Communication about Prognosis during Patient-Initiated Second Opinion Consultations in Advanced Cancer Care: An Observational Qualitative Analysis

Prognostic communication is essential for patients with advanced cancer to enable informed medical decision-making and end-of-life planning. Discussing prognosis is challenging, and might be especially complex for oncologists conducting a second opinion (SO). Survival data are often lacking, and consulting oncologists need to consider previously conveyed information and patients’ relationship with the referring oncologist. We qualitatively investigated how advanced cancer patients and consulting oncologists discuss prognosis during audio-recorded SO consultations (N = 60), including prognostic information received from the referring oncologist. Our results show that patients regularly expressed implicit cues to discuss prognosis or posed explicit questions tentatively. Consulting oncologists were mostly unresponsive to patients’ cues and cautious to prognosticate. They also seemed cautious when patients brought up the referring oncologist. Consulting oncologists checked which prognostic information patients had received from the referring oncologist, before estimating prognosis. They agreed with the first opinion or rectified discrepancies carefully. Altogether, this study exposes missed opportunities for open prognostic discussions in SOs. Consulting oncologists could explicitly explore patients’ information preferences and perceptions of prognosis. If desired, they can provide tailored, independent information to optimise patients’ prognostic awareness and informed medical decision-making. They may additionally support patients in dealing with prognosis and the uncertainties associated with it.


Introduction
Cancer patients are confronted with a life-threatening disease, leading to considerable distress [1]. For many patients, the diagnosis involves complex decision-making about treatment and substantial uncertainty regarding their prognosis, i.e., the (in)curability of the disease and life expectancy [2][3][4][5][6]. Information about prognosis generally includes a high degree of uncertainty, as the precision of prognostic estimates and their predictive values for individual patients are limited [7,8]. In an era of rapidly developing anti-cancer

Study Design
The presented qualitative analysis was part of a larger prospective mixed-methods study about patient-initiated SOs in medical oncology (SO-COM), which examined communication during SOs and their psychological impact on patients [31]. The SO-COM study included audio recordings of SO consultations and self-report surveys from patients, referring oncologists and consulting oncologists (for more detailed information, see Lehmann et al. [31]). For the current analysis, we used the audio recordings of the SO-COM study to assess prognostic discussions by advanced cancer patients and consulting oncologists.

Recruitment and Sample
Medical oncologists affiliated with two Dutch tertiary referral centres were informed about the SO-COM study and invited to participate. Consecutive cancer patients scheduled for a SO with consenting oncologists were informed by hospital staff and consequently contacted by the study team. Both consulting oncologists and patients provided written informed consent for participation.
Patients scheduled for a SO were eligible for the SO-COM study if they had any type of solid tumour, were older than 18 years and had sufficient Dutch language proficiency. The SO-COM study originally comprised audio recordings of 69 patients. For the present analysis, we selected recordings of patients with advanced cancer stages, resulting in N = 60 consultations.
All procedures were in accordance with the Helsinki Declaration and were approved by the institutional medical ethics review board of the study contractor (Amsterdam University Medical Centres, NL63087.018.17; study number 2017_271) and the local review boards of both study sites.

Data Collection
The Standards for Reporting Qualitative Research were followed [43]. All audio recordings were fully listened to and relevant segments were anonymously transcribed verbatim. Relevant segments of the consultation involved (1) direct discussion of prognosis (i.e., (in)curability, life expectancy or the word "prognosis") or (2) communication about prognosis-related topics (e.g., life extension, survival gain, delay of tumour growth, treatment duration and effectiveness). The latter category was included because it may contain implicit information about prognosis, thereby indirectly affecting patients' prognostic awareness or triggering discussions of (in)curability or life expectancy. Relevant segments of significant others accompanying advanced cancer patients were included as well since these influenced prognostic communication during the consultation. All transcripts were imported into the coding software MAXQDA2020 [44].

