This retrospective, real-world study retrieved and analyzed data posted by patients and caregivers on social media concerning the treatment of advanced BC. Data posted between January 1, 2015, and March 4, 2021, were considered for the study. Posts on publicly available domains, written in English, and geolocalized in the United States were included. Posts from all public sites, including social media sites such as Twitter and forums such as patient association forums, were included. In contrast, posts on Facebook and Instagram were not included, since not all posts on these sites are publicly available.

Data (social media posts) were retrieved from publicly available social media sites by identifying and extracting posts, eliminating irrelevant data, and then filtering the posts to obtain only messages concerning advanced BC. The Brandwatch extractor (Cision Ltd, Chicago, IL) software was used to identify all public posts available on the web using combinations of words related to BC (the full query is available in Multimedia Appendix 1). These discussions were extracted with the associated metadata (eg, publication date or country) and anonymized. Irrelevant posts such as those from discussion forums not related to BC, those not pertaining to patients or caregivers, and those not featuring advanced BC were then eliminated by applying a three-step process.

Initially, posts from irrelevant sources such as potential advertising sites or forums related to pets and animals were removed. Then, a machine learning algorithm was applied to the data set. The algorithm recognized three different variables (lexical field, syntactic aspects of the post, and semantic style) to identify and classify patients and caregivers according to their respective vocabulary and grammar. Next, a manual review was performed to remove inconsequential posts unrelated to patient and caregiver perceptions. Finally, the messages were filtered using keywords characteristic of advanced BC (eg, stage IV BC or terminal BC). Once these relevant posts had been identified, the users or usernames associated with these posts were considered to be directly concerned with advanced BC. Thus, all messages from these users in the data set mentioning BC were retained, even if they did not mention advanced BC.

The algorithm used in this study was previously developed using a training set of 12,330 messages related to different health domains (eg, dermatology, tobacco use, and oncology). The method consists of a pipeline featuring two extreme gradient boosting [21] classifiers (one for caregivers’ experiences and one for patients’ experiences) applied successively. This method allowed identification of whether a post belonged to a patient, a caregiver, or neither. Both classifiers were based on features combining pronouns and lexical fields describing relatives and pathologies (eg, “my [pronoun] father [relative] has cancer [pathology]”). We trained the algorithm by first identifying the caregivers; this was carried out on the whole data set. To determine patients’ messages, we then reapplied the algorithm on the rest of the data set (excluding the already identified caregiver messages). Evaluation of performances yielded F1-scores (a measure of accuracy combining precision and recall) of 88.0% and 87.0% for the caregiver and patient classifier, respectively. In this work, manual review following the application of the algorithm ensured validation of the results. Prediction mistakes were corrected by the annotator.

The data sets corresponding to the patients and caregivers were then filtered using keywords associated with cancer therapy, such as “chemotherapy” and “immunotherapy.” The complete list of search terms is available in Multimedia Appendix 1. Posts containing both “chemotherapy” and “immunotherapy” were classified in both therapeutic categories.

This study used data from publicly available sources; thus, private groups or web pages were not included in our data extraction process. We did not seek permission since users automatically consent to the reuse of their data when they post on public platforms. Moreover, the study’s findings contain no identifiable information and are presented in aggregate. Names, usernames or handles, geographic locations, and any other sensitive data were not included.

Advanced BC treatments, either chemotherapy or immunotherapy, were mentioned in 299 posts; 222 mentioned chemotherapy and 77 mentioned immunotherapy (Figure 1).

The details of the posts mentioning treatments by patients and caregivers are described in Table 1. There were 80 posts by 69 patients and 142 posts by 127 caregivers mentioning chemotherapy from 38 discussion sources (Table 1). In addition, there were 42 posts by 31 patients and 35 posts by 32 caregivers mentioning immunotherapy from 13 discussion sources (Table 1). The forums and social media sites where patients and caregivers discussed specific BC treatments are described in Table 2.

