interact randomly; rather, they retweet selectively based on their respective social identity.
By jointly analyzing the content of tweets and interaction patterns, we can address a series of specific questions, such as: Do advocates discuss the results that are presented during congresses with oncologists? Are non-profit organizations (NPOs) advocating for targeted therapies? Who is raising concerns regarding access to clinical trials? This combined approach allows us to gain a deeper understanding of the dynamics at play within this online community.
What is being discussed and by whom?
The topic model initially identified 116 clusters. As mentioned above, we collectively reviewed the top 50 most representative tweets per topic to validate topic coherence and assign meaningful labels. Based on the hierarchical representation of the topics, we then regrouped those that were close to one another in the hierarchy and referred to the same issue. We obtained forty-nine groups (Fig 3), which were further consolidated into thirteen overarching categories. These thirteen categories are organized into three broad areas of discussion, focusing on cancer, health issues, and other subjects (Table 2). The labeling process is shown in the open dataset attached to the paper.
The figure represents the frequency of the 49 subtopics. The color indicates the overarching topics to which the subtopic belongs. There are 13 overarching topics gathering in three areas of discussion: cancer, health issues and other subjects.
For instance, the ‘News in Oncology’ topic encompasses tweets related to the latest advances in research or FDA decisions. It differs from the ‘Cancer Awareness’ topic, which consists of tweets aimed more at disseminating information and raising public awareness about the fight against cancer, rather than focusing on ongoing scientific research.
While these two topics are centered on cancer, other health-related topics addressed broader issues such as the COVID-19 pandemic, the healthcare system, telemedicine, or events in the academic world, including awards, thesis grants, and job postings.
Finally, the ‘Other’ categories encompass discussions on daily news and hobbies (e.g., U.S. presidential elections, sports), social issues (e.g., racial and gender inequalities), as well as more casual exchanges such as greeting messages. The Table 2 illustrates through a few examples the diversity of topics from purely social and mundane to very technical and narrow.
Tweets classified under cancer topics account for 50% of the corpus, excluding outliers. Focusing on the sub-themes, more than 5% of tweets belong to the ‘Breast cancer’ theme. Next come ‘messages of support’ (4% of tweets) and the theme ‘lung cancer awareness’. While social media appears primarily as a tool for communication and mobilization to raise awareness about cancer, it also serves as a place for ‘ordinary’ socializing. Mundane topics dominate 25% of the labelled tweets. A more ad hoc and restrictive research strategy would not have identified the importance of purely social interactions within the community. Information sharing between accounts is therefore inseparable from the establishment and maintenance of interpersonal relationships [26]. Echoing the findings of Bourret et al. [18], It is through the articulation between these two types of interaction that communities are constituted. In particular, users build an advocacy system relying on groups of mutual acquaintance and knowledge-sharing. We observe these inner structures when measuring the density of retweet connections between advocates, as visualized in Fig 5. The high within-group connectivity among advocates suggests sustained interaction patterns consistent with ongoing relationships, a finding corroborated by our qualitative observations of advocates referencing shared conference attendance, collaborative projects, and direct personal communications in their tweets.
The homophilic nature of retweet networks is not limited to advocates. Homophily refers to the tendency of individuals to associate and interact preferentially with others who share similar characteristics [27]. In our context, we operationalize homophily as the propensity for Twitter users to retweet content from accounts belonging to the same category (e.g., oncologists retweeting oncologists, advocates retweeting advocates). This homophilic behavior is observed among advocates and oncologists. They both tend to retweet mainly information from their peers within their group (Fig 5). While we observe marked homophily in retweet patterns (Fig 5), we acknowledge this is a purely descriptive finding.
However, this trend is much more prevalent among oncologists than advocates. Comparing the accounts followed by Janet Freeman-Daily, an activist concerned with the ROS1 oncogene, and Dr West, an oncologist specializing in the same area, sheds light on the origin of this disparity (Table 3). While Janet Freeman-Daily follows roughly the same number of advocates and oncologist accounts, Dr West follows more than four times as many oncologist accounts as advocate accounts.
We can therefore see that Janet Freeman-Daily has expanded her network of contacts so that the different categories of actors concerned by oncogene-driven cancers are significantly represented, which is much less the case with Dr West, who is nevertheless very supportive of advocates, encouraging them to join forces and offering to meet with them.