Analysis
A qualitative analysis was performed using an inductive coding approach to allow for new themes to arise [45]. Data analysis was conducted by MK (last year medical student, trained in qualitative methods), NV (psychologist and researcher with expertise regarding prognostic communication, trained in qualitative methods) and MH (psychologist and researcher with expertise regarding second opinions and qualitative methods). NV and MK independently coded fifteen transcripts to enhance data triangulation, using open coding [46]. After every five transcripts, results were jointly discussed by NV, MK and MH to reach a consensus on an initial coding scheme. Throughout this process, the coding scheme was further refined and hierarchical layers emerged. Previously analysed transcripts were reviewed to make any necessary coding adjustments based on the updated coding scheme. When only a few new codes were added during the third consensus discussion, the coding scheme was finalised. MK coded all remaining 45 transcripts. In addition to segments that MK perceived as ambiguous, one out of every five transcripts was double-coded by NV and discussed to reach an agreement. After completion of the open coding process, prominent topics were derived from the final coding scheme. All three authors reviewed these to generate higher-order descriptive themes of prognostic communication in SOs for advanced cancer patients. Extensive discussions further explored potential links between recurring themes. Illustrating quotes were identified from the transcripts throughout the data analysis to substantiate findings.
Fifty-six of the analysed consultations included relevant segments (n = 56/60). In more than 80% of these consultations, the prognosis was discussed directly (n = 46/56 Discussions of prognosis were frequently initiated by patients or their significant others. They were often explicit about their wish for prognostic information, but still posed questions cautiously. The timing for starting these conversations varied across consultations. Several patients brought up prognosis immediately at consultation onset, suggesting that a need for more specific estimates had motivated them to request the SO (Q1). Yet, patients who had primarily sought the SO for other reasons, such as assistance in treatment decision-making, also uttered a wish for prognostic information (Q2). Rather than posing explicit questions about prognostic estimates, patients and their significant others often seemed to pursue a conversation about prognosis implicitly, by mentioning hope for optimistic scenarios or participation in future events. They did not ask for (dis)confirmation, nor did they subsequently request specific prognostic information. Consulting oncologists mostly did not respond to such implicit cues about prognosis (Q5). Q5 Significant other: 'Yes, she is sick, but she feels relatively well. And we hope we can go on like that for 40 to  In a few cases, consulting oncologists took the initiative to discuss prognosis with advanced cancer patients. When they did, it pertained to the incurability of the disease. Regarding life expectancy, oncologists were generally reluctant to provide estimates when patients asked for it. They additionally stressed the indeterminacy of prognosis (Q7 and Q8). Q7 Consulting oncologist: 'Well, and as for the future, the difficult thing is that crystal ball [to foresee the future]. They don't have one in [the referring hospital], but I don't have it here either.' -Consultation #7/male patient/55 years old/kidney cancer Q8 Consulting oncologist: 'Yes, that is always so difficult, because of course we can't tell you what your life expectancy is. That's so difficult. As doctors, we're actually not that good at estimating life expectancy. Certainly not when the patient sitting in front of us is in a good condition. When patients are hospitalised and they're very ill, we could say, "This is not going to last longer than a few days". We can do that. But everything else, we can't. So I never do [provide estimates]. I also just don't know.' -Consultation #8/female patient/41 years old/melanoma Consulting oncologists commonly explained the indeterminacy of prognosis by pointing out the influence of other uncertain factors, such as treatment effectiveness. Others expressed concerns about being held responsible for their prognostic statements (Q9).
Q9 Consulting oncologist: 'The difficulty of communicating prognosis is that it will start living a life of its own. The doctor specifies a certain amount of time and then it's always like, "But the doctor said this and in the end, it turned out to be that".' -Consultation #9/female patient/60 years old/breast cancer In a few consultations, consulting oncologists more strongly expressed their apprehension about prognostication in late-stage cancer, which they suggested would reflect medical incompetence (Q10). These comments may relate to the lack of survival statistics for advanced cancer generally, and for patients receiving multi-line therapy or surpassing prognostic estimations specifically. Q10 Consulting oncologist: 'Well, it's like that indeed: we can't say anything. You can see that looking at your entire disease history. The fact that someone is still responding to a treatment so well after 2 years, 2.5 years even, is unique already. The way you are right now, that's unique. Um, so in other words, you should disregard everyone who would predict something right now and view them as incompetent.' -Consultation #10/female patient/36 years old/breast cancer

