Advanced Cancer Patients’ Changes in Accurate Prognostic Understanding and Their Psychological Well-Being

Abstract

Context:

Clinicians often worry that patients’ recognition of the terminal nature of their illness may impair psychological well-being.

Objectives:

To determine if such recognition was associated with decrements to psychological well-being that persisted over time.

Methods:

Eighty-seven advanced cancer patients, with an oncologist-expected life-expectancy of less than 6 months, were assessed before and after an oncology visit to discuss cancer restaging scan results, and again at follow-up (median time between assessments, approximately 6 weeks). Prognostic understanding (PU) was assessed at pre- and post-visit, and a change score was computed. Psychological well-being was assessed at pre, post, and follow-up, and two change scores were computed (post minus pre; follow-up minus post).

Results:

Changes toward more accurate PU was associated with a corresponding initial decline in psychological well-being (r = −0.33, p < .01), but thereafter was associated with subsequent improvements (r = .40, p < .001). This pattern remained controlling for potential confounds. Patients showed different patterns of psychological well-being change (F = 3.07, p = .05; F = 6.54, p < .01): among patients with improved PU accuracy, well-being initially decreased, but subsequently recovered; by contrast, among patients with stable PU accuracy, well-being remained relatively unchanged, and among patients with decrements in PU accuracy, well-being initially improved but subsequently declined.

Conclusion:

Improved PU may be associated with initial decrements in psychological well-being, followed by patients rebounding to baseline levels. Concerns about lasting psychological harm may not need to be a deterrent to having prognostic discussions with patients.

Introduction

Among patients with advanced, life-limiting cancers, an accurate understanding of the terminal nature of one’s prognosis is essential for informed treatment decision-making and end-of-life care planning.^1–4 In fact, accurate prognostic understanding (PU) is associated with positive outcomes including higher likelihood of advance care planning and receipt of value-consistent end-of-life care.^5,6 Therefore, clinicians helping patients develop more accurate PU is advantageous to improving end-of-life care.

Clinicians often worry, however, that helping patients recognize their terminal prognosis might damage patients’ psychological well-being.^1,3 In fact, some studies suggest that more accurate PU is associated with worse well-being.^7–10 These findings have limitations due to the methodological limitations of this prior work; most notably, cross-sectional designs and/or one-time assessments of PU and well-being. Such designs are problematic as they cannot capture any dynamic, changing nature of the underlying emotional response. The underlying emotional response might be one of initial shock and worse psychological well-being, followed by patients rebounding to baseline well-being, a pattern born out in studies of other negative experiences (e.g., bereavement).^10–12 Single assessments of PU and psychological well-being cannot gauge whether patients accurate prognostic understanding is recently gained or long standing, and whether psychological well-being fluctuated following accurate changes in prognostic understanding.

The present study accordingly tests the unanswered question of whether gaining more accurate PU is associated with decrements to psychological well-being that persists over time.^7,8,13 The study used a repeated measures design with PU examined twice, before and after a cancer restaging scan results discussion. Psychological well-being was assessed at those same times, and again at a follow-up assessment. Specifically, the present study examined if changes in PU over time were associated with corresponding, and subsequent changes, in psychological well-being. This design, capitalizing on changes in the variables over time, has the advantage of homing in on patient recognition of prognosis as it is happening (as opposed to PU that is long-standing and which patients have emotionally accepted) and its consequence on psychological well-being.

Methods

Data for these analyses were drawn from the Coping with Cancer-II study, a multi-institution observational study conducted from 2010 to 2015 to examine end-of-life communication and its influence on prognostic understanding of advanced cancer patients. Participating sites are listed in Table 1. Study procedures were approved by the Institutional Review Boards at all participating sites and all participants provided written informed consent.

