Accuracy of curability expectations in patients with gastrointestinal cancers

Abstract Background Assessment of illness and treatment understanding among cancer patients has largely focused on those with advanced disease. Less is known about patient expectations at earlier stages of cancer and potential modifiers of accurate understanding. Methods We assessed accuracy of cure expectations in patients across all stages with gastrointestinal (GI) cancers. Accuracy was determined by independent reviews of patient health records by oncologists on the investigative team. Impact on cure accuracy of selected clinical variables and health‐information preferences was analyzed. Results Hundred and thirty five patients were included for analysis, with 100% interrater agreement for accuracy between oncologist reviewers. Sixety five patients (48%) had accurate cure expectations from their cancer treatment. Accuracy was lower in Stage IV versus Stage I–III disease (35% vs. 63%, p < 0.01), lower in unresectable versus resectable disease (35% vs. 67%, p < 0.01), and higher in patients with early‐stage disease who received adjuvant chemotherapy versus those who did not (78% vs. 53%, p = 0.04). Accuracy did not differ by health‐information preferences and remained stable over time. Of 63 patients who died, baseline accuracy differed by location of death (p = 0.03), with greater accuracy in those who died with home hospice (56%). Accuracy was lower in those who were hospitalized in the last 30 days of life versus those who were not (25% vs. 59%, p = 0.01). Conclusions Inaccurate cure expectations are prevalent across all stages of GI cancers, but particularly among those with metastatic or unresectable disease. High‐quality, iterative communication strategies may facilitate patient illness and treatment understanding throughout the disease course.


| INTRODUCTION
Understanding the intent of cancer treatment is imperative for patients' informed consent and medical decisionmaking. Patients therefore, deserve to receive prognostic and treatment-related information, when desired, from their oncology teams in a clear and personalized manner. 1,2 Prognostic awareness and acknowledgment of the uncertainty about life expectancy can help patients plan for their future and make decisions according to their personal values and goals. 3,4 Though receipt of prognostic information may be distressing, [5][6][7][8] skillful communication by clinicians can support patients to integrate this information while preserving hope and psychological wellbeing. [9][10][11] Despite the benefits of serious-illness and prognostic discussions to increase awareness, [12][13][14] patients may still have misunderstandings about their illness severity and the intent of cancer treatment. In a landmark study by Weeks et al, most patients with newly-diagnosed metastatic lung or colorectal cancers expected to be cured from chemotherapy. 15 Unrealistic expectations regarding curability may result from ineffective communication, the choice to not believe that one's disease is incurable, and/or the use of optimism as a coping strategy. 16 Nevertheless, oncology care teams need to conduct high-quality, empathic, and iterative conversations about the realistic goals of cancer treatment, in order for patients and families to feel supported and informed throughout the disease course.
Historically, research has focused on evaluating illness and treatment understanding in patients with metastatic or advanced cancer. Evidence about expectations of patients with earlier stage cancer is scant. In addition, little is known about an individual patient's evolution of their expectations over time and potential modifiers of, and associations with, accurate understanding. Moreover, the prior research predated the groundbreaking progress of molecularly-driven and immune-based cancer therapies that have improved survival outcomes, and was limited to a one-time survey assessment for patients who received chemotherapy. 15 A contemporary evaluation of illness and treatment understanding in our current era of cancer therapeutics and in the context of real-world clinical practice is needed. We therefore, analyzed the accuracy of cure expectations in patients across all stages of gastrointestinal (GI) cancers and characterized accuracy in relation to defined clinical variables and healthinformation preferences. We further examined whether expectations changed over time and if accuracy varied by end-of-life care outcomes.