Generic Terminology and Disclaimers for Precise Estimates
When consulting oncologists did provide information about life expectancy, they mainly used imprecise terms that were open to interpretation (Q11 and Q12). For example, they used words like "short" or "fast" to express some sense of time, as well as wide non-numerical time frames (e.g., "months to years"). Apart from discussions about (in)curability and life expectancy, nearly all analysed consultations involved communication about topics strongly related to prognosis. This concerned issues such as life extension, survival gain, delay of tumour growth, treatment duration and its effectiveness. Consulting oncologists appeared less reluctant to provide numerical information about these subjects as compared to life expectancy (Q16 and Q17).
Q16 Consulting oncologist: 'So it is possible . . . There are patients who still benefit from it [chemotherapy] after 6, 7, 8, 9, even 10 years.' -Consultation #12/female patient/58 years old/melanoma Q17 Consulting oncologist: 'My first advice is to use the therapies that have proven useful, chemotherapy and such, as you will live 4 times longer with chemotherapy than without it.' -Consultation #13/female patient/48 years old/colon cancer When discussing prognosis-related topics, consulting oncologists did not explain the implications for patients' (in)curability or life expectancy. Yet, some patients appeared to infer information about their prognosis from these related topics (Q18). A few times, this subsequently triggered a conversation about prognosis.
Q18 Consulting oncologist: 'Your life will be divided into periods of 2 months. You will get 3 chemotherapies and a week of pills. That will be repeated 3 times before you get the next CT scan. Then we'll decide whether we'll give you another 3 cycles. After 6 cycles we will take a break. About 40% of the consultations involved conversations about prognostic information previously provided by the referring oncologist (n = 24/56). These discussions were predominantly initiated by patients and their significant others, who sometimes simply repeated what the referring oncologist had said (Q19 and Q20). Consulting oncologists in our sample did not explicitly explore these implicit cues to talk about previous prognostic information. Occasionally, patients brought up the referring oncologist to express their dissatisfaction with previous prognostic communication (Q22). This most often concerned how prognosis had been disclosed, rather than the content of the information.
Q22 Patient: 'At the start they [referring doctors] said, "You've got up to 5 years". They said that at the start and one week later they said to us, "Well, maybe, 80 percent will die within one year in your case". So I found that a little bit, yes, that was quite heavy.' Consulting oncologist: 'Another setback.' Patient: 'Yes, that's a setback. I also said to that doctor, "Don't do that anymore". Up to 5 years, then we're already thinking it's going to be years, but it won't be years according to those information booklets.' -Consultation #16/female patient/69 years old/oesophagus cancer In a few consultations, patients and their significant others introduced the referring oncologist into the conversation to explicitly ask the consulting oncologist for additional prognostic information, sometimes persistently. Again, they seemed to indicate an understanding of the complexity of prognostication for oncologists (Q23). In a few cases, consulting oncologists initiated communication about the referring oncologist. Some did so to check which prognostic information had already been communicated, before disclosing the prognosis themselves. Generally, these attempts to obtain information about a previous prognosis seemed careful. In reaction to patients' reports, consulting oncologists often supported prognostic information from the referring oncologist (Q24). Notably, one patient refused to disclose what the referring oncologist had said and specifically requested an independent prognostic assessment from the consulting oncologist (Q25). In none of the analysed consultations did the consulting oncologist firmly contradict the referring oncologist. Consulting oncologists who expressed doubts about (parts of) the information that was previously conveyed, only exposed minor discrepancies and did so carefully (Q26). When subtly rectifying previous estimates, consulting oncologists often appeared to explain the discrepancy between both estimates (Q27).