The emphasis placed on specific uses of the social network varies according to the type of actor involved. Oncologists, not surprisingly, exhibit a greater propensity for posting more focused messages related to advancements in oncology. Advocates predominantly focus on messages aimed at raising awareness about cancer issues, including the availability of novel diagnostic tools. This is what we observe from the analysis of the correlation matrix (Fig 4) measuring how certain topics are over/under-represented among a certain group. When comparing two columns, we get to understand the differences between the activity of oncologists and health professionals, patients and advocates, etc.
The emerging ascending clustering shows which types of actors look similar in the topics they use, and conversely, which topics are used by the same categories of actors. Positive values in red show an over-representation of a given topic among this group of users. Conversely, negative values depicted in green indicate an under-representation.
For instance, the patients are more likely to use Twitter to discuss their hobbies than cancer advocates. Conversely, cancer advocates express concerns about the relationship with oncologists and health professionals, the participation of patients in biomedical research, and other ‘oncopolicy’ issues [26].
‘Will there be a patient advocate or representative on the board? Transparency is vital and patients should have a voice at the table!’ (Advocate, 2018-05-16)
Even if cancer advocates use their accounts more professionally than patients do, they still differ from oncologists in that they pay attention to a broader scope of issues. This is consistent with the literature on Patient Advocacy Groups (PAGs) [28]. Like other PAGs, cancer advocates provide information and support toward patients for empowering them, educate health professionals, and contribute to scientific research, as illustrated by the following exchange where an oncologist is seeking information and receives an answer from an advocate (Dec 29, 2014):
[oncologist] Is there an up to date % distribution of ROS1, ALK, EGFR, KRAS, vs other (non-target)?
[advocate] Yes, but on home computer (I’m traveling), hard to find w/search. Will send Friday. ROS1 ~ 1% of NSCLC.
The effort of advocates to bridge different social spheres and address cancer-related issues is illustrated in the retweet network (Fig 5). Compared to patients, advocates tend to retweet information from a diverse range of actors, especially from the medical community. Put it differently, we can hypothesize that advocates serve as bridges between the medical community and patients.
It shows the strength of ties in and between the groups of actors. If we take the ‘advocates-NPO’ pair, for instance, the density is calculated by dividing the number of NPOs that each advocate has actually retweeted at least once by the number of possible pairs. This provides clues as to the degree of attention paid by groups on the vertical axis to groups on the horizontal axis.
The previous quantitative approach offers a general overview of the ecosystem surrounding cancer-related hashtags. It helps identify key points of interest that we will explore further using a qualitative approach. Focusing on biomarkers, we delve into tweets to observe how innovations related to precision oncology circulate on Twitter within this context of peer-to-peer discussions, information campaigns, and friendly exchanges.
Biomarkers emerging on Twitter
When considering biomarkers along with their related therapies, we observe that 2.4% of tweets refer to precision oncology. The majority of these tweets focus on the biomarkers associated with the most widely used precision oncology therapies: HER2 [29] and EGFR [30].
Between 2000 and 2022 [31,32], the United States Food and Drug Administration (US FDA) approved a total of 14 drugs targeting the human epidermal growth factor receptor 2 (HER2), with 11 of these approvals being granted post-2012. Concurrently, the US FDA approved 9 drugs targeting the epidermal growth factor receptor (EGFR), of which 5 received approval between 2012 and 2022.
The time-series graphs (Fig 6) incorporate black points when therapies received approval from the US FDA. These graphs depict the frequency of tweets referencing precision oncology over time. The annual meetings of the American Society of Clinical Oncology (ASCO), marked by dashed lines, coincide with significant increases in Twitter activity. This pattern suggests how ASCO meetings act as key events that stimulate discussions on precision oncology across social media platforms.
Vertical dashed lines are aligned with the annual ASCO conference. Black points indicate FDA approval dates for targeted therapies.
Regarding the categories of actors, oncologists authored 40% of tweets referencing biomarkers, while advocates and the media each contributed 17%: as shown in Fig 2, advocates are about four times fewer in number than oncologists, meaning they proportionally have a higher activity. However, oncologists and advocates played the most significant roles in the dissemination of information with more than half of their posts being retweets. In contrast, retweets account for just 15% of posts published by the media.