| METHODS
This prospective cohort study was conducted at Memorial Sloan Kettering Cancer Center (MSK) as a pre-planned component of a clinical initiative to normalize, systematize, and improve the quality of "primary" (non-specialist) palliative care by oncology teams supported by palliative care specialists. [17][18][19] The study cohort consisted of patients with newly-diagnosed GI cancers, regardless of disease stage, receiving care from medical oncology clinics from July 2017 to December 2020. For this analysis, study participants were followed until death or August 31, 2021, whichever occurred sooner. There were no exclusion criteria. MSK's Institutional Review Board waived informed consent based on minimal risk to participants.
As previously described, the larger clinical initiative comprised a structured, sequenced, and iterative series of assessments by outpatient oncology nurses to elicit patients' core health-related values, symptom burden, and understanding of their illness and intent of cancer treatment. 17,18 To assess illness and treatment understanding (ITU), patients were asked three questions derived from a validated questionnaire used in a large national observational study of outcomes including perceptions of care among patients with metastatic colorectal and lung cancer. 15 Specifically, patients used a 4-point Likert scale to rate the likelihood ("not at all likely," "a little likely," "somewhat likely," "very likely") that their treatment would: (1) Cure their cancer; (2) prolong life; and (3) help with problems caused by the cancer. 15 Along with these items, all patients were asked if they preferred to receive information about their health in detail or as an overview without much detail, 17 and a subset of study patients were asked to assess the quality of communication provided by their treating oncology care teams, using a validated questionnaire (e.g., "how often does your care team… listen carefully to you, explain things in a way you could understand, give you as much information as you wanted about your cancer treatments?" and "how good is your care team at… involving you in treatment discussions about your care, asking you about important things in life, discussing your values/goals?"). [20][21][22] Questionnaires were administered in person in clinic, initially at the second follow-up clinic visit and then quarterly to evaluate changes in patient responses over time.
The primary objectives of this analysis were to assess the accuracy of patients' reported expectations for the likelihood of cure× from their cancer treatment at baseline, and to determine if accuracy rates varied by clinical variables and health information preferences. Two gastrointestinal medical oncologists (R.A. and A.S.E.) and one surgical oncology fellow (P.S.) independently assessed accuracy of patients' responses. Accuracy was determined by first review of patient data, including: GI cancer type, disease stage, recurrent or newly diagnosed disease, resectability, and receipt of neoadjuvant versus adjuvant versus metastatic treatment. The oncology reviewers then used clinical practice standards to classify the intent of cancer treatment for each patient's clinical scenario as either curative or non-curative.
Patient responses were classified as accurate if their cure expectations were consistent with the intent of treatment determined by the oncology reviewers, and patient responses were classified as inaccurate if their cure expectations were unrealistic compared with reviewers' projections. For example, a patient with Stage IIIA colon cancer receiving adjuvant chemotherapy who reported that their cancer treatment was "somewhat likely" to be curative was deemed to have answered accurately; a patient with Stage IV pancreas cancer receiving chemotherapy who reported that their cancer treatment was "very likely" to be curative was deemed have answered inaccurately; a patient with Stage IV gastric cancer receiving combined chemotherapy and immunotherapy who reported that their cancer treatment was "not at all likely" to be curative was deemed to have answered accurately.
Independent assessments of the accuracy of patients' responses were then compared to each other to determine the degree of interrater agreement. When assessments of accuracy were not in agreement, these clinical cases were jointly reviewed to arrive at a consensus. Patients' responses were subsequently coded as accurate or inaccurate.
For study participants who completed the ITU questionnaire more than once (i.e., at baseline and a subsequent quarterly interval), change in accuracy was evaluated by comparison of baseline and the last completed questionnaire response. For those who died within the study period, electronic health records were reviewed to assess the following outcomes: location of death, hospice enrollment, receipt of cancer treatment within the last 30 days of life, and hospitalizations within the last 30 days of life. Accuracy of cure expectations was then evaluated relative to each end-of-life outcome.

| Statistical analysis
Participant responses to the ITU, information preferences, and quality of communication questionnaires were summarized descriptively. Rate of cure expectation accuracy was estimated among all patients at baseline. Baseline cure accuracy was compared between groups with the Chi-square test, including by demographic and clinical variables, for patient-reported information preferences (for information in detail vs. no detail), and for end-of-life outcomes in those who died. Longitudinal data on curability expectations were used to identify changes in accuracy over time (either more or less accurate compared to reviewer classification) and reported descriptively. All tests were evaluated for statistical significance at alpha level 0.05. Statistical analysis was performed using SAS version 9.4.