Table 1:

Sample Characteristics

	n (%)
Age (n = 87)	Mean = 60.65, SD = 8.83
Education, years (n = 86)	Mean = 15.17, SD = 3.12
Sex
Male	23 (26.4%)
Female	64 (73.6%)
Married
Yes	54 (63.5%)
No	31 (36.5%)
Insured
Yes	73 (83.9%)
No	14 (16.1%)
Race
White	79 (90.8%)
Black	8 (9.2%)
Ethnicity
Latino	4 (4.6%)
Non-Latino	83 (95.4%)
Geographic region New England
New England	69 (79.3%)
Mid-Atlantic/South	2 (2.3%)
South-west/West	16 (18.4%)
Cancer type
Lung cancer	28 (32.2%)
GI cancer	17 (19.5%)
Other	42 (48.3%)

Note. Sample size = 87; missing data present in some demographic variables; reported percentages based on available data; SD = standard deviation. Participating sites for the Coping with Cancer II study included: Weill Cornell Medicine Meyer Cancer Center (New York, NY); Memorial Sloan Kettering Cancer Center (New York, NY); Dana-Farber/Harvard Cancer Center (Brigham and Women’s Hospital, Dana-Farber Cancer Institute, and Massachusetts General Hospital, Boston, MA); Yale Cancer Center (New Haven, CT); Virginia Commonwealth University Massey Cancer Center (Richmond, VA); Simmons Comprehensive Cancer Center (Dallas, TX); Parkland Hospital (Dallas, TX); University of New Mexico Cancer Center (Albuquerque, NM); and Pomona Valley Hospital Medical Center (Pomona, CA).

Patient eligibility criteria included: Black or White race; age 21 years or older; locally advanced and/or metastatic gastrointestinal, lung, or gynecologic cancer; and disease progression after one or more chemotherapy regimens. Patients were excluded if they had cognitive impairment, were too weak to participate in study interviews, or if they were receiving hospice or palliative care.

The present study utilized data from participant surveys conducted before (pre) and after (post) an oncology visit to discuss cancer restaging scan results, and again at a follow-up assessment. The median time elapsed between pre and post assessment was 43.5 days (interquartile range: 17.5 to 70) and between post and follow-up assessment was 43 days (interquartile range: 29.5 to 103). The analytic sample consisted of 87 patients who had complete data for the two primary variables at their repeat assessments: prognostic understanding at pre and post assessment; and psychological well-being at pre, post, and follow-up assessment. The analytic sample showed no difference from those patients enrolled in the parent study but not included in the present analyses, in terms of age or marital status; however, consistent with general trends in research retention,¹⁴ the analytic sample had higher education (p < .001) and a greater proportion of women (p = .05), insured patients (p < .05), whites (p < .01), and non-Latinos (p < .01). These demographic variables were therefore considered in the primary analyses as potential confounds.

At study entry, patient demographics and disease characteristics were recorded. Patient assessments were conducted via structured interviews carried out by trained interviewers. PU were measured at pre and post visit assessment in the same manner used in an earlier analysis of these data examining PU as an outcome of clinical communication at the end-of-life.¹⁵ Four aspects of PU were assessed: 1) terminal illness acknowledgment (TIA), 2) recognition of incurable disease status, 3) knowledge of advanced stage of disease, and 4) life expectancy. Items and response options are listed below. The responses were coded as a 1 indicating accurate understanding, or a 0 indicating inaccurate understanding, based on data demonstrating the studied sample had a median life-expectancy of less than 6 months from assessment.⁴ The four items were summed to create a composite score (possible range 0 to 4), and a difference score reflecting change in PU was computed by subtracting pre-score from post-score, with higher change scores indicating more accurate PU change.

