Online accounts of gene expression profiling in early‐stage breast cancer: Interpreting genomic testing for chemotherapy decision making

Abstract Background Genomic techniques are being developed within oncology and beginning to be experienced within routine cancer care. Little is known about how these tools feature in patients’ experiences of treatment decision making. Objective This research explores the ways in which women interpret and discuss gene expression profiling for breast cancer treatment decision making, as articulated within online accounts. Design This study used a qualitative approach to analyse written exchanges focusing on gene expression profiling in the UK (Oncotype DX test). Accounts are taken from online forums hosted by two UK cancer charity websites, comprising 132 discussion threads from a total of seven forums. Authors qualitatively analysed the data and developed key themes drawing on existing literature from medical sociology. Findings Women used online spaces to share and discuss results of gene expression profiling. Women interpreted results in the context of indirect experience of cancer treatment, and sociocultural depictions of cancer and chemotherapy. Users largely represented the test positively, emphasizing its ability to “personalize” treatment pathways, though many also pointed to inherent uncertainties with regards the possibility of cancer recurrence. Discussion and Conclusions We highlight the complex contexts in which genomic techniques are experienced, with these shaped by personal biographies, online environments and pervasive cultural narratives of cancer and its treatment. We highlight tensions between the claims of genomic testing to aid treatment decision making and patient reflections on the capability of these techniques to resolve uncertainties surrounding treatment decisions.


| INTRODUC TI ON
A key feature of contemporary medicine is the incorporation of molecular information within clinical pathways to understand and act upon disease. 1 This is particularly visible within oncology, where it is anticipated that the identification of specific genetic alterations within tumours will lead to treatment regimes tailored to individual patients. 2 As an example of this novel approach, gene expression profiling may now be offered as part of routine National Health Service (NHS) breast cancer care for a subgroup of patients. This technique is used to assist decision making around adjuvant chemotherapy, a treatment administered following surgery to reduce the likelihood of cancer returning. For some early-stage breast cancers,* the predicted benefit of chemotherapy for preventing recurrence may be unclear when assessed on protein receptor status and tumour grade alone. Available as part of NHS care as of April 2015 (initially in England), the Oncotype DX test uses gene expression profiling to predict the risk of cancer recurrence in these patients and identify those who are most likely to benefit from adjuvant chemotherapy.
Oncotype DX assesses the activity of 21 genes in breast cancer tissue. The corresponding results are prognostic, indicating the likelihood of a woman's cancer returning within 10 years when treated with hormone therapy alone. The cancer is assigned a continuous "recurrence score" (from 0 to 100), and a risk category for recurrence: low (<18), intermediate (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) or high (≥31). 3 The recurrence score is a predictor of benefit from the addition of chemotherapy to hormone therapy for disease-free survival. 4 In clinical practice, both the score and risk categorization are used by clinicians and patients to assist chemotherapy decisions. For those positioned at low risk of recurrence, studies have suggested that these patients are unlikely to derive great benefit from adjuvant treatment. 5 Patients in this category are not recommended to proceed with chemotherapy, 6 which is itself associated with (sometimes severe) side-effects and suffering. 7 Chemotherapy is recommended for those patients with a high recurrence score, as it has been shown to bestow significant advantage for disease-free survival compared with hormone therapy alone. 4 For those placed in the intermediate category, recommendations for chemotherapy are less clear (though a recent clinical trial has indicated that women with an intermediate score may be spared chemotherapy 8 ). In the case of an intermediate score, treatment recommendations often involve further discussions with the patient, alongside consideration of wider clinical parameters and patient preferences. 6,9 In 2013, guidance published by the National Institute for Health and Care Excellence (NICE) acknowledged the uncertainties surrounding treatment decision making for patients with early-stage breast cancer of this type and recommended that Oncotype DX be adopted by the NHS: Breast cancer patients face significant emotional and psychological strain when considering chemotherapy.
It can be particularly distressing for patients in whom the decision to have chemotherapy is unclear… Tools or tests that help people decide whether or not to have chemotherapy are likely to be greatly appreciated by patients. 3 Here, the availability of gene expression profiling is positioned as a positive development for individual patients, by "helping people to decide" whether to proceed to chemotherapy. Discerning who may (not) benefit most from chemotherapy is also important from a policy perspective, with the "over-treatment" of breast cancer having implications for health service costs and delivery. 3,10 Though social scientific research has explored clinicians' experiences of interpreting gene expression tests 9 and their impact on professional roles and identity, 11 less attention has been given to the ways in which these tools feature in patient decision making with regards treatment, or the role they play in experiences of cancer more widely. This is important for policy and practice; it has been established that chemotherapy decisions are shaped by social contexts, familial relationships and wider health histories, but little is known about how novel prognostic techniques intersect with these. In this article, we explore some of these factors as articulated by women within online accounts.
Existing sociological research has shown that medical decision making by patients is complex and situated. Treatment decisions may be thought of as "distributed" 12 -shared amongst patients, their families, clinicians and wider social networks, and as occurring across time and space. 13 Although patient participation in decision making is advocated within medical practice, it has been reported that patients vary in the degree to which they wish to take full ownership over treatment decisions in health care. 14,15 Alongside contemporary shifts in the provision of care, individuals are also seeking advice and support for medical decision making beyond the clinic through virtual platforms. 16 The Internet can be a source of second opinions, advice regarding symptoms or side-effects, and information about tests and treatments for those experiencing cancer. 17 Indeed, access to others' experiences of a shared health condition has been highlighted as a key aspect of online information seeking, with first person accounts of illness shaping treatment choices and the very experiences of ill health. 18 As categorizations of (some) cancers and treatment pathways become more diffuse, patients are today presented with large amounts of information about their condition and a range of different options, including novel therapies and clinical trial participation. In this context, insight into other patients' experiences and choices may be more salient, as individuals become more active in choosing treatments and options for long-term management.
To improve understanding of how novel techniques might be shaping patient decision making, in what follows we explore women's experiences of gene expression profiling (Oncotype DX) in early-stage breast cancer, as discussed within postings on UK online forums.
*Generally breast cancers that are human epidermal growth factor receptor 2 (HER2) negative, oestrogen receptor (ER) positive and lymph node-negative.