The contributions of these categories vary depending on the biomarker, over time and by community. For example, we observed an overrepresentation of HER2, one of the oldest genomic biomarkers for which there are several targeted therapies, including trastuzumab, the active ingredient in the drug Herceptin [28], which is considered a poster child of precision oncology. In contrast, we observed an underrepresentation of ROS1, a rare biomarker, in tweets published by the media.
Conversely, advocates posted relatively little about HER2 but were highly active in discussions about ROS1. Between 2016 and 2018, advocates even outpaced oncologists in the volume of tweets about ROS1. The ROS1 advocates mobilization coincided with the FDA approval of crizotinib on March 11, 2016, for the treatment of patients with metastatic Non-Small Cell Lung Cancer (NSCLC) whose tumors are ROS1-positive.’ It is also correlated with the creation in 2015 of The ROS1ders which is the first oncogene-focused patient advocacy group [18]. ROS1ders founders, such as other oncogene-focused patient advocacy groups in the ‘war on disease’ (EGFR Resisters, ALK positive, Kras Kickers), engaged in a form of activism that centers on therapeutic actionability. As patients who experienced difficulties to be diagnosed or access targeted therapies, the mission they set themselves is to improve the condition of possibility for optimal care pathway and to ‘promote the research and development of new generations of drugs for their rare cancer’ [18] to overcome drug resistance.
Our quantitative analysis of tweets has revealed several key findings. Firstly, oncologists and advocates constitute distinct groups as they tend to share tweets from people within their own category more than from people from other categories. However, this trend is much more pronounced among oncologists, whose tweets largely concern advances in oncology. In contrast, advocates are interested in a wider range of topics and retweet information from a greater variety of sources. This suggests that they play a mediating role between different stakeholders, particularly between oncologists and patients.
To give more precise meaning to the observed interactions, we will focus on one of The ROS1ders oncogene organizations, present on Twitter through one of its founders, Janet Freeman-Daily (mentioned above), and carry out a qualitative analysis of a number of exchanges. Here, we use Janet Freeman-Daily as the focal point for our qualitative inquiry, however, our analysis extends well beyond her individual tweets. Freeman-Daily’s conversations served as entry points into a broader network of exchanges involving dozens of other advocates, oncologists, patients, and healthcare professionals.
The ROS1ders as a spin-off of Twitter networking
FDA approved crizotinib for the treatment of ROS1-positive Non-Small Cell Lung Cancer (ROS1-positive NSCLC) on March 11, 2016. Note however that crizotinib for treatment of ROS1-positive NSCLC was already covered by US health insurance in 2012 on the strength of preliminary clinical trial data published in January 2012. This targeted therapy was commercialized as Xalkori® and already used for the treatment of ALK-positive patients. FDA approved another targeted therapy called entrectinib (commercial name Rozlytrek) on August 15, 2019.
Before the FDA approval, on November 14, 2011, Dr West published a first tweet mentioning that ROS1-positive NSCLC could respond to crizotinib: ‘New molecular target for #LungCancer: ROS1 rearrangements in only ~2%, but they respond well to Xalkori® (crizotinib).’ Only one advocate relayed the news, possibly because, according to Janet Freeman-Daily (personal communication), “few lung cancer patient advocates were active on Twitter before 2012. Lung cancer advocacy started to grow after crizotinib for ALK became available in 2011, and more people who had lung cancer started feeling well enough to become active in advocacy”. The oncologist published with more success a similar tweet on January 7, 2012. The news was repeated by other oncologists, NPOs, health centers or industries and the effects of crizotinib on ROS1 patients were discussed during the 2012 ASCO conference. One year later, in February 2013, OxOnc Development LP and Pfizer announced in a tweet that they agreed to conduct ‘a clinical trial of Pfizer’s crizotinib for the treatment of ROS1-positive patients in Japan, China, Taiwan and South Korea.’
At that time, The ROS1ders patient group did not yet exist. Janet Freeman-Daily was a blogger, like others affected by the same cancer. An engineer by training, she was interested in research in general, and more specifically on lung cancer, as shown by her subsequent involvement in 2015 as a research advocate in a Specialized Programme of Research Excellence on lung cancer, which was funded by the National Cancer Institute and conducted at the University of Colorado.