Main Findings
This study explored how advanced cancer patients and consulting oncologists discuss prognosis during SO consultations. Our analysis revealed that patients and their significant others usually initiated prognostic communication and introduced the referring oncologist into the conversation, using explicit as well as implicit communication.
Consulting oncologists were overall cautious in these discussions, as they were mostly unresponsive to patients' cues, stressed the indeterminacy of prognosis and provided ambiguous rather than explicit estimates. When asking about previously conveyed prognostic information, consulting oncologists often supported the first opinion and rectified discrepancies carefully.
In line with previous work, the current study demonstrated advanced cancer patients' need for prognostic information. Yet, we also exposed their struggle to start prognostic conversations straightforwardly. Patients expressed implicit cues, for example by presenting their expectations for the future and hope for positive outcomes, and posed even their explicit questions cautiously. Previous research suggests that patients commonly utter their informational and emotional needs in implicit rather than direct ways. Such indirect communication has been observed in conversations about other sensitive topics too, like alternative treatment options and death [47][48][49]. Patients' latent communication might indicate an ambivalence towards receiving prognostic information. Especially as death is impending, it appears that patients want to be informed about their prognosis, but at the same time do not want to know [48,50,51]. They seem to recognise the practical reasons for prognostic disclosure, but fear the emotional impact of knowing [51]. Alternatively, our findings could reflect patients' politeness, wishing not to complicate the conversation by introducing a sensitive subject [52]. Patients' use of implicit cues might be problematic, as the literature indicates that physicians experience difficulty in detecting indirect expressions [52][53][54]. The latter was reiterated in our study. Possibly, consulting oncologists are unaware of patients' hidden cues. Alternatively, they might not perceive an urgency to respond to implicit requests, or feel inadequate to meet patients' prognostic information needs [48,52,55].
When consulting oncologists did respond to patients' requests for prognostic information, they were particularly cautious about estimating life expectancy. Reluctance to prognostic communication has been observed in regular consultations before [8,30], indicating that barriers to discussing prognosis remain substantial across different settings. Insufficient time is often reported as a complicating factor [29,56,57], yet might be less relevant in the relatively long SO consultations in our sample (M = 39.57 min). Undoubtedly, prognostication is especially complex when conducting SOs in advanced cancer care. While the determination of incurability might be fairly clear-cut, overall survival varies widely and oncologists regularly overestimate individual life expectancy [26,[58][59][60]. Therefore, they might not always be able to fulfil patients' desire for certainty. In addition, personal barriers may play an important role in oncologists' cautiousness, as previous studies showed their discomfort regarding sensitive topics. For prognosis specifically, oncologists worry about being blamed by patients, inflicting distress or failing to answer questions [48,52]. The uncertainty of estimates may further diminish their confidence in discussing this topic [59]. Hence, our results may reflect a struggle between wanting to meet patients' needs, while avoiding definite prognostication and its feared consequences [61]. It is understandable that oncologists use strategies that might mitigate this discomfort, like ambiguous language [62]. However, this should not prevent opportunities for constructive conversations about prognosis. Interestingly, we observed that some patients expressed awareness of the difficulty of prognostication, suggesting that they might understand the limits to oncologists' predictions. This may encourage oncologists to engage in prognostic discussions without fear that patients expect precise estimates. Additionally, several studies show no or even positive associations between prognostic disclosure and patients' emotional well-being and the oncologist-patient relationship, which may reassure oncologists as well [18,29,[63][64][65][66][67].
We observed that consulting oncologists were also cautious when patients mentioned prognostic information previously provided by the referring oncologist. They carefully asked patients about previous estimates before disclosing prognosis and often supported or justified prognostic statements provided by the referring oncologist. Possibly, consulting oncologists agreed with the first medical opinion. Another explanation could be that they perceived discrepancies as negligible, taking into account that minor differences may needlessly confuse or harm patients [37,38]. Alternatively, consulting oncologists might downplay inconsistent information to avoid openly disagreeing with the first opinion, thereby guarding patients' relationship with the referring oncologist. This potentially reflects consulting oncologists' awareness of the sensitivities of the SO setting [35,37]. Importantly, inconsistencies in prognostic estimates between the first and the second opinion may not necessarily originate from a medical disagreement between both oncologists. It is well known that patients' recall of information may be impaired, especially in bad news consultations [68]. Moreover, patients may perceive and react to medical information differently at different stages in their disease trajectory, as each consultation changes patients' emotions and attitudes towards their illness. Patients seeking SOs could be more "ready" to hear prognosis and therefore experience prognostic information as novel or different [37,38].
In contrast with the aforementioned results, we noticed less caution among consulting oncologists on a few occasions. First, some consulting oncologists disclosed life expectancy while remarking that patients had probably already heard their prognosis from the referring oncologist. Yet, they did not explicitly explore patients' prognostic awareness beforehand. Second, consulting oncologists were less reluctant to provide numerical estimates about subjects other than life expectancy. Perhaps they were unaware of the implicit message in prognosis-related topics. Alternatively, consulting oncologists felt more comfortable in talking about, for example, treatment effectiveness, because they had more evidence available or perceived such information as less confrontational [11,48,62,69]. Nevertheless, both behaviours could confront patients with novel, possibly unwanted, yet impactful information. This may startle or upset them, and consulting oncologists should address such patient distress [61,70,71]. Furthermore, the discussion of prognosis-related topics specifically might confuse patients' understanding of the prognosis, if not substantially clarified. Patients could silently infer prognostic information from these topics, preventing oncologists from rectifying incorrect ideas. Correspondingly, earlier research suggested that oncologists' propensity to talk about treatment details and use of medical jargon (e.g., delay of tumour growth) distracts from prognostic discussions and exacerbates patients' misconceptions [30,62,69,72].