| Rationale
The online research presented here took place alongside qualitative interviews for a wider study exploring experiences of genomic techniques within contemporary oncology research and practice. The design of the research has been informed by members of two patient and public involvement (PPI) panels, who have raised predictive and diagnostic genomic testing as a topic of concern. Conversations with panel members suggested that these tests may produce false negative/positive or inconclusive results and indicated that uncertainties inherent within genomic testing should be discussed with patients and their families. As a recent introduction to NHS management of breast cancer, and with scant qualitative exploration of patient experiences, Oncotype DX testing was identified as a relevant technique through which to explore such issues within the remit of the wider research project.
The selection of online forums to access experiences of this healthcare technology was also informed by meetings with PPI panel members. Members have discussed the issue of diversity in experiences of cancer care, and the difficulties that may be faced by those who are "socially excluded" in articulating questions or complaints about care with health professionals. Online forum data provide access to such reflections, with Internet communication often used by individuals to follow-up clinical diagnoses and compare clinical information with other users. 18 Further, online forums provide access to a range of experiences from a large number of geographically diverse individuals, who may be excluded from face-to-face forms of qualitative research due to disability or their omission from opportunities to participate. 19 The method also responds to a recognition within the social sciences that Internet use is enmeshed with contemporary experiences of health and illness. 16,20 Indeed, as landscapes of disease categorization and treatment shift in line with advances in prognostic testing and therapies, decision making in cancer care is becoming more complex. 21 This has implications for online spaces and their functions, with these subject to ongoing reconfiguration.

