Introduction

Accurate illness understanding is crucial for patients with advanced cancer to make decisions that align with their preferences and values. In fact, it has been reported that accurate prognostic understanding is associated with higher quality of advance care planning and end-of-life care [1-6]. In addition, previous studies on early palliative care showed that informing patients with an integrated approach would lead to better quality of life as well as better prognostic understanding [7, 8]. Therefore, health care providers make efforts to promote patients’ accurate understanding, which could result in improvement of their quality of life. However, patients with advanced cancer tend to have inaccurate prognostic understanding and overestimate their life expectancy. Previous studies have reported that proportions of patients with inaccurate prognostic understanding varied between 20% and 75% [9-15]. Because most of these studies were cross-sectional, where the assessment was conducted after some time from the time of diagnosis, it remains unclear how the prevalence of inaccurate prognostic understanding in patients with advanced cancer changes from diagnosis over time.

We previously reported the prevalence of inaccurate prognostic understanding at the time of diagnosis and its associated factors in patients with newly diagnosed advanced lung cancer [14] In this report, accurate prognostic understanding was associated with female and employment status among baseline patient characteristics. Several studies have also reported factors associated with accurate prognostic understanding in patients with advanced cancer such as being female, older age, and being educated [5, 9, 16]. However, because of the lack of longitudinal studies, little is known about whether the associated factors differ depending on time from cancer diagnosis.

In our study, we found that accurate prognostic understanding was strongly associated with worse emotional well-being at the time of diagnosis [14]. Several studies also suggested the association between accurate prognostic understanding and psychological distress [9, 17-19]. Recently, George et al. reported that becoming more accurate in prognostic understanding was associated with initial decline and subsequent recovery in psychological well-being in patients with advanced cancer with a life expectancy of less than 6 months [20]. However, there is currently no study reporting the association between accurate prognostic understanding and psychological distress over time, and it is still unclear whether psychological distress associated with accurate prognostic understanding remains long or not, especially after the time of cancer diagnosis. As Burns et al. reported, illness understanding in patients with advanced cancer can fluctuate during the disease trajectory [21]. Therefore, we hypothesized that not just being accurate in prognostic understanding at one moment but being consistently accurate during the trajectory was associated with psychological distress via patients’ full acceptance of the incurable nature of the disease.

In this multicenter study, we surveyed the prevalence of inaccurate prognostic understanding in patients with advanced lung cancer longitudinally starting from cancer diagnosis. We examined associated factors with prognostic understanding over time, mainly focusing on whether maintaining accurate prognostic understanding was associated with their psychological distress over time.

Patients and Methods Study Design and Participants

This longitudinal study is a part of a larger study, which we initially planned, and we previously reported our initial research focusing on prognostic understanding only at the time of diagnosis [14]. Briefly, patients were prospectively recruited upon disclosure of their diagnosis of advanced lung cancer at 16 hospitals in Japan between December 2013 and March 2016. Patients were eligible if they were (a) diagnosed with clinical stage IIIB or IV lung cancer (the seventh edition of lung cancer stage classification), (b) 20 years of age or older, and (c) able to write and comprehend Japanese. Exclusion criteria were the following: (a) had significant cognitive impairment, or (b) had already received anticancer treatments, including chemotherapy, radiotherapy, surgery, or immunotherapy. Participants were asked to complete the questionnaires promptly after submitting written informed consent, 3 months later, 6 months later, 12 months later, and 24 months later. Patients’ clinical and sociodemographic characteristics were collected through reports from treating physicians. This study was approved by the institutional review board of all the participating hospitals, and all participants submitted written informed consent. We approached the patients as soon as possible after disclosure and explanation of their cancers, considering the emotional state of the patients. Participants were recruited in principle within 2 weeks after disclosure.

Self-Reported Outcomes

To assess the perceptions of prognostic understanding, we used self-report questions that were adopted from those used in previous studies. Patients were asked to answer the question, “What is your cancer like?” by selecting either one of the following two options: “completely curable” or “complete cure is difficult.” Because advanced lung cancer is still an incurable cancer, patients who selected terms “complete cure is difficult” were categorized into accurate prognostic understanding and those who selected terms “completely curable” were categorized into inaccurate understanding. By applying this question without asking just whether their cancer was curable (yes or no) like previous studies [10], we aimed to judge participants’ prognostic understanding more accurately.