While most tweets in 2013 celebrated the efficacy of crizotinib, some highlighted the difficulties in accessing the drug. For instance, Janet Freeman-Daily tweeted that she learned about ROS1 and trial locations from an online forum (’I found my ROS1 trial myself, then told my oncologist’). She also mentioned having to travel every four weeks to participate in one of these trials, incurring costs that not all patients could afford. Lastly, she pointed out that the ROS1 population was ‘too small to seek FDA approval under current [2013] rules,’ which was a problem since health insurance will refuse to cover non-approved therapies. She was not the only patient who has engaged in analyzing the situation and ongoing developments.
Reading through the tweets, we observe the rise of a group of people sharing their experiences and knowledge of the disease and related targeted therapy. They talked about side effects of Xalkori, they also discussed with oncologists on the clinical trial process, the reasons why ROS1-positive patients were excluded from trials concerning ALK-positive patients or other therapeutic options.
Exchanges were not confined to the Twittersphere; they also involved in-person encounters at large annual oncology meetings such as ASCO and ESMO (respectively the American and the European Society of Medical Oncology), and more specific ones, such as those focusing on a given kind of cancer. Twitter facilitated these encounters, as advocates tweeted and retweeted messages announcing, for instance, their planned attendance to a forthcoming conference: ‘looking forward to meeting some of the #lcsm family.’ In turn, oncologists posted tweets extolling the virtue of meeting members of hashtag communities such as #lcsm: a radio-oncologist, for instance, defined #lcsm as a ‘great online community for support and advocacy!,’ and a translational oncologist tweeted that advocates performed ‘truly impressive work.’
In fact, the qualitative inquiry on oncogene-advocacy groups [18, p.7–8] reveals that the ‘foundational’ meeting of the ROS1-positive patient group took place at the LUNGevity Hope Summit in May 2015 and seems to have been initiated by a tweet posted by Freeman-Daily in March 2015: ‘Any other #ROS1 folks going to the @LUNGevity Hope Summit in May? Let’s meet up!.’ The ROS1ders’ mobilization was supported by the oncologists they had met on Twitter, but also in-person at oncology conferences. As explained by patient advocate Lisa Goldman (blog 2015), oncologists encouraged ‘ROS1ers to band together’ because the existence of a patient advocacy group would help to stimulate the interest of other researchers and Health authorities in the ROS1 mutation by gathering ‘enough cohort of ROS1 Patients.’ As mentioned in another tweet by a ROS1der advocate, this blend of online and in person activities was ‘a group effort – many ROS1ders were involved in networking with researchers, translating, supporting patients.’
After this first meeting, ROS1 advocates organized a new meetup and even participated as speakers at the American Society of Clinical Oncology in June 2015 and created a Facebook page and a private Facebook group in July 2015, events which were announced on Twitter.
A pair of ROS1 success stories at #ASCO15. Meet more pts [i.e., patients] at booth 3005’ (June 2015)
If you have CANCER with a ROS1+ mutation please join our Facebook page. any type of cancer can join. #facesoflungcancer #ROS1 (September 2015)
In short, Twitter played a crucial role in bringing together an organized group of ROS1 patients. The relationships between certain particularly dedicated patients and oncologists, documented through tweet exchanges dating back to 2013 and corroborated by retrospective interviews [18], preceded the formation of The ROS1ders and appear to have facilitated the connections that made the group possible.
ROS1ders decided to communicate with each other in a closed Facebook group rather than on Twitter. However, as we will see, this does not put an end to Twitter exchanges; advocates such as Janet Freeman-Daily maintain links between the two platforms.
Twitter networks as a resource for advocates and patients
A close inspection of the exchanges on the Twittersphere simultaneously mobilizing advocates, oncologists, and caregivers shows the different ways in which social media activism partakes in the shaping of patient outcomes. The following vignette provides an example of a direct intervention in the diagnosis and management of the disease:
A 2019 Twitter exchange begins with Janet Freeman-Daily asking about the downsides of immunotherapy. An oncologist replies that in her clinic she regularly sees two downsides: immunotherapy does not work for everyone — very disappointing for patients and caregivers who can be devastated and utterly deflated given the hype surrounding immunotherapy— and side effects can (albeit rarely) be quite serious.