The following are the items and response options used to asses each PU aspect. TIA was examined using the item, “How would you describe your current health status?”, and participants responded, “relatively healthy” (0), “relatively healthy and terminally ill” (1), “seriously ill but not terminally ill” (0), “seriously ill and terminally ill” (1), and “don’t know” (0). Recognition of incurable disease status was examined using the item, “Which of the following best represents what your oncology providers have told you about a cure for your cancer?,” and participants responded, “my cancer will be cured” (0), “my cancer may be cured if treatments are successful” (0), “my cancer cannot be cured but we will try to control the cancer with treatment” (1), “my cancer cannot be cured and I am not able to have any further cancer treatment” (1), and “don’t know” (0). Knowledge of stage of disease was assessed by asking participants, “What stage is your cancer?”. Response options included, “no evidence of cancer” (0), “early stage of cancer” (0), “middle stage of cancer” (0), “late stage of cancer” (1), “end stage of cancer” (1), and “don’t know” (0). Life expectancy was assessed with the prompt, “Many patients have thoughts about how having cancer might affect their life expectancy, either based on what their doctors have told them, what they have read, or just their own sense about how long they might live with cancer. When you think about this, do you think in terms of.” Patients indicated if it was “months” (1), “years” (0), or don’t know”.

Psychological well-being was assessed at pre, post, and follow-up, using the 4-item Psychological Symptoms subscale (gauges depressed mood, worry, sadness, and fear) of the McGill Quality of Life Questionnaire.^16,17 The items were rated on a 10-point Likert scale with higher scores indicating greater severity of symptoms. The items were averaged after reverse scoring, and a difference score was computed reflecting change in psychological well-being from pre to post and post to follow-up. Positive changes scores reflected improving, while negative change scores reflected worsening, psychological well-being.

Prognostic discussion and scan results were assessed and examined as potential confounds. Prognostic discussion was assessed at post-scan assessment, where patients reported whether they had a prognostic discussion during the past visit (yes, 1; no, 0). At post-scan, patients also reported on whether they were told that the scan results showed the cancer to be “worse” (1), “better” (0), “same” (0), or “other” (0).

Analytic Plan

Descriptive statistics were used to summarize sample characteristics and study variables. Pearson correlations examined bivariate associations between PU change and psychological well-being change. Regression analyses examined these associations while controlling for confounding factors. Such confounding factors included those demographic variables reported in a recent systematic review as being associated with prognostic understanding:¹⁸ age, sex, education, race, and ethnicity. Finally, to examine if psychological well-being change scores differed between patients showing different types of PU change one-way ANOVA was used.

Results

Patients were on average approximately 61 years old (SD = 8.83) and had 15 years of education (SD = 3.12; see Table 1). Approximately 74% were women, 64% were married, 91% were white, and 95% were non-Latino. Eighteen patients (20.7%) reported having a prognostic discussion during the last visit, and 21 patients (24.1%) reported receiving scan results showing cancer progression.

The average psychological well-being of the sample remained stable, showing no significant change from pre (M = 7.11, SD = 2.38), to post (M = 7.18, SD = 2.20), to follow-up (M = 7.14, SD = 2.77; F = .05, p = .95). Mean psychological well-being change scores from pre to post was .08 (SD = 2.09), and post to follow-up was −.04 (SD = 2.56).

Mean PU scores at pre and post assessment were 1.67 (SD = 1.19) and 1.74 (SD = 1.16), respectively. Mean PU change scores from pre to post assessment was .07 (SD = .74). PU change scores ranged from −2 to +2, with some patients showing improvements in understanding, others showing unchanged understanding, and yet others showing worse understanding. Two patients (2.3%) showed a change of −2; 12 patients (13.8%) showed a change of −1; 54 patients (62.1%) showed no change; 16 patients (18.4%) showed a change of 1; and 3 patients (3.4%) showed a change of 2.

Primary analyses

Changes toward more accurate PU (i.e., an improvement in accurate PU) was associated with a corresponding decline in psychological well-being (i.e., pre to post; r = −0.33, p < .01), but thereafter was associated with subsequent improvements in psychological well-being (i.e., post to follow-up; r = .40, p < .001). Controlling for potential confounds including relevant demographic factors (age, sex, education, race, ethnicity), prognostic discussion, scan results, and time elapsed between assessments, did not change the pattern of results found in these bivariate analyses: changes toward more accurate PU continued to be associated with a corresponding decline in psychological well-being (ß = −0.36, p < .01), and thereafter associated with subsequent improvements in well-being (ß = .42, p < .01; n = 80 due to missing data in control variables).