| Study participants
A total of 135 patients with GI cancers were available for analysis. Demographic and disease characteristics are described in Table 1. The median age of the study cohort was 61 years, with more men (60%) than women. Most patients identified as white (75%) and Non-Hispanic/Non-Latinx (94%). Most patients were married (66%). There was diversity among patients who identified with a religion (73%). Slightly more than half of the patients in the cohort had Stage IV disease and slightly less than half had Stage I-III disease. All patients completed the ITU questionnaire at baseline. The information preferences and quality of communication questionnaires were completed at baseline by 125 (93%) and 60 (44%) patients, respectively. In addition, 73 (54%) patients completed the ITU questionnaire on at least one follow-up visit (median 3 questionnaires completed, range 2-7) and were included for longitudinal analysis. Table 2 depicts accuracy rates by clinical and patientreported variables, and Table 3 summarizes patient questionnaire responses. Agreement in accuracy assessments between oncologist reviewers was 100%. Regardless of disease stage, approximately 70% of patients thought it was very or somewhat likely for their treatment to cure their cancer, 80% thought it was very or somewhat likely for their treatment to alleviate cancer-related problems, and over 90% believed that treatment was very or somewhat likely to prolong their life ( Table 3).

| Accuracy of curability expectations
Sixty-five patients (48%) were determined to have accurate expectations for the likelihood of cure from their cancer treatment at baseline ( Of patients who completed the baseline information preferences questionnaire (n = 125), most patients preferred to receive information about their health in detail (86%) and with a family member or friend present (82%) ( Table 3). The accuracy rate for patients who preferred a general overview was lower than for those who preferred detailed health information (33% vs. 52%, p = 0.14); however, these results were not statistically significant ( Table 2).
Of patients who completed the baseline quality of communication questionnaire (n = 60), nearly all patients selected responses to indicate that they felt heard, were able to share their personal values and goals, had the opportunity to ask cancer-related questions and discuss their concerns, were provided treatment information in a way that could be understood, and felt included by their care team for decision-making ( Table 3). Among patients who died (n = 63), baseline accuracy significantly differed by location of death and hospitalization within the last 30 days of life but did not differ if patients received cancer treatment during the last 30 days of life ( Table 5). Cure accuracy was higher for those who died at home with hospice (56%) compared to those who died in hospitals (14%), or an inpatient hospice facility (20%), or at home without hospice (0%) (p = 0.03). Cure accuracy was lower in patients who were hospitalized compared to those who were not hospitalized within the last 30 days of life (25% vs. 59%, p = 0.01).

| DISCUSSION
In this single-center, prospective cohort study, we examined the accuracy of curability expectations among patients with GI cancers regardless of stage, and the influence of selected clinical factors on accuracy. Our analysis revealed that half of the patients had accurate baseline expectations. Accuracy of curability expectations was greater in patients with earlier stage disease compared to metastatic disease, in patients with surgically resectable compared to unresectable disease, and in patients with early-stage disease who received adjuvant chemotherapy compared to those who did not. Accuracy did not differ by type of systemic treatment (immunotherapy vs. chemotherapy), GI cancer diagnosis, or DNR status. Our study contributes real-time and longitudinal information about illness and treatment expectations from a large cohort of patients, inclusive of all stages of disease, receiving modern-day cancer therapeutics.
The strengths of our data include disease heterogeneity, high interrater agreement for coding accuracy, and use of questions derived from validated questionnaires. 4,15,17,20 Moreover, surveys were administered in the setting of our institution's primary palliative care initiative that centered on eliciting patient values as well as assessing for and addressing unmet needs such as uncontrolled symptoms and information about disease and treatment intent. 17 That is, our assessment of patients' expectations from their cancer treatment did not occur in isolation, but was dovetailed with an endeavor to deliver palliative care and enhance serious-illness conversations in actual clinical practice, thereby informed by these proactive discussions. As patients had the opportunity to reflect and provide responses longitudinally while receiving their cancer care, our results thus provide up-to-date and valuable additions to the evidence on patients' expectations and understanding.
Our methodology was also refined, as each individual patient's disease course and clinical characteristics were reviewed by oncology experts to determine accuracy, instead of categorically classifying survey responses as accurate or inaccurate for patients with advanced disease without likelihood of achieving cure. 15 For example, when utilizing this more personalized clinical approach, a patient with Stage IV colorectal cancer with a single hepatic metastasis who reported that it was "somewhat likely" that chemotherapy could provide cure would be classified as accurate, whereas this would have been an inaccurate response based on methodology used in Weeks et al. Our results primarily suggest that patients receiving treatment for metastatic disease may have more difficulty understanding the intent of this treatment than those receiving adjuvant treatment for potentially curable cancer. This may be in part due to existential distress often experienced by patients with metastatic disease, who receive treatment while facing mortality sooner than anticipated or hoped. Yet, it is understandably human for patients to feel this way, especially as drug development continues and as newer therapies reinvigorate optimism for tumor response. [23][24][25] There are a number of factors that could influence patient perceptions, 16,26 and prior research has even demonstrated that receiving life expectancy statistics does not impact patients' prognostic expectations. 27 Additionally, patients with earlier-stage disease who receive adjuvant chemotherapy may have a better understanding than those who do not, perhaps because oncologists may spend more time to explicitly review the goals of treatment. When oncologists recommend adjuvant chemotherapy for the aim, but not guarantee, of eradicating any residual micro-metastatic disease, this may lead patients to an accurate understanding of