Strengths and Limitations
The current study was the first to explore prognostic communication in the unique setting of SOs in advanced cancer care, uncovering how patients and consulting oncologists talk about prognostic information previously provided by the referring oncologist. Open and double coding of observational data, as well as recurrent extensive consensus discussions with a multidisciplinary team, supported the quality and reliability of the results. A limitation of this research concerns its limited generalisability, as only two hospitals in the Netherlands participated. Oncologists in other contexts might carry out SOs and/or prognostic communication differently. Furthermore, the participating patient group may not be representative of the entire population of advanced cancer patients. Lastly, we did not distinguish between prognostic communication by patients and their significant others, as those findings were often intertwined and we aimed to explore prognostic discussions in SOs comprehensively. Nevertheless, separate analyses of patients and significant others may yield additional insight into the dynamics of prognostic communication.

Practical Implications and Recommendations
The results of this study corroborate the sensitivity of prognostic discussions for advanced cancer patients and consulting oncologists, as both regularly struggled to discuss prognosis straightforwardly. This yields missed opportunities for open prognostic discussions, which could lead to prognostic unawareness, misunderstandings and hampered medical decision-making among advanced cancer patients. Therefore, prognostic communication should be optimised. Current guidelines suggest that oncologists need to explicitly assess patients' prognostic information preferences first while also considering the minority of patients that do not want to discuss life expectancy [17,18,61,66,73]. Consulting oncologists should listen actively for implicit cues and pose clarifying questions, empowering patients to verbalise their needs and ideas about prognosis. If it is desired, and depending on patients' preferred level of specificity, consulting oncologists could offer explicit (e.g., numerical) prognostic information based on the available evidence. It should be made clear that statistics apply to groups of patients and are not specific to individuals. Consulting oncologists may provide the worst, typical and best-case scenarios of survival to balance realism with hope [18,66,73]. However, especially in the setting of SOs for advanced cancer, survival data are often lacking. This may require oncologists to carefully provide assessments based on their clinical experience. It is crucial to explain the nature and limitations of any estimate and check patients' understanding of it [18,61,66,73]. When discussing prognosis-related topics, oncologists should clarify the relevance to patients' life expectancy. Overall, an expert, emphatic and collaborative approach to communicating prognosis is recommended [18,73,74]. For future research, it is essential to complement current guidelines by examining which strategies to communicate prognosis benefit patients most.
Considering the triadic relationship characterising SOs, consulting oncologists could additionally explore what prognostic information patients have previously acquired from the referring oncologist. Although patients' reports of the first opinion may be biased, it will enable consulting oncologists to assess what patients understood from it and whether that matches their own assessment. Consulting oncologists may agree with prognostic information gained from the referring oncologist or find (minor) inconsistencies, but should always ensure an honest, independent opinion if requested. Our findings, as well as previous literature, suggest that consulting oncologists are already committed to safeguarding patients' and their own relationship with the referring oncologist [35,36]. Hence, we propose being transparent about significant inconsistent information. Consulting oncologists have the opportunity to complement, nuance or adjust patients' perception of their prospects, given that patients might be more "ready" to hear prognosis when receiving a SO [37,38]. Helping patients understand discrepancies could prevent misconceptions, and facilitate patients' awareness of prognosis and the uncertainties associated with it. Lastly, supporting patients in dealing with prognostic uncertainty seems vital. Consulting oncologists could provide them with a sense of control by explaining what to expect regarding the disease trajectory, offering a clear plan forward and emphasising the continuity of care [75][76][77].

Conclusions
This study substantiates advanced cancer patients' need for and oncologists' cautiousness about prognostic discussions during SO consultations. Barriers to prognostic communication seem heightened in this sensitive setting, yet the relevance of transparent conversations about prognosis might be too. In order to overcome missed opportunities, it is essential for consulting oncologists to explicitly explore patients' information preferences and perception of their prognosis. If it is desired, consulting oncologists may provide tailored, independent information to optimise patients' prognostic awareness and well-informed medical decision-making. Additionally, offering patients support in dealing with prognosis and the uncertainties associated with it seems valuable.