To further probe the association, patients were grouped based on the type of change shown in PU. Three groups were created: improved PU group (i.e., change score of +2 and +1; n = 19), no change PU group (change score of 0; n = 54), and worse PU group (change score of −1 and −2; n = 14). Testing if psychological well-being change differed across groups approached significance for pre-to-post change scores (F = 3.07, p = .05), and was significant for post-to-follow-up change scores (F = 6.54, p < .01). Plotting psychological well-being by group exhibited divergent patterns of well-being (see Figure 1). Among the improved PU groups, well-being initially decreased from 7.03 (SD = 2.23) to 6.30 (SD = 1.80), but subsequently increased to 7.63 (SD = 2.08). In contrast, the worse PU group’s well-being initially increased from 6.30 (SD = 2.62) to 7.36 (SD = 2.04), but subsequently decreased to 5.63 (SD = 3.60). Interestingly, the no change PU group showed relatively more unchanged, stable psychological well-being [7.34 (2.37) to 7.45 (2.32), to 7.36 (2.66)].

Figure 1.

Psychological Well-Being Across Pre, Post, and Follow-Up, for Patients Showing Different Types of Prognostic Understanding Change. Error bars represent standard errors.

In a sensitivity analysis, we examined associations between change in each individual item used to assess PU and psychological well-being change. Overall, the patterns found for the individual PU items largely reflected the pattern found for the total PU score (i.e., changes toward more accurate PU was associated with a corresponding decline in psychological well-being but thereafter was associated with subsequent improvements in psychological well-being): knowledge of advanced stage of disease (pre to post, r = −.18, p = .09; post to follow up, r = .31, p < .01), life expectancy (pre to post, r = −.25, p = .02; post to follow up, r = .16, p = .14), recognition of incurable disease status (pre to post, r = −.26, p = .02; post to follow up, r = .16, p = .13), and TIA (pre to post, r = .01, p = .91; post to follow up, r = .14, p = .21).

Discussion

This study examined if improvements in the accuracy of prognostic understanding among patients with advanced, life-limiting cancer is associated with decrements in psychological well-being that persists over time.^1,15,19 Results showed that patients who became more accurate in their prognostic understanding had initial decrements to psychological well-being (i.e., more fear, sadness, worry, and depressed mood), even after controlling for confounds such as demographics, prognostic discussion, and scan results.^7,8 This decline may partly explain the hesitation often displayed by clinicians for having prognostic discussions — clinicians likely notice among patients the increased fear and sadness that comes with recognizing the life-limiting nature of prognosis.^1,3

However, the results showed that the decrements in psychological well-being did not persist over time, and patients appeared to rebound.^9,10 PU change, although associated with worse well-being change at the corresponding time, was subsequently associated with better well-being change. This pattern suggests that on average, the unfavorable effects on patients’ psychological well-being may not last, and clinicians may not need to be concerned about causing lasting damage to patient psychological well-being.

It is noteworthy that calls for clinicians to engage in more prognostic discussions often highlight as a talking point, that helping patients understand their prognosis will not unfavorably impact psychological well-being.^1,3 The present results however suggest that that may not fully capture the dynamic impact of gaining prognostic understanding on patients’ psychological well-being. Instead, specifically noting that patients may initially feel worse, but are able to regain prior levels of psychological well-being, may be more accurate and consonant with what clinicians observe in their patients.¹³

In fact, the pattern shown by patients of initial decrements in psychological well-being followed by a return to baseline, is similar to normative trajectories of emotional response found following other major negative experiences (e.g., trauma, loss).^10–12 Additionally, the psychological literature has also highlighted the value of negative emotions in coping with stressors, showing that attending to and experiencing negative emotions can help habituate to the stressor, and inform and direct more beneficial coping strategies.^20,21 Therefore, the initial decrements to psychological well-being (such as sadness, worry, fear) following increased prognostic understanding may not be something to be warded off, but perhaps best seen as a normative coping process, that may ultimately facilitate better prognostic awareness, decision-making, end-of-life planning, and well-being.^22,23 Clinicians may support patients during initial worse psychological distress by inquiring how patients feel, attending to and validating their emotions, and expressing ongoing support, no matter what. Such clinician support may change the pattern of distress experienced by patients for the better. Future research examining prognostic discussions would therefore be apt to examine how differences in clinician support could be associated with less distress immediately and over time.

