When the authors McCombs and Shaw introduced the theory of agenda-setting, they explained how the media, through a process of selection and framing of the news, affect the issues in the public mind, but also, how they think about them (McCombs & Shaw, 1972). Rogers and Dearing (1988) consider that agenda-setting is the outcome of the mutual influence of the media agenda, the public agenda, and the political agenda. Each of these agendas constitutes issues and concerns specific to its stakeholders. Within this system, the traditional media are relatively few in number; their programs reach a large audience which significantly influences the public as well as the political agenda (Rogers and Dearing, 1988). Nonetheless, the media landscape is very different in today's digital environment. McCombs has admitted that the internet has created a new configuration within the agenda-setting model (2005, p. 544). At the time of this author's observation, social media did not have yet the importance they have currently acquired. The age of social media has thus decentralized communication stemming from traditional media (broadcasting) by putting forward the information shared by a network (Zimmerli & Pelletier, 2021).

According to Albalawi & Sixsmith (2015), the social media agenda constitutes a new entity to be taken into account in establishing the agenda-setting. Thereby, the agenda-setting results from a process of influence between the media agenda, the political agenda, the public agenda as well as the social media agenda. Indeed, the interactions occurring between these four spheres can affect each other. For example, several studies have demonstrated the mutual influence between the media agenda and the social media agenda (Bekkers et al., 2011; Sayre et al., 2010). Extending this notion, Albalawi & Sixsmith (2015) argue that the social media agenda also has set its own organization and agenda. The latter is built through almost continuous social interactions between the users of these platforms (Albalawi & Sixsmith, 2015).

Regarding health, the individual agenda is set by the personal interests of the patients, their close members, or by ordinary citizens, who share and create health content on their social media. By diffusing publicly their experiences, questions, or preoccupations on health, individuals, participate, voluntarily or not, in the establishment of the social media agenda, as well as the more general agenda of public debate (Albalawi & Sixsmith, 2015; Feezell, 2017; Han et al., 2019; Sugawara et al., 2012). Literature has already pointed out the importance of patients that rely on social media in order to create communities and support networks with patients affected by the same disease (Broca & Koster, 2011; Gupta & Schapira, 2018). However, the existence of communities of patients is not the same for all medical conditions. While researchers have indicated that even if certain patients - particularly those affected by cancer - have received a significant amount of online support, this is not always the case for any kind of disease or any type of cancer (Gage-Bouchard et al., 2017; Gage & Panagakis, 2012; Loeb et al., 2017; Loeb et al., 2018; Slater et al., 2008; Sugawara et al., 2012). As to the organizational aspect of the agenda, health institutions, associations, and organizations, employ social media to conduct prevention campaigns, fundraising, communicate about an institution or, more broadly to set the public agenda without resorting to traditional media, field journalism (Albalawi & Sixsmith, 2016); this also occurs on the topic of cancer (Cao et al., 2017; Plunkett & Ryan, 2018; Neiger et al., 2013; Nguyen et al., 2013). The public agenda reflects the public interest deriving from the mutual influences of individual and organizational agendas.

For this reason, the most discussed types of cancers on social media depend on the activity and interests of various users of the network. For instance, if a community of patients engages more actively on a specific type of cancer, the disseminated content will influence the organizational agenda as well as the public agenda of social media. The interactions between these spheres will also impact the three key components of the agenda setting: the media agenda, the public agenda, and the policy agenda (see Figure 1). To put it differently, if certain types of cancer appear much more frequently than others in the social media agenda, then - through the process of mutual influence between the different agendas - this differentiated treatment will be also present in the establishment of the media agenda, the policy agenda as well as the public agenda. Jensen et al. (2014) have already indicated the influence of the media agenda on the public agenda. In particular, by comparing the perceptions of individuals with the media coverage, they revealed that distortions of perception of the incidences of cancer in relation to real incidences coincided with the media distortions. Individuals' perceptions of the risk and frequency of a type of cancer can be explained by what Fischhoff et al. (1993) refer to as accessibility bias: “a tendency to estimate the frequency of events by the ease with which they are remembered” (p.187). Therefore, the crucial role of social media in the processes of influence between different agendas leads us to enhance the studies on traditional media by focusing this time on social media coverage.