Switching from a generic to a concrete question, another participant, a caregiver and advocate, mentions that her father has just been diagnosed with stage 4 lung cancer and that his oncologist wants to initiate standard chemotherapy for first line treatment and eventually (given the presence of pdl-1, a biomarker for checkpoint immunotherapy treatment) immunotherapy as second line treatment. She asks for comments.
Freeman-Daily asks whether the patient has been tested for oncogene-driving mutations such as EGFR, ROS1 or ALK. If positive to any of them first-line treatment should be targeted therapy rather than chemo; even if pdl-1 is high immunotherapy should be saved for later. The advocate asks the oncologist for confirmation.
In the meantime, the caregiver replies that her father is indeed positive for ROS1, and the oncologist attending this exchange states that in that case she recommends as first line a targeted drug (crizotinib) targeting ROS1, encouraging the patient to seek a second opinion.
Other ROS1 advocates enter the fray arguing that: crizotinib should be first line for ROS1 lung cancer stage 4 stage, before chemo and immunotherapy, even if pd-l1 is high. Given that targeted therapy is more effective and has fewer side effects, the patient should ask his oncologist to consult with a ROS1 specialist.
This exchange is particularly interesting because, as described in other media [3], it combines various types of interaction in a dense way.
- A caregiver requests advice, to which an advocate responds with a medical response that reveals a certain level of expertise, while still seeking validation from a professional.
- On the basis of medical arguments and patient quality of life, a group of advocates establish the specific response previously given as a general rule that should be respected by oncologists.
- An activist is using the open channel of communication with oncologists to draw on their experience and enhance their understanding of the side effects of certain treatments.
Twitter as a space for collective investigation
To what extent can Twitter serve as a space for developing initiatives aiming at acting directly on regulatory and policy issues such as access to new therapies or the development of new practice guidelines in accordance with precision oncology? In this respect, one of the downsides of a platform such as Twitter is the word-limit that each tweet can contain, with obvious consequences for the staging of meaningful discussions.
To work around this problem, in addition to using specific hashtags to highlight and connect relevant tweets, the organizers of social media hashtag communities such as #LCSM resorted to scheduling bimonthly one-hour events (’tweetchats’) devoted, each time, to specific topic questions posted by a chat moderator. The core moderator team of #LCSM chats consisted of four MDs and two patient advocates – one of them was Janet Freeman-Daily, whereas on the MDs was Dr West. The topics covered a wide variety of issues about lung cancer clinical research and patient care. More than 150 chats took place between July 2013 and December 2021. Participants included researchers, clinicians, healthcare providers, advocates, people living with lung cancer and their family members. As stated by the organizers, ‘perhaps the most appealing aspect of these chats is that they brought together a variety of stakeholders in the lung cancer community on equal footing and removed silos to communications’ (https://lcsmchat.com/).
What follows are a few selected excerpts from chat discussions, which provide readers with a concrete, albeit necessarily approximate, idea of their content. A 2016 chat topic, for instance, concerned the issue of enrollment in clinical trials as the evidence-based gateway to new treatments, such as targeted and immune therapies, that could offer hope of survival for patients at the end of the therapeutic trajectory. The fact that trials were highly regulated ‘due to prior shortcomings’ raised the issue of whether the actual level of regulation was appropriate or excessive. The moderator, Freeman-Daily, expressed the opinion that given the lack of wide access to compassionate use of the new therapies, she ‘would like to see ways to include more patients who are out of options.’ Another advocate concurred arguing that ‘compassionate use shouldn’t be such a hurdle […], it’s called compassionate use, right?’ Freeman-Daily replied that ‘compassionate use is tied to clinical trials, but availability is entirely at pharma’s discretion,’ and ‘many patients with brain metastases are excluded from trials because they’re too great a risk for the drug company.’ According to another moderator, Dr West, ‘it was fair to say that patients, in particular metastatic ones, should have some input in how much risk they will accept without paternalistic limits.’