Health-related quality of life of patients was measured with the Functional Assessment of Cancer Therapy–Lung scale, which is composed of physical well-being, social well-being, emotional well-being, functional well-being, and lung cancer symptom burden (lung cancer subscale) [22]. Higher scores indicate better quality of life and lower symptom burden.

To assess psychological distress for patients, we asked participants to complete the Hospital Anxiety and Depression Scale (HADS), which is a 14-item self-report questionnaire that contains two subscales measuring anxiety (HADS-A) and depression (HADS-D) [23, 24].

Statistical Analysis

After descriptive analyses, bivariate analyses were conducted to compare variables of the patients with consistently accurate understanding with those with inaccurate or inconsistent understanding. We defined patients with inconsistent understanding as those who had inaccurate understanding about prognosis at least once throughout the time points observed. Fisher's exact tests were used for categorical variables. Wilcoxon signed-rank tests were used for continuous and ordinal variables. Multivariate logistic regression analyses were conducted to assess variables associated with consistently accurate understanding. We selected the variables that were found significant in the univariate analyses at 3 months: sex, Lung Cancer subscale score, and Emotional Well-Being subscale score. The cases with missing values were excluded on an analysis-by-analysis or pair-by-pair basis. Additionally, considering the missing data, we performed mixed-effects model for repeated measures (MMRM) analyses for the proportions of accurate prognostic understanding and the associations between being consistently accurate and psychological distress. An unstructured covariance structure was used to model the within-patient errors. Kenward-Roger correction was used to estimate denominator degrees of freedom. All p values were two-sided, and p < .05 was considered statistically significant. Analyses were conducted using JMP 13 software (SAS Institute, Cary, NC) and GraphPad Prism 6 (GraphPad Software, San Diego, CA).

Result Patient Characteristics

The flow of the participants is shown in supplemental online Figure 1. Out of 248 patients enrolled, 225 reported their perceptions of prognosis. Patients’ characteristics are shown in Table 1. The median age is 70 years at baseline. Most patients were male, smokers, and married and had stage IV advanced lung cancer with Eastern Cooperative Oncology Group performance status of 0 or 1. Chemotherapy was scheduled for most patients (86.0%). Only 2.2% of patients were scheduled to receive supportive care only. Characteristics of the patients who were able to be followed at 3 and 6 months after diagnosis were similar to those at the time of diagnosis. The majority of these patients were treated with chemotherapy at the later time points and those who received best supportive care increased up to 8.7% at 6 months after diagnosis. When we examined the differences of variables between responders and nonresponders/withdrawals at 3 months after diagnosis, poorer performance status, poorer quality of life, and higher HADS-A scores were significantly associated with nonresponders/withdrawals (supplemental online Table 1).