In an attempt to grasp the visibility of the different types of cancer in the public media space, research studies have mostly been conducted on written press (Freimuth et al., 2006; Greenberg et al. 1979; Jensen et al., 2014; Jensen et al., 2010; Stryker et al., 2007), occasionally on newspapers addressed to specific audiences (Cohen et al., 2008; Stryker et al., 2007), or even on a corpus of multiple traditional media (Slater et al., 2008). A few years later, Hurley et al. (2014) examined the different types of cancer mentioned on the web by centralizing his approach on online information websites (Google News, Yahoo! News, CNN, etc.). Even if the main field of study for all these researches has been conducted in the United States, some studies carried out in China (Cai et al. 2008; Peng & Tang, 2010), in Japan (Miyawaki et al., 2017), or Canada (Hoffman-Goetz & Freidman, 2005) have yielded results with many similarities. Indeed, these authors note that the most publicized cancers concern breast, lung, female and male reproductive systems. To a lesser extent, colon cancers and leukemias also appear in the most publicized cancers (Table 1).

Despite the studies carried out in North America on English-speaking corpora and those from the Asian continent, there is a substantial lack of current research on the French-speaking side on this subject. Besides, the importance of the uses of social media in relation to health and their impact on the establishment of the public agenda leads us to extend this research on the different types of cancer initially carried out on the so-called traditional media by applying them to a French-speaking corpus drawn from Facebook and Instagram. The objective is to determine the French-speaking coverage of the types of cancer mentioned on these social media. For this reason, this study asks the following research question:

RQ1: With what frequencies different cancer types are mentioned on Facebook and Instagram?

Among the studies that have analyzed the media coverage of different types of cancer, some of them have delved deeper into the analysis by comparing this media coverage with the incidence rate of these types of cancers; all of them recognize the existence of distortions (Freimuth et al., 2006; Greenberg et al., 1979; Jensen et al., 2010; Miyawaki et al., 2017; Slater et al., 2008). Indeed, the results of these studies point out the phenomena of over- or under-representation - respectively called "media amplification" or "media attenuation" in reference to Combs and Slovic (1979 - of certain types of cancer in media compared to their incidence rate (Jensen et al., 2014). Also, researchers have indicated that media coverage of cancer types reflects more closely incidence rates than death rates (Cohen et al., 2006; Slater et al., 2008)

From 1979 to 2010, studies conducted on US media outlets have strived to indicate the over-representation of breast cancer and leukemia, but also the under-representation of cancers of the male reproductive system as well as lymphomas. This news distortion (Jensen et al., 2014) of media coverage has evolved over time, with certain types of cancer moving from over- to under-represented, such as cancers of the female reproductive system. While coverage of cancer types has been little analyzed online so far, to our knowledge, it has never been studied under the prism of social media. Indeed, studies associating cancer with social media have generally focused on content generated on a particular type of cancer or communities created around specific hashtags, often on Twitter (see, e.g., Feliciano et al., 2020; Himelboim & Han, 2014; Loeb et al., 2018; O’Hanlon, 2019; Sugawara et al., 2012). This study focuses on the frequency of the different types of cancer in correlation with the most popular social media of today, Facebook and Instagram. Indeed, Facebook is the most widely used social network in the world, with 2,740 million active users. According to a survey of Statista, Instagram is deemed the 4th most popular social network in the world (2021). The results are similar in a European context, as Drahošova & Balco (2017) stated, Facebook, used by 89.20% of Europeans, and Instagram, used by 48.60% of Europeans, are among the three most consumed social media in Europe.

Considering that social media have an impact on the establishment of the public agenda, it seems essential to delve into this theme. This study, therefore aims, to answer the following questions:

RQ2: Are the most mentioned types of cancer on Facebook and Instagram also the most common in the population? If distortions are detected between social media and reality, what types of cancers are involved and how significant this gap with reality is?

Aiming to answer these research questions, using the Mention web listening tool, a corpus of 3,288 French-language Facebook and Instagram posts was collected. Data collection took place over 2 periods of 4 days: from 4 to 7 June 2020 and from 10 to 13 August 2020. Two distinct periods were randomly selected from the available results thus limiting the bias inherent in the choice of a single period. The number of days was chosen so as to obtain a substantial corpus for the analysis. Furthermore, the main days of prevention of various cancers have consciously been avoided. Finally, the collection of publications via the Mention tool on these two social media can be merely achieved on public pages on Facebook, and Business accounts on Instagram: for reasons of protection of the private sphere. This research, therefore, does not include posts created from private accounts.