The access issue came up in a subsequent chat. Off-label prescription of targeted therapies approved for other conditions represents a possible workaround when facing financial and regulatory obstacles to accessing life-extending medication, but during a previous chat (2014) Dr West argued that off-label use was not in his opinion ‘a realistic expectation for drugs that will be very pricey and have a defined population to target.’ Freeman-Daily, mentioning the successful off-label use of crizotinib for ROS1, replied that ‘off-label use shouldn’t be ruled out.’ Still, according to Dr West there was a ‘huge difference of going for well-defined population vs. a broad population and just see what happens,’ to which Freeman-Daily replied that while ALK+ clinical trials now are ALK+ exclusive, some of the drugs would work for ROS1, leading to the conclusion that ‘off-label is the only ROS1 option.’
Obviously, chats such as the aforementioned ones do not take place within a regulatory framework nor are they part of decision-making meetings but can nonetheless be seen as discursive arenas where arguments are put to a test in view of subsequent, more formal debates within clinical or regulatory settings. They inform the agenda and expectations of patient advocacy groups. Reading through the tweets, we observe the rise of a group of people sharing their experiences and knowledge of the disease and related therapies. Advocates talk about therapy’s side effects, they also discuss with oncologists about the clinical trial process, and the reasons why patients are excluded from trials concerning relevant therapeutic options. Several tweets, including those coming from advocates directly target representatives of regulatory and clinical institutions by including hashtags such as #us_fda or #ncidirector. For instance, a tweet by Freeman-Daily asked the FDA to ‘nix proposed regulations’ for laboratory developed tests that could limit patient access to therapies. In 2020, she was invited to give a presentation for the ‘Project Patient Voice’ of FDA’s Oncology Center of Excellence, an ‘online platform for patients and caregivers along with their healthcare providers.’ The presentation led to a number of tweets and retweets commenting on this event, consolidating the connection between social media and regulatory arenas.
The core team of chat moderators was aware of the reflective dimension of their initiative, which led them to devote a chat to the ‘key shortcomings” that the ‘#LCSM community’ should address to advance developments in lung cancer. Several suggestions are worth reporting. Reiterating the positive assessment of the advocates’ role and activities, an oncologist stated that he ‘would like to work on growing the community on social media.’ Freeman-Daily mentioned in this respect that ‘patient advocates need to refine their story-telling skills to be focused, concise, and bring hope in a way that attracts the media,’ while also making ‘better use of other social media tools’ such as ‘live patient-doc video chats or podcasts.’ Dr West added that ‘we want to be positive as a community and not be perceived as always feeling slighted (even if it’s fair to).’ Yet another oncologist suggested to profit from the ‘opportunities to attend relevant lung cancer conferences for insight into patient-centered outcomes.’ In 2021, however, moderators decided to suspend the Chat, mentioning the ‘waxing and waning’ of participation, technical difficulties that discouraged participants, health issues by potential participants, and a ‘growing disenchantment with social media,’ with a ‘vast majority’ of tweets generated by the chat team rather than participants (https://lcsmchat.com/). Moreover, in 2024 the demise of Twitter, replaced by X, led Freeman-Daily to leave X, citing ‘misinformation, anti-science, lack of moderation’ [4].
Conclusion
Our longitudinal digital ethnography shows how Twitter served as an influential platform for online cancer communities over a period of more than ten years. By analyzing millions of tweets categorized by themes, user identities, and interaction patterns, we documented how cancer communities actively engaged with scientific, medical, and policy developments in oncology, particularly regarding biomarker-driven targeted therapies. Oncologists, patient advocates, and media emerged as the leading disseminators of information, with advocates playing a crucial role in bridging complex scientific knowledge and patient experiences, contributing to shifts in biomedical classifications, and fostering patient empowerment through evidence-based activism.
The detailed ROS1 case study highlighted how interactions on Twitter often preceded or coincided with traditional scientific dissemination channels, notably around FDA approval events and major oncology conferences like ASCO. Such community-driven conversations were instrumental in spotlighting concerns about clinical trial design, drug accessibility, patient mobilization, and the broader socioeconomic dimensions of targeted therapies. Notably, patient advocates significantly shaped these discussions, forming organized advocacy groups online, directly interacting with medical professionals, and prompting attention toward clinical research and regulatory adjustments.