Among the 222 posts mentioning chemotherapy, only 21 of 69 patients (30%) and 35 of 127 caregivers (28%) mentioned an age. In addition, 40 patients (58%) and 121 caregivers (95%) indicated a sex. Of the 40 patients with known sex, 20 (50%) were female and 20 (50%) were male, whereas of the 121 caregivers with known sex, 87 (72%) were male and 34 (28%) were female. Among the 77 posts mentioning immunotherapy, 18 of 31 patients (58%) and 30 of 32 caregivers (94%) indicated a sex, while 11 of 31 patients (35%) and 8 of 32 caregivers (25%) mentioned an age (Table 1). For internet users with a known sex, the majority were male for both patients (12/18, 67%) and caregivers (21/30, 70%).

To our knowledge, this is the first study to examine perceptions of advanced BC systemic treatments in social media posts by patients and their caregivers. Despite recent therapeutic advances for advanced BC, little is known about patient and caregiver perceptions of these therapies. Our results provide valuable insights into their perceptions. Concerning chemotherapy, we found that 71% (n=57) of patient posts expressed no perceptions. They described chemotherapy objectively, as an inevitable part of their health care journey. In contrast, caregivers were more likely to express their opinion of chemotherapy, with 44% (n=62) of their posts being negative, 8% (n=11) mixed, and 7% (n=10) positive. Only 9% (n=19) of all posts contained positive perceptions about chemotherapy, and these were mainly focused on effectiveness. Chemotherapy disadvantages were discussed in 39% (n=87) of posts and were mostly related to side effects and perceived lack of effectiveness. Conversely, patients expressed their opinions about immunotherapy, with 55% (n=23) of posts being positive and 29% (n=12) expressing no perceptions. Positive comments focused on treatment effectiveness, few side effects, and extending the patient’s life. Interestingly, caregivers were more likely to express a negative perception about immunotherapy than patients, accounting for 37% (n=13) and 10% (n=4) of the posts, respectively. Negative perceptions about immunotherapy focused on perceived lack of effectiveness, side effects, and persistent sequelae.

In our study, more patients and caregivers shared their perceptions of chemotherapy (222 posts) than immunotherapy (77 posts). This is expected since platinum-based chemotherapy has been the preferred standard first-line treatment for patients with advanced BC for a long time [3,4]. Furthermore, immunotherapy was only authorized as part of the advanced BC treatment pathway in 2016; therefore, during the first 2 years of the study, chemotherapy was the only treatment option for advanced BC. More positive perceptions were noted for immunotherapy among overall posts (patients and caregivers), possibly because these are newer treatments with favorable safety profiles and their increased use in advanced BC has received positive press, including the recent positive results reported with avelumab as the new standard of care in first-line maintenance of advanced BC [22].

Among caregiver posts, a mostly negative perception of advanced BC treatments was revealed. This negative perception may be explained by the fact that caregivers often feel poorly equipped to support patients, with limited knowledge about BC and treatments [6,23]. Furthermore, treatment side effects severely impact both patient and caregiver quality of life and can be expected to negatively influence treatment perceptions [24]. Considering the pivotal role that many caregivers assume in the lives of patients with BC and the importance of their involvement in patient care, their level of understanding should be acknowledged by clinicians and other members of the multidisciplinary care team. It is thus crucial that caregivers be informed and provided with the support required to effectively assist patients with their cancer treatments.

While patients with BC may use social media to share their experiences, there is a paucity of literature using social media data to gauge patient perceptions [25]. Overall, we found that caregivers engaged more frequently and actively on social media than patients. These results are consistent with a recent systematic review in which the authors noted that most patients with BC were older men with lower electronic literacy [25]. Therefore, it is the caregivers, on behalf of patients, who may be actively engaging on social media to obtain further information. The increased social media presence of caregivers could also be due to the severe grief or burden related to end-of-life care that they experience, with messages often posted several years after the patient’s death [26]. Interestingly, most caregivers identified in the Renner et al [26] study were women, who have been found to seek emotional support in online health communities more often than men [26,27].