The web listening tool was configured by a Boolean query developed from a rigorous selection of keywords related to cancer. The wide variety of use of the word "cancer" on social media denotes minimizing as much as possible the "pollution", for example by removing anything within the purview of astrology. A pre-test carried out on the first body of entries (N = 250) helped refine the initial Boolean query.

Thereafter, with the aim of comparing the types of cancers mentioned in French-speaking publications on social media with the actual incidence of different cancers, the corpus was sorted manually; only publications from French-speaking countries in the West, i.e., France, French-speaking Switzerland, Belgium, Luxembourg, and the region of Quebec in Canada, have been retained. They have emerged from our corpus as the main countries publishing French-language content about cancer. Publications from French-speaking countries in Africa have been withdrawn from the corpus because these publications were less numerous compared to those from Western geographic areas, but also because the varying incidence of cancer types was significantly different from that of the countries retained for our analysis. The geographical origin of the publication was sourced from information diffused on the Facebook page or the Instagram account of the transmitter of the message.

Data cleansing was performed on the entire corpus, in order to remove duplicates on the same social media; publications that were no longer present at the time of the analysis, or even cancers that occurred in an animal. In addition, only the content that was directly visible on social media was analyzed. The final corpus is composed of 2,446 publications (Table 2).

As this study deals with the discussion of cancer on social media, we chose to select the two platforms most used by our study population and to process them together.

Categories have been constructed on the basis of the 36 types of cancer identified via a reference site in the fight against cancer (Ligue Contre le Cancer). From this point, a series of keywords has been identified for each type of cancer. The categories listed were submitted and corrected by an expert in the field, the head of the oncology department of the Geneva University Hospitals (Gage-Bouchard et al., 2018; Tapi Nzali et al., 2017). The entire corpus was then sorted through content analysis in order to identify the types of cancer mentioned in each post.

If a cancer type was mentioned even once, the post was then classified in the corresponding category; even if the entry of this type appeared merely via the logo of the owner’s Facebook page/Instagram account, via a hashtag, or even if the entry of a particular type of cancer was evident through an image or a video. Furthermore, if more than one type of cancer was mentioned, then the publication could be classified into more than one category. Whereas, if there was no precise indication of cancer type, but the word "cancer" in general was merely mentioned, the post was classified under the category "General cancer". The 2,446 publications of the corpus generated 2,680 categorizations.

As the two researchers of this study conducted the post classification, Krippendorff's Alpha reliability test was carried out on a sample of 20% of the total publications (507 entries). This coefficient allows measuring the agreement between two reviewers during a content analysis (Krippendorff, 2011). Hence, an alpha index (α) was calculated for each type of cancer in order to test and validate the conducted categorizations. An index equal to 1 signifies a degree of perfect agreement between researchers (Krippendorff, 2011).

The RQ1 aims to find out how often different types of cancer are mentioned on Facebook and Instagram. Table 3 shows the classification of the different types of cancer identified according to their frequency of occurrence in the studied corpus.

The six most mentioned types of cancer on Facebook and Instagram are breast cancer (18.3%), pediatric cancer (9.7%), leukemia (4.2%), skin cancer (3.9%), brain cancer (2.8%), and lung cancer (2.0%). All other cancer types were categorized in less than 2% of the publications, representing a categorization number equal to or less than 54 out of 2,679. Breast and pediatric cancer, as well as leukemia, represent nearly one-third (32.2%) of all categorizations.

Table 3 shows that the category "Other" includes 12 types of cancer with a very low number of entries (less than 9), indicating that these are very little discussed in the internet public sphere (Papacharissi, 2002).

RQ2 aims to detect interreality distortions between social media and reality, but also to quantify them. In order to achieve this, the types of cancers listed according to their frequency of appearance on social media (cancer coverage in social media) were compared to the classification of their incidence (as a reminder, the number of new cases) in 2020, in the countries selected for the study. The incidence, prevalence, and mortality rates stem from the database of the International Agency for Research on Cancer (2021) which allows a targeted comparison of the posts of Facebook and Instagram corpus within the geographical zone under study.