| Data
To gather online accounts of women's experiences, two authors (ER and JS) searched for the term "Oncotype" within publically accessible online forums, hosted by two UK cancer charity websites: one supporting all cancer types and the other a breast cancer charity. Posts included in analysis were limited to those authored from April 2015, when the test was approved for NHS use, until May 2017. Discussion threads identified by the search were copied into Microsoft Word documents to facilitate qualitative analysis. This application was favoured over Computer Assisted Qualitative Data Analysis Software (CAQDAS) to enable easier movement between analytic memos, which were recorded alongside the text itself using the "comments" function.
Searches yielded a large amount of data. By way of example, one of the seven online forums featuring the word "Oncotype" contained 68 threads within the date range. Discussion threads on this forum contained between 4 and 75 individual posts. To generate a manageable dataset, data from two forums on the breast cancer charity website were excluded from analysis. These two forums focused on (a) experiences of recent diagnosis and (b) discussion of more general topics beyond cancer. Threads where Oncotype DX did not form the substantive content of discussion were excluded from the recent diagnosis forum, for example if the technique was merely named when recounting treatment pathways. The largest threads excluded here were from women seeking emotional support for specific aspects of their cancer experiences, for example questioning whether their emotions were "normal" (137 pages), and when feeling "low" (118 pages). A total of 32 threads, comprising 950 pages, were excluded, as well as an additional ongoing thread containing over 5000 posts; this also centred around support in the context of recent diagnosis. In the general discussion board, a total of two threads were excluded comprising 433 pages. The first was a thread welcoming users to the forum, and the second concerned a user seeking advice with regards her mother's cancer treatment.
Included threads encompassed comments from those who had not undergone the test themselves, but who had outlined anecdotal or media reported information about the Oncotype DX test, and from those who had unsuccessfully attempted to access gene expression profiling. From the cancer-wide website, 53 discussion threads from one forum were taken forward to analysis, and from the breast cancer charity website, 79 threads from six forums were included. This gave a final dataset of 132 discussion threads, comprising 639 pages.

| Data analysis
Analysis took a thematic approach, aligned with the analytic process described by Braun and Clarke. 22 The content of entire discussion threads selected for analysis was read by each author conducting the search, who "constantly compared" 23 the text within and between threads. This was performed with reference to existing sociological literature on cancer illness narratives, biomedicalization and treatment decision making. This process was also informed by ongoing PPI activities, during which differences in individual patients' desires to engage with clinical information have been emphasized, as have the uncertainties faced by patients when given information about diagnosis and prognosis. With this literature and PPI insight in mind, ER and JS each developed key themes from their set of data, which were shared electronically and deliberated during several in-person meetings. Each author then examined both sets of themes and associated extracts, and grouped relevant quotes and concepts within refined key themes. These were then shared between all authors and discussed and developed drawing on verbatim data, to cultivate overarching focal points for the presentation of findings.

| Ethical considerations
Online methods of data collection for sociological research are subject to wider ethical guidelines associated with the discipline, with safeguarding the interests of those involved in or affected by the research remaining paramount. 24

| FINDING S
In what follows, we show how the Oncotype DX test was represented by women using online forum spaces. We then go on to discuss how women described test results and their role in chemotherapy decision making, with this shaped by indirect experiences of cancer treatment, and sociocultural depictions of cancer and chemotherapy. Finally, we outline how users discussed the test in relation to inherent uncertainties with regards the success of treatment and possibility of cancer recurrence. In doing so, we highlight tensions between the claims of genomic testing to aid treatment decision making, and the uncertainties and anxieties which the procedure could provoke or leave unresolved. Related to this, several users emphasized the test's ability to provide "personalized" information about cancer. For example, one woman advised others to undergo the test because "it's based on you and you can then make an informed decision" (Stacey1954, Jan 2016). Another described that "what it does is show whether your specific tumour cells would benefit from chemo" (Telophene, Jun 2017). Due to the ability to provide what many discussed as a "specific" prediction of cancer recurrence, some represented Oncotype DX as providing certainty:

| Representations of gene expression testing within online environments
The onco[type DX] test is a very good test, which provides statistical evidence based upon your tumour. It's tailor made and was designed so that it takes the guess work out of whether you will benefit.