However, several limitations should be acknowledged. First, our corpus focuses on the most active users (834 accounts representing <3% of total users but >80% of tweets), capturing the engaged core of the community rather than peripheral participants. This sampling strategy may overrepresent professional and committed voices while underrepresenting casual users or patients who engage more passively.
Second, while our mixed-methods approach combines breadth with depth, the qualitative analysis centered on Janet Freeman-Daily and her network may not fully represent the diversity of advocacy styles and patient experiences within precision oncology communities.
Third, our data reflect a predominantly Anglo-American context, with discussions centered on FDA approvals, ASCO conferences, and US-based advocacy organizations. The experiences and perspectives of non-English-speaking communities or those navigating different healthcare systems are largely absent from our analysis.
Fourth, the transition from Twitter to X in 2023 and subsequent platform changes limit the reproducibility and future extension of this research. It becomes harder to get access to X’s academic API. Besides, our findings reflect a specific digital era that, as Janet Freeman-Daily noted, “no longer exists.”
References
- 1. Katz MS, Staley AC, Attai DJ. A History of #BCSM and insights for patient-centered online interaction and engagement. J Patient Cent Res Rev. 2020;7(4):304–12. pmid:33163550
- 2. Katz MS, Anderson PF, Thompson MA, Salmi L, Freeman-Daily J, Utengen A, et al. Organizing online health content: Developing hashtag collections for healthier internet-based people and communities. JCO Clin Cancer Inform. 2019;3:1–10. pmid:31251658
- 3. Akrich M. From communities of practice to epistemic communities: Health mobilizations on the internet. Sociol Res Online. 2010;15(2):116–32.
- 4. Freeman-Daily J. Why this cancer patient and advocate is walking away from X. (Hint: it’s not politics). Gray Connections [Internet]. 2024 [cited 2025 Apr 11]. Available from: https://grayconnections.net/2024/11/16/why-this-cancer-patient-and-advocate-is-walking-away-from-x-hint-its-not-politics/
- 5. André F, Rassy E, Marabelle A, Michiels S, Besse B. Forget lung, breast or prostate cancer: Why tumour naming needs to change. Nature. 2024;626(7997):26–9. pmid:38347121
- 6. Rabeharisoa V, Moreira T, Akrich M. Evidence-based activism: Patients’, users’ and activists’ groups in knowledge society. BioSocieties. 2014;9(2):111–28.
- 7. Brooker P. Computational ethnography: A view from sociology. Big Data & Society. 2022;9(1).
- 8. Marres N, Gerlitz C. Interface methods: Renegotiating relations between digital social research, STS and sociology. Sociol Rev. 2016;64(1):21–46.
- 9. Pellino G, Simillis C, Qiu S, Rasheed S, Mills S, Warren O, et al. Social media and colorectal cancer: A systematic review of available resources. PLoS ONE. 2017;12(8):e0183031.
- 10. McLaughlin ML, Hou J, Meng J, Hu CW, An Z, Park M, et al. Propagation of information about preexposure prophylaxis (PrEP) for HIV prevention through Twitter. Health Commun. 2016;31(8):998–1007. pmid:26756069
- 11. Sun Y. How conversational ties are formed in an online community: a social network analysis of a tweet chat group. Inf Commun Soc. 2020 [cited 2025 Apr 14]. Available from: https://www.tandfonline.com/doi/full/10.1080/1369118X.2019.1581242
- 12. Platt JR, Brady RR. #BCSM and #breastcancer: Contemporary cancer-specific online social media communities. Breast J. 2020;26(4):729–33. pmid:31493301
- 13. Ortiz-Guerra RA, Jaime-Casas S, Martinez-Cannon BA, Ariza-Avila JC, González-Morales AP, Bardan-Duarte A, et al. Overview and characterization of penile cancer content across social media platforms. Front Oncol. 2023;13:1301973. pmid:38169747
- 14. Carroll CL, Kaul V, Sala KA, Dangayach NS. Describing the digital footprints or “sociomes” of asthma for stakeholder groups on Twitter. ATS Sch. 2020;1(1):55–66. pmid:33870269
- 15. Kim J, Chivukula M, Feldman M, Wei S, Baskota S, Bui M, et al. Disseminating education about HER2-low breast cancer through Twitter Chats and Tweetorials. Am J Clin Pathol. 2023;160(Supplement_1):S97–8.