The analysis of Table 4 reveals that ten types of cancer benefit from an adequate representation between the frequency with which they are mentioned on social media and their actual incidence in the same year: breast cancer (+2), liver cancer (1), corpus uteri cancer (0), lymphoma (-1), esophageal cancer (-1), testis (- 1), gallbladder (-1), lungs (-2), colon (-2) and stomach (-2).

However, it is also possible to observe two types of distortions between reality and social media. The first type, over-representation, concerns brain cancer (+11), cervix uteri (+11), leukemia (+8), ovarian cancer (+8) as well as penis cancer (+5). The second type of distortion, the under-representation is observed in the case of kidney cancer (-17), bladder cancer (-16), larynx cancer (-7), prostate cancer (-7), myeloma (-6), thyroid cancer (-6), mesothelioma (-4), pancreatic cancer (-4) and skin cancer (-3).

These cases of distortions of reality do not display all the same degree of severity. Indeed, Figure 2 shows the greater or lesser importance of the distortion: a difference score was calculated between cancer coverage in social media and their actual incidence (interreality distortion). The cases of over-representation are indicated in yellow (difference score of +3 and more), while in red are portrayed the cases of under-representation (difference score of -3 or less). The intermediate cancers, illustrated in green and placed in the middle of the scale of values (difference score equal to or between +2 and -2), indicate an absence of distortion.

Kidney cancer and bladder cancer present alarming levels of interreality distortion, assuming that under-representation is more dangerous to individuals than over-representation. However, as we have seen previously, over-representation can also be problematic. Brain cancer and cervical cancer are the most affected by this over-exposure on social media. Explanatory paths are proposed below to clarify the existence of these distortions.

Our results indicate that breast cancer is the most-talked-about type of cancer on Facebook and Instagram (18.3% and 1st in the ranking); in the continuity of previous works (see, e.g., Cai et al., 2008; Jensen et al., 2010; Slater et al., 2008) which, even though performed mainly on newspapers, place this cancer at the top of the cancer types mentioned. In this analysis, this cancer has an adequate representation in relation to its incidence rather than an over-representation as it has been identified within the press. It is also noteworthy that breast cancer is the most diagnosed cancer in women. In fact, nearly one in four (24.2%) cases of cancer detected in women worldwide are affected particularly by this type of cancer (International Agency for Research on Cancer, 2021). In addition, numerous studies have highlighted the existence of a large community of patients formed by advocacy or mobilization groups around this cancer (Slater et al., 2008). Still very active today, the strong online engagement of these communities contributes to the visibility of this cancer (Loeb et al., 2017; O’Hanlon, 2019; Seale et al., 2006).

The strong coverage of breast cancer on social media joins that observed more generally for female cancers. Indeed, our results highlight a high presence of cancers of the female reproductive system (cancer of the cervix, ovary, and uterus) as well as a major over-representation for cancers of the cervix uterus (+11) and ovary (+8). Conversely, prostate cancer is little and even under-represented (-7) on the social media under study. This result coincides with Vagra et al. (2018), who demonstrated an over-representation of female cancers compared to male cancers. It also further underlines a distortion from reality, since over the course of a lifetime, one in two men is at risk of developing cancer, whereas this is the case for one in three women (Frost et al., 2014). This distortion can be explained by two main factors. First of all, we must bear in mind the greater use of social media by women than by men (Bender et al., 2021; Kimbrough et al., 2013) as well as the over-representation of women in online participation in support groups (Nikoloudakis et al., 2018). As mentioned previously in the case of breast cancer, this distortion can also be explained by the mobilization of associations and institutions in charge of health against female cancers (Slater et al., 2008). This prevalence of attention to female cancers rather than male cancers is long-standing, since historians have shown that from the onset of this disease, it is perceived as mainly affecting women. Despite the current evidence of high incidence rates of cancer among men, female cancers retain a major place on the public health agenda. The evolution of treatments and screening over time for gynecological cancers (Hurley et al., 2014; Moscucci, 2016), but also the latest advances in terms of HPV vaccination help explain the over-representation of cervical cancers observed (Wenjia et al., 2020).