DiamondMary, May 2016
In expressing its ability to provide "tailor made" information, some Scepticism of these more widely used tools was also visible amongst others, with MayP (January 2017) describing these as "only a general indicator based on past recovery data that can be quite old." In the same thread, Pumpkin noted that where "generalised tools could be very wrong," Oncotype DX is "an individual test for you," which she interpreted as providing more surety with regards her treatment decision.
Betty45, who had not experienced gene expression profiling herself but was living with a heart condition she attributed to chemotherapy, displayed a particularly positive view of the technology. She invoked a hopeful future by saying "I think the answer for a large percentage of us will be the Oncotype DX test which shows whether chemo will work or not" (Nov 2015). Some thus positioned the test as not only providing certainty, but also as having the potential to transform breast cancer care. 26,27 Although this was not reflective of all users, with many also depicting the test not as a tool that eliminated guesswork, but one that "refined" guesswork, positive reflections on the test's role in decision making were visible throughout forum posts. In many cases, these could efface the complex and relational contexts within which decision making arose, which were also articulated within forum discussions. This is discussed further below.

| Gene expression testing and treatment decision making
Oncotype DX's production of a single figure to indicate recurrence risk, and corresponding recommendation to proceed (or not) to chemotherapy, was welcomed by many women. Some attributed a particular power to the test score in shaping their treatment choices.

| Attending to uncertainty?
Despite some women attributing the gene expression profiling result with particular authority in shaping treatment choices, others expressed scepticism of the test's ability to aid treatment decision making. Some users noted that despite clinician opinion, statistics and test results, ultimately there "are no crystal balls" (GrannyG, April 2017). In some cases, this was linked to the elusive and insidious nature of cancer as a disease, with posts describing "rogue" or "stray" tumour cells. Users' experiences, including memories of historical treatment, reinforced this sense of the unpredictability of cancer. None of us can see into the future, so we have to make a decision on the information we have.
As we have seen above, the "information we have" went beyond test results and clinical judgement, to indirect experience and expecta- As we have seen, the emphasis on the personal was informed by cultural narratives and memories of cancer and treatment, and an embodied sense of vulnerability to cancer cells. 33 In these contexts, women's responses to gene expression profiling test results are therefore not easily predicted; for example, those with a low recurrence score did not always indicate that they would eschew chemotherapy. In what follows, we discuss our findings, and their implications for sociological explorations of treatment decision making in the context of novel biomedical techniques. Whilst online forums have provided insight into intimate accounts of treatment decision making as shaped by gene expression profiling, their use to access women's experiences does have limitations. We cannot say with certainty that we were able to capture accounts from a diverse group of women, with Internet access likely to reflect wider social and structural inequalities. 36 This research, however, did not intend to be generalizable in a statistical sense, but to shed light on the breadth and potential complexity of decision making in the context of a novel genomic technique. 37 The absence of in-depth narratives, with analysis focusing on short posts which were sometimes devoid of context, has also meant that we were not able to explore wider impacts on decision making in great depth.

| D ISCUSS I ON
Further research is required to learn more about how patients seek and share information with others, and the impact of online forum use itself on decision making about cancer treatment. This may offer guidance to patients and clinicians about how online forums might be best used at this difficult time. These issues are being addressed by complementing this online research with ongoing qualitative interviews.

| CON CLUS ION
This study of accounts of gene expression profiling has shed light on how women are engaging with and negotiating novel genomic techniques as they become integrated within routine cancer care in the UK, and the resources they draw on in this regard. Importantly for clinical practice, we have shown that the women represented in this research did not always interpret Oncotype DX scores straightforwardly, with these results taking on varying significance according to factors including personal encounters with cancer, and potential regret for declining treatment. This emphasizes the importance of holistic treatment decision making between patients and clinicians, which may engage with loved ones' experiences of the disease, "gut feelings," emotions and anticipated futures.
Online forums have proved to be a valuable resource to explore perceptions of gene expression profiling as articulated by women in the midst of chemotherapy choices. These are also emblematic of the contemporary distribution of decision making, which has the potential to become reconfigured as genomic techniques and "personalized" treatment regimes become further integrated within clinical care. In-depth qualitative research will provide deeper insight into the emotional and embodied elements of these treatment choices, and their interplay with genomic techniques alongside more established means of informing treatment decisions in cancer care.

ACK N OWLED G EM ENTS
We are grateful to the anonymous online forum users whose comments are featured in this research and to forum moderators for consenting to our use of this material for our study. Thank you to Breast Cancer Care and Macmillan Cancer Support. We also acknowledge participants in our PPI panels for their contributions to our research and to the anonymous reviewers for their insightful feedback.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.