Table 1. Sociodemographic and clinical characteristics Variable At diagnosis (n = 225); n (%) After 3 months (n = 147); n (%) After 6 months (n = 120); n (%) Age, yr Median, range 70 (25–92) 70 (34–87) 71 (34–88) Sex Male 160 (71.4) 106 (72.6) 80 (67.2) Female 64 (28.6) 40 (27.4) 39 (32.8) Smoking status Never 51 (23.0) 34 (23.4) 34 (28.8) Current/former 171 (77.0) 111 (76.6) 84 (71.2) Employment status Employed 60 (27.8) 28 (19.7) 24 (20.9) Unemployed 156 (72.2) 114 (80.3) 91 (79.1) Marital status Married 145 (67.4) 98 (71.0) 80 (70.2) Single/widowed/divorced 70 (32.6) 40(29.0) 34 (29.8) Household size 1 person (living alone) 37 (17.0) 18 (12.8) 17 (14.5) 2 or more 181 (83.0) 123 (87.2) 100 (85.5) ECOG performance status 0 106 (47.5) 54 (37.5) 39 (34.8) 1 82 (36.8) 69 (47.9) 56 (50.0) 2 23 (10.3) 16 (11.1) 12 (10.7) 3 10 (4.5) 2 (1.4) 4 (3.6) 4 2 (0.9) 3(2.1) 1 (0.9) Clinical stage (the 7th edition) IIIB 40 (17.9) 27 (18.5) 22 (18.5) IV 184 (82.1) 119 (81.5) 97 (81.5) Planned or ongoing treatment Chemotherapy 190 (86.0) 131 (91.0) 87(75.7) Radiation 2 (0.9) — 5 (4.3) Combined chemoradiation 24 (10.9) 4 (2.8) 2 (1.7) Best supportive care 5 (2.2) 2 (1.3) 10 (8.7) Treatment-free period — 7 (4.9) 11 (9.6) FACT-L total score, median (IQR) 80 (68.7–91.2) 84 (73.5–96.9) 83.9 (73.2–93.0) Physical Well-Being subscale score, median (IQR) 19.6 (14.0–23.2) 18.0 (14.0–23.0) 20.0 (17.0–24.0) Social Well-Being subscale score, median (IQR) 19.0 (14.0–22.4) 17.5 (14.0–21.8) 18.7 (14.0–21.0) Emotional Well-Being subscale score, median (IQR) 15.0 (11.0–19.0) 19.8 (14.0–23.3) 17.0 (13.0–20.0) Functional Well-Being subscale score, median (IQR) 14.5 (11.0–21.0) 16.0 (11.0–20.0) 16.2 (12.0–20.0) Lung cancer subscale score, median (IQR) 13.0 (11.0–15.3) 14.0 (13.0–16.0) 14.0 (12.3–16.0) HADS-Anxiety, median (IQR) 5 (3–8) 5 (2–7) 5 (3–7) HADS-Depression, median (IQR) 6 (3–10) 6 (3–9) 5 (3–9) Abbreviations: ECOG, Eastern Cooperative Oncology Group; FACT-G, Functional Assessment of Cancer Therapy-General; FACT-L, Functional Assessment of Cancer Therapy–Lung; FACT-L TOI, Functional Assessment of Cancer Therapy–Lung Trial Outcome Index; HADS-A, the Hospital Anxiety and Depression Scale-Anxiety; HADS-D, the Hospital Anxiety and Depression Scale-Depression; IQR, interquartile range. Prognostic Understanding Over Time

Not only at the time of diagnosis but also at the later time points, approximately one fifth of patients reported that their cancer was curable despite its incurable stages, as shown in Table 2 and supplemental online Table 2. How their understanding changed over time is shown in Figure 1. Three months after diagnosis, out of 114 patients who accurately reported that complete cure was difficult at diagnosis, 103 patients remained accurate in prognostic understanding, whereas 11 patients became inaccurate. Out of 26 patients who inaccurately reported that their cancer was curable at diagnosis, 15 patients remained inaccurate, whereas 11 patients became accurate. At 6 months after diagnosis, of 99 patients whose longitudinal reports about prognostic understanding at diagnosis, 3 months and 6 months were all available, 68 (68.7%) patients were consistently accurate in prognostic understanding at all the three time points, whereas 10 (10.1%) patients were consistently inaccurate. Out of the remaining 21 patients whose prognostic understanding was inconsistent, 10 patients had become accurate and 11 had become inaccurate at 6 months after diagnosis.

Table 2. Perceptions of prognosis over time Time point “Completely curable” (inaccurate), n (%) “Complete cure is difficult” (accurate), n (%) At diagnosis (n = 225) 48 (21.3) 177 (78.7) After 3 mo (n = 147) 27 (18.4) 120 (81.6) After 6 mo (n = 120) 24 (20) 96 (80) After 12 mo (n = 75) 16 (21.3) 59 (78.7) After 24 mo (n = 32) 4 (12.5) 28 (87.5)

Prognostic understanding in patients with advanced lung cancer over time.