Illness and treatment understanding questionnaire
Give you as much information as you wanted about your cancer treatments? Talking to you about your feelings concerning the possibility that you might get sicker?
9 (1,10) Talking to you about the details concerning the possibility that you might get sicker? Three patients who completed the illness and treatment understanding questionnaire did not answer the question regarding curability but were determined to have clearly incurable disease, and these patients were classified as having an inaccurate understanding. b Instructions indicate to pick one answer only.
c Instructions indicate to pick one or more answers. Twenty-six patients had no response to question (3) and responses are reported for n = 99. d Responses to these questions were recorded on a 0-10 Likert Scale, along with an option for indicating "did not discuss." Three patients indicated that they did not discuss any communication aspects related to this question.

T A B L E 3 (Continued)
treatment intent more often than in patients in whom no such chemotherapy is prescribed. Our results also support findings from existing literature on the relationship between information preferences and curability expectations, 28 the stability of prognostic awareness over time, 29 and the impact of curability expectations on end-of-life care outcomes. 30 The body of evidence to which the present study adds can help inform how oncology care teams use patients' expectations to guide conversations at times of disease progression and/or clinical decline, with a greater attention on clarifying patients' illness understanding as it relates to their personal values, prognosis, and end-of-life care preferences.
A contemporary evaluation of the accuracy of cancer patients' expectations highlights the importance of iterative and tiered (e.g., at diagnosis, recurrence or progression, and near the end of life) communication strategies that support patients as they cope with their illness and prognosis. Even when prognosis is uncertain, oncology care teams should be cognizant of how patients may interpret what has been communicated and how they may perceive their illness, as misconceptions can lead to downstream effects on end-of-life care. Ongoing research is needed to develop personalized, simple, and effective methods of conveying prognosis and facilitating understanding, when and as desired by patients and caregivers. [31][32][33][34] Future communication interventions may require specific roles for nursing professionals, 17,35,36 oncologists, and palliative care specialists, with a framework that can be acted upon early and that can adapt to all stages and types of cancers.
Our study has several limitations. This study was conducted at a single academic cancer center and our findings may not be generalizable to other healthcare systems. Though MSK's primary palliative care initiative was implemented in only a select number of GI oncology clinics, 17 we studied a diverse patient population across all stages and GI cancer types. In addition, despite high interrater agreement, a ceiling effect was possible in our methodology; additional reviewers of patient responses may have resulted in a lower agreement level on accuracy assessments. Higher patient-reported ratings of physician communication quality have previously been associated with less accurate and more unrealistic expectations for the goals of treatment. 15,37 A ceiling effect may also therefore, explain why most patients reported that their oncology care teams provided high-quality communication about their health and personhood. While retrospective, end-of-life outcomes data were available for almost all study patients. Finally, our analysis focused only on quantitative, Likert-scale, patient responses to validated surveys on ITU and curability. As described previously, 16 such analyses may limit the strength of inferences or conclusions, as inaccurate responses may reflect a lack of information, inadequate communication, and/or patient-level barriers to integration of information or reporting of understanding. Future investigations will incorporate qualitative data and other associations, such as the degree of belief in miracles, 38 to explain inaccuracies in patient  understanding and combine with quantitative responses for mixed-methods analysis.
In summary, we explored accuracy of curability expectations in patients with GI cancers regardless of stage and identified lower accuracy rates among those with metastatic disease, unresectable disease, and early-stage patients who did not receive adjuvant chemotherapy. Additional studies will evaluate more disease types and the associations of distress (physical, emotional, spiritual), including worry about death and dying, 39 on patient expectations. Accurate illness and treatment understanding by patients can support them and their care team to align around the goals and plans for oncologic care.