- 16. Pemmaraju N, Utengen A, Gupta V, Thompson MA, Lane AA. Analysis of first-year Twitter metrics of a rare disease community for blastic plasmacytoid dendritic cell neoplasm (BPDCN) on social media: #BPDCN. Curr Hematol Malig Rep. 2017;12(6):592–7. pmid:29064025
- 17.
Plique G, Breteau P, Farjas J, Théro H, Descamps J, Pellé A, et al. Minet, a webmining CLI tool & library for python [Internet]. Zenodo; 2023 [cited 2023 Oct 25]. Available from: https://zenodo.org/doi/10.5281/zenodo.4564399
- 18. Bourret P, Akrich M, Paterson F, Cambrosio A. Oncogene-driven advocacy: Collective expertise and therapeutic actionability. Health. 2025.
- 19. Rabeharisoa V, Doganova L. War on diseases: Patient organizations’ problematization and exploration of market issues. In: Healthcare activism. Oxford University Press; 2021 [cited 2025 Jan 13]. p. 55–85. Available from:
- 20. Frickel S, Gibbon S, Howard J, Kempner J, Ottinger G, Hess DJ. Undone science: Charting social movement and civil society challenges to research agenda setting. Sci Technol Hum Values. 2010;35(4):444–73.
- 21. Grootendorst M. BerTopic: Neural topic modeling with a class-based TF-IDF procedure [Internet]. arXiv. 2022 [cited 2025 Apr 11].
- 22. McInnes L, Healy J, Melville J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction [Internet]. arXiv. 2018 [cited 2026 Aug 28].
- 23. Campello RJGB, Moulavi D, Sander J. Density-based clustering based on hierarchical density estimates. In: Pei J, Tseng VS, Cao L, Motoda H, Xu G, editors. Advances in Knowledge Discovery and Data Mining. Berlin, Heidelberg: Springer Berlin Heidelberg; 2013. p. 160–72.
- 24. Suehnholz SP, Nissan MH, Zhang H, Kundra R, Nandakumar S, Lu C, et al. Quantifying the expanding landscape of clinical actionability for patients with cancer. Cancer Discov. 2024;14(1):49–65. pmid:37849038
- 25.
Boyd D, Golder S, Lotan G. Tweet, tweet, retweet: conversational aspects of retweeting on Twitter. In: 2010 43rd Hawaii International Conference on System Sciences [Internet]. Honolulu, HI: IEEE; 2010 [cited 2025 May 24]. p. 1–10. Available from: https://doi.org/10.1109/HICSS.2010.412
- 26. McPherson M, Smith-Lovin L, Cook JM. Birds of a feather: Homophily in social networks. Annu Rev Sociol. 2001;27(1):415–44.
- 27. Cambrosio A, Campbell J, Keating P, Polk JB, Aguilar-Mahecha A, Basik M. Healthcare policy by other means: Cancer clinical research as “oncopolicy”. Soc Sci Med. 2022;292:114576. pmid:34826765
- 28. Akrich M, Rabeharisoa V. On the multiplicity of lay expertise: An empirical and analytical overview of patient associations’ achievements and challenges. In: Eyal G, Medvetz T, editors. The Oxford Handbook of Expertise and Democratic Politics. Oxford University Press; 2023 [cited 2024 Jun 18]. p. 103–33. Available from:
- 29.
Bazell R. Her-2: The making of herceptin, a revolutionary treatment for breast cancer. 1st ed. New York: Random House; 1998. 214 p.
- 30. Liu X, Wang P, Zhang C, Ma Z. Epidermal growth factor receptor (EGFR): A rising star in the era of precision medicine of lung cancer. Oncotarget. 2017;8(30):50209–20. pmid:28430586
- 31. Jørgensen JT. Twenty-five years with HER2 targeted therapy. Ann Transl Med. 2024;12(3):53. pmid:38911570
- 32. Scott EC, Baines AC, Gong Y, Moore R Jr, Pamuk GE, Saber H, et al. Trends in the approval of cancer therapies by the FDA in the twenty-first century. Nat Rev Drug Discov. 2023;22(8):625–40. pmid:37344568

관련 뉴스