Associations Between Being Consistently Accurate in Prognostic Understanding and Psychological Distress

When we examined the associations between changes of prognostic understanding from the time of diagnosis and psychological distress at 3 months after diagnosis (supplemental online Table 3), patients who remained accurate in prognostic understanding showed the lowest Emotional Well-Being subscale score even compared with those who became accurate after 3 months. To address the question whether psychological distress associated with accurate prognostic understanding remains when the understanding was consistently accurate, we examined the association between being consistently accurate in prognostic understanding and psychological distress. As shown in Figure 2, we found that being consistently accurate was significantly associated with the lower Emotional Well-Being subscale score at both 3 and 6 months after diagnosis (p = .010 and p = .014, respectively). Furthermore, when we performed longitudinal analysis using MMRM, being consistently accurate from the time of diagnosis to 6 months after diagnosis was significantly associated with the lower Emotional Well-Being subscale score at all the time points (supplemental online Table 4). To see whether being consistently accurate in prognostic understanding is even associated with increasing psychological distress, we examined the changes in the Emotional Well-Being subscale score from the time of diagnosis (supplemental online Table 5) and found no significant difference in the changes between the groups. It is suggested that psychological distress at diagnosis remains in patients whose prognostic understanding is consistently accurate compared with patients whose prognostic understanding is inaccurate or inconsistent although we observed some improvement of emotional well-being at 3 months after diagnosis in both groups.

Associations between being consistently accurate in prognostic understanding and Emotional Well-Being subscale score.

Associated Factors with Being Consistently Accurate in Prognostic Understanding

Next, we investigated associations between being consistently accurate in prognostic understanding and the other variables, in univariate analyses (Table 3) and found that female gender and greater lung cancer–specific symptom burden measured by Lung Cancer subscale score were associated with being consistently accurate at 3 months after diagnosis (p = .032 and p = .004, respectively). When we examined whether physicians’ disclosures were associated with patients’ consistently accurate prognostic understanding, patients to whom predictive life expectancy was explained were more likely to be consistently accurate at 3 months after diagnosis (supplemental online Table 6). In multivariate analyses (Table 4), female gender and lower Lung cancer subscale score were significantly associated with being consistently accurate in prognostic understanding at 3 months after diagnosis (p = .008 and p = .005, respectively). At 6 months, the lower Emotional Well-Being subscale score was significantly associated with being consistently accurate in prognostic understanding (p = .006).

Table 3. Associations between being consistently accurate in prognostic understanding and variables Variable 3 months after diagnosis 6 months after diagnosis Inaccurate/Inconsistent (n = 37) Remained accurate (n = 103) p value Inaccurate/Inconsistent (n = 31) Consistently accurate (n = 68) p value Age, yr Median, range 69 (62–76) 71 (63–76) .707 70 (63–76) 71 (63.3–75) .994 Sex Male 32 70 .032 24 48 .628 Female 5 33 7 20 Smoking status Never 7 25 .649 5 18 .310 Current/former 30 77 26 49 Employment status Employed 8 17 .465 6 13 1.000 Unemployed 28 83 24 51 Marital status Married 24 72 .829 21 47 1.000 Single/widowed/divorced 11 28 8 18 Household size 1 person 2 16 .156 2 9 .494 2 or more 34 86 28 58 ECOG PS 0 20 53 .626 11 22 .474 2 2 9 1 8 3 1 3 1 2 4 0 2 0 1 Clinical stage IIIB 4 21 .222 4 15 .411 IV 33 82 — 27 53 — Ongoing treatment Chemotherapy 35 88 .238 23 47 1.000 Radiation 2 2 1 4 Combined chemoradiation 0 2 0 2 Best supportive care 0 7 1 4 Treatment-free period 4 7 FACT-L total score, median (IQR) 91 (77.8–103.7) 82.7 (72.7–96.3) .055 87.8 (3.7–100) 83 (71–93) .209 Physical Well-Being subscale score, median (IQR) 22 (15–25) 18 (14–23) .090 23 (18–26) 20 (16–24) .053 Social Well-Being subscale score, median (IQR) 16.3 (12–21) 17.8 (14–21.8) .291 18 (14–21) 18.7