Analyses examining patients showing different types of prognostic understanding change further suggested that psychological well-being is closely associated with patients’ expectations of their prognosis. Among patients with unchanged PU, psychological well-being showed minimal change across time. In contrast, among patients with changes in PU — whether it be towards more accurate or inaccurate PU — well-being showed more variability. Interestingly, patients with improved PU showed a divergent pattern of change from patients showing worsened PU. Among patients with improved PU, psychological well-being initially decreased and subsequently returned to baseline. Notably however, among patients with changes towards more unrealistic, inaccurate PU, well-being initially improved, but subsequently declined. This pattern shown in the latter group is consistent with the notion that false expectations/unrealistic hope may only offer temporary respite from emotional distress, eventually leading to worse psychological well-being (as patients are confronted with their progressive illness, sooner or later). Future research with larger samples is needed though, before definite claims can be made about these subgroups, as the number of patients per subgroup was relatively small in the present study.

The limitations of this study include issues related to sample selection and retention, which may limit generalizability. It is possible that patients who chose to participate in this study may be different from those who did not participate in terms of their psychosocial characteristics (e.g., coping abilities; social functioning). Similarly, psychosocial factors may vary between those patients who participated in follow up assessments compared to those who were lost to follow up. If present, such systematic differences in psychosocial variables could limit the generalizability of study findings. Another limitation of the study is its observational design, which leaves open confounding explanations. It is possible that confounds beyond the ones we controlled for may explain the associations. Additionally, it is possible that the direction of causality is counter to what is presumed, with psychological well-being driving changes in prognostic understanding. As emotional states may influence thinking and perceptions (including those of clinical communication), this alternate direction of causality is possible.²⁴ However, this cannot explain the association found between PU change and subsequent psychological well-being change, as PU change preceded well-being change. Also noteworthy as a limitation of this study is the variability in timing of assessments; a uniform assessment schedule with equal time between assessments across all participants may have allowed additional inferences regarding patterns of change in distress across specific durations of time. Finally, low diversity and sample size are also notable limitations of the present study. Diversity is particularly important in the present context as perceptions of prognosis and other end-of-life issues may vary among patients with diverse racial, ethnic, and cultural backgrounds.^25,26 The majority of the analytic sample was white and non-Latino, and therefore, future research examining the generalizability of these findings in more culturally diverse patients is needed.

The present study findings indicate several directions for future research. It would be important to examine how patterns of distress associated with prognostic understanding change varies across different racial and ethnic groups and among patients with varying coping strategies. Future research should also examine how prognostic understanding and patterns of distress may be different across patients with different available treatment options, including treatments associated with much hope and excitement (e.g., immunotherapies) — it is conceivable that patterns of distress may be different among those receiving treatments with much associated hype and promise, with less distress initially and greater subsequent disappointment should treatments fail.

The strengths of the present study include its longitudinal design, and examination of changes in prognostic understanding and psychological well-being and their associations. To our knowledge, this is the first study to demonstrate that PU change is associated with a pattern of initial decrements to psychological well-being followed by a rebound. This pattern — which matches the normative psychological coping process to negative experiences^10–12 — provides a context for clinicians in working with their patients. Patients with life-limiting illnesses experiencing more sadness or worry following improved understanding of their prognosis, may not be something to be averted, and may not need to be a deterrent to having prognostic discussions. Rather, this lowered psychological well-being may be part of a normative coping process, from which patients may rebound. Clinicians attending to patient emotion could likely assist patients in managing their emotions as they face, and come to terms with, their prognosis.