This study design has several strengths. A large sample size collected over a 6-year period was analyzed. The results include data from a variety of social media sources and could provide another dimension to research on treatment perceptions. Accessing publicly available social media data is quick, inexpensive, and has no access restrictions.

However, our innovative research approach does entail several limitations. The posts extracted were limited to publicly available sites, which excluded popular social media networks such as Facebook and Instagram, meaning that many data were not included. Furthermore, relevant posts may have been inadvertently discarded during the filtering process. Duplication may have also been possible if users were active on more than one forum. Additionally, our analysis is based on the spontaneous declarations of internet users about their experience of the disease or their treatment. Although this type of data collection allows us to be representative of the population of internet users that post on social media, it is not necessarily representative of the general population.

A further limitation with using social media posts is that posts only have limited information. Critical information to place the post in context (such as the disease stage or treatment details) may be missing. This lack of data also makes it difficult to compare our results with those of traditional epidemiological studies. Furthermore, few forum users shared demographics such as age, sex, and location in the publicly accessible data that were used for this study, making it impossible to judge whether the data are representative of patients with advanced BC and their caregivers in the United States. The data quality depends on patient and caregiver electronic literacy, their experiences and perceptions, and their capacity to understand and accurately communicate BC information, including the type or stage of BC and treatment administered. Patients and caregivers do not necessarily include all details about their treatment, such as type of treatment, duration, lines of treatment, and response information. These self-reported data may be subject to recall bias. In addition, we cannot verify the authenticity of the published posts.

It is also possible that since the data came from social media, posts may be more negative [28,29]. Twitter has more anonymity than sites such as Facebook, meaning that more negative behavior could be provoked [30,31]. Since most of our data came from Twitter, this could partly explain our findings. Finally, our study is prone to selection bias, as included patient and caregiver posts may not represent all patients with BC and their caregivers. Indeed, engagement with social media depends on age and sex, ethnicity, socio-professional class, and income, as well as levels of education and technological and health literacy.

This study revealed areas that need to be addressed. Patients and caregivers indicated that they lacked information about patient experiences with advanced BC and its treatments. This is consistent with the fact that studies on social media reported that BC remains underrepresented online compared with other cancers [25,32]. There is therefore a need for clear, accurate, and accessible information about BC treatments for patients and caregivers.

Currently, chemotherapy is the recommended first-line treatment for patients with advanced BC. The negative perception of chemotherapy identified in this study needs to be investigated and considered, as it may influence the choice of treatment of patients seeking advice in social media forums. Therefore, a reflection work could be initiated in partnership with physicians who treat patients with advanced BC. This reflection work could help to identify the levers of improvement and communication to best manage the potential stress and anxiety associated with chemotherapy for patients and caregivers. Subsequently, it would be interesting to study the impact of chemotherapy perception on the adherence to treatment and the quality of life of patients and caregivers using social media [33].

This study also highlights that social media posts from patients and caregivers may provide real-world insights into treatment perceptions and quality of life, as previously shown in other studies [34]. It would also be interesting to cross-reference this or a future study applying our methodology with other qualitative studies on patients with advanced BC to compare the different signals and analyze their potential complementarity [16]. The extension of our research method to other countries or regions may be also valuable to identify initiatives that could improve treatment perceptions, quality of care, and quality of life for patients with BC and their caregivers.

Real-life data from social media posts may generate further insights into the impact of BC treatments on patients and caregivers not captured in standardized clinical study questionnaires. In advanced BC, chemotherapy remains the cornerstone of first-line therapy. Despite this, there appear to be some negative perceptions of chemotherapy among patients with advanced BC and more so among their caregivers. Addressing these negative perceptions of treatment may improve treatment adoption. Additional support and information could be offered to patients and their caregivers on BC therapy and how to manage side effects. This may allow them to have a more positive experience, which has increased importance given the survival benefits associated with first-line platinum-based chemotherapy followed by avelumab maintenance in those whose disease has not progressed on chemotherapy.