Our results show the need to communicate more about prostate cancer (-7). The communication efforts of advocacy groups and associations around this cancer are recent; this observation is consistent with that of Loeb et al. (2017), who report a considerable delay in the promotion of awareness surrounding this male cancer. After studying the two months of the year devoted to breast cancer and then to prostate cancer on Twitter, these authors noticed a significant difference both in the amount of messages exchanged over these two periods and in the type of transmitters: mainly patients in the first case, and doctors or influencers specializing in the issue, in the second (Loeb et al., 2017). Certain societal factors, such as social stigma, can influence this lack of digital voice among patients with prostate cancer and, more generally, male cancers.

This study also shed light on the significant place pediatric cancer occupied on social media. It constitutes the second most common type of cancer in our corpus. Additionally, it is important to underline that two of the over-represented cancers in our corpus, brain cancer (+11) and leukemia (+8), are the two types of cancer that affect mostly children (Ligue Contre le Cancer, 2021). Compared with previous studies usually focused on traditional media, leukemias have always appeared as over-represented cancers; stressing the existence of a strong community around this disease (Jensen et al. 2010; Sugawara et al., 2012). As for the strong presence of pediatric cancer on social media, we can state several explanatory factors. In the corpus analyzed, we noticed that opening a public page on social media allows family or relatives to share their experiences, emotional fatigue, mobilize funds or acquire online support. Gage-Bouchard et al. (2017) also observed a plethora of Facebook pages launched by hospitals or charities offering information and resources related to pediatric cancer. Moreover, glioblastoma - a subtype of brain cancer that particularly affects children - is talked about a lot due to its extremely aggressive nature, often considered the most deadly.

Another explanatory factor, concerning not only childhood cancers, is the creation of a brain tumor social media hashtag (#BTSM), induced by the successful example of the breast cancer social media hashtag (#BCSM), managed to bring currently together a large community (Feliciano et al., 2020). Brain cancer is also heavily mentioned when it comes to radiation emitted by cell phones, and even more so in current debates around 5G (Service de recherche du Parlement européen, 2019). Regarding leukemia, as the treatment period for this disease can be quite long, patients have the time to engage in a discourse about it on social media (Sugawara et al., 2012).

The fourth most-mentioned cancer (3.9% of categorizations), skin cancer, is heavily present on social media. The corpus of our study takes place over a summer period, which probably contributed to its strong evocation, despite this fact; this cancer presents a reality score of -3, signifying that it is slightly under-represented compared to its real impact. To understand the very high incidence rate of this type of cancer, it should be stated that it includes not only melanoma-type cancers but also non-melanoma-type cancers. The former are very severe but less frequent, whereas the latter are less dangerous but extremely frequent. Totaling these two forms of cancer up increases their incidence rate: 80% were non-melanoma type of cancers while the remaining 20% concerned melanoma (International Agency for Research on Cancer, 2021). Insofar as the main cause of occurrence of this cancer derives from ultraviolet radiation, strong prevention activities have been implemented especially through social media, a suitable means of transmitting messages to the younger generation. The prevention of skin cancer is also widely exploited for commercial purposes, for example, by cosmetic companies.

Our results reveal a strong under-representation for kidney cancer (-17), bladder cancer (-16), and to a lesser extent for prostate (-7). This result corresponds to that of Loeb et al. (2018) who highlighted the presence of a few tweets on Twitter concerning urological cancer. Provided that the incidence of the latter is lower than for other cancers, there is a significant gap in our analysis between the reality and the online presence of kidney and bladder cancers, also observed in the study of Sugawara et al. (2012). Worth mentioning that some cancers are less "attractive" than others when it comes to communicating with visuals (Slater et al., 2008); this phenomenon is probably further reinforced on social media (Harcup & O'Neill, 2017).

This analysis also enabled us to notice the importance of addressing cancer disease in general (46.8%) without referring to a specific type of cancer. This result is consistent with findings of Miyawaki et al. (2017) on a Japanese corpus - half of the articles dealt with the disease in general, without focusing on a particular type of cancer - and that of Cohen et al. (2008) on a corpus of US news journals where almost 30% dealt with cancer in general.