Nationwide Survey on Family-perceived Experiences of Patients With Cancer of Unknown Primary Site

Kyoko Ishida (  k_ishida@nagoya-u.jp ) Nagoya University Graduate School of Medicine, Department of Nursing, Doctoral Course https://orcid.org/0000-0001-6829-8163 Kazuki SATO Nagoya University, Nursing for Advanced Practice, Department of Integrated Health Sciences Hirokazu KOMATSU Nagoya City University, Department of Hematology and Oncology Tatsuya MORITA Seirei Mikatahara Hospital, Department of Palliative and Supportive Care, Palliative Care Team and Seirei Hospice Tatsuo AKECHI Nagoya City University Graduate School of Medical Science, Department of Psychiatry and Cognitive-Behavioral Medicine Megumi UCHIDA Nagoya City University Graduate School of Medical Sciences, Department of Psychiatry and Cognitive-Behavioral Medicine Kento MASUKAWA Tohoku University Graduate School of Medicine, Department of Palliative Nursing, Health Sciences Naoko IGARASHI Tohoku University Graduate School of Medicine, Department of Palliative Nursing, Health Sciences Yoshiyuki KIZAWA Kobe University Graduate School of Medicine, Department of Palliative Medicine Satoru TSUNETO Kyoto University Graduate School of Medicine, Department of Human Health Sciences Yasuo SHIMA Tsukuba Medical Center Hospital, Department of Palliative Medicine Mitsunori MIYASHITA Tohoku University Graduate School of Medicine, Department of Palliative Nursing, Health Sciences Shoko ANDO Nagoya University Graduate School of Medicine Faculty of Medicine: Nagoya Daigaku Daigakuin Igakukei Kenkyuka Igakubu


Introduction
Cancer of unknown primary site (CUP) is a clinically aggressive disease with early dissemination [1][2], resulting in poor prognosis with a median survival time of 6 months [3]. While research on the medical aspects of CUP has been increasing [4][5][6][7][8], information on the experiences of CUP patients and families is scarce. To date, some qualitative studies demonstrated that CUP patients experience uncertainty and di culty when accepting the CUP diagnosis [9,10] and that CUP patients experience anxiety, disbelief, and frustration in addition to uncertainty from onset to initial treatment [11]. Quantitative studies are limited, and only one large quantitative study compared experiences between CUP and other cancer patients, which revealed that CUP patients were less likely to have understood the explanations of their condition [12]. Furthermore, CUP patients' families may experience considerable burden [13,14], but very little research has focused on the family members of CUP patients [10]. Empirical evidence suggests that family caregivers sometimes become more depressed than patients themselves [15] and that they face continued burden after the patient's death [16,17]. A greater understanding of the experiences of CUP patients' families is valuable for developing an appropriate strategy to care for both CUP patients and their families.
The main purpose of this study was thus to compare the family-perceived burden of CUP patients and those with common cancers (lung, colon, and stomach cancer). The secondary aims were 1) to explore the association between family-perceived burden on family members and the patients' quality of life (QOL) at end-of-life and depression among bereaved family members and 2) to identify factors related to family-perceived burden.

Methods
This study was conducted as a part of the Japan Hospice and Palliative Care Evaluation (J-HOPE) with the aim to evaluate the quality of palliative care services from the perspective of bereaved families [18,19]. This was a cross-sectional nationwide survey using anonymous, self-administered questionnaires. For the present study, data from two surveys, namely J-HOPE 2016 (conducted May-July 2016) and J-HOPE 4 (July-September 2018) [18,19], were combined to achieve the necessary sample size.

Subjects
Potential subjects were bereaved family members of cancer patients who had died at each participating institution between November 2013 and January 2016 (J-HOPE 2016) and between February 2016 and January 2018 (J-HOPE 4). Participating institutions were certi ed inpatient hospices/palliative care units (PCUs), home care services, and acute hospitals belonging to the Japan Hospice and Palliative Care Foundation. The eligibility criteria were as follows: adult bereaved family members of a patient that died of cancer, the deceased patient was aged ≥20 years, and the patient was an inpatient in a PCU for >3 days. The exclusion criteria were as follows: a PCU stay <3 days, inability to identify a family member, treatment-associated death or death in an intensive care unit, suffering serious psychological distress, inability to complete the self-report questionnaire because of health issues, and overall judgment of the healthcare professionals.

Procedures
Questionnaires were sent by mail to identi ed family members directly by each participating institution, and uninterested participants were asked to check the "No Participation" box and return the incomplete questionnaire. Participants were provided detailed information in the form of an enclosed letter of intent about participation being voluntary, absence of medical disadvantage due to nonparticipation, and assurance of complete anonymity. We asked that the respondents be family members who were the main caregivers of the patient and who had the most knowledge about the treatment provided to the patient. Completed questionnaires were retuned by the participants to the research secretariat, and data management at the secretariat was independent of all participating institutions.
Return of a completed questionnaire was de ned as consent for study participation. A reminder was sent to non-responders one month after the questionnaire was sent out.

Measurements
The questionnaire for this study was developed by the authors on the basis of a literature review [9,10,20,21], a previous qualitative study using interviews with 9 CUP patients [11], and extensive discussions among the authors. Content validity was con rmed by a pilot test with three bereaved family members and unanimous agreement among the authors.
Burden on family members associated with cancer testing and diagnosis (family-perceived burden scale: Burden) The Burden scale was developed to assess the family's experiences related to cancer testing and diagnosis and comprised eight items.
Each item was rated on a four-point Likert scale (1: Not at all, 2: Almost never, 3: Sometimes, and 4: Often). Subscale scores were de ned as the mean scores of the included items, and the total score was de ned as the mean of the subscale scores: each was scored on a scale of 1-4, and higher scores indicated greater Burden.

Good death inventory (GDI)
Patients' QOL at end-of-life was assessed using the short version of the GDI, which measures achievement of a good death from the perspective of bereaved family members. The short GDI consists of 18 representative items from each domain of the original version and its validity and reliability have been previously con rmed [22]. Responses were rated on a scale from 1 to 7, with a high total score indicating good death [22]. Patient Health Questionnaire 9 (PHQ-9) Depression was assessed using PHQ-9; the reliability and validity of the Japanese version have been previously con rmed [23][24][25][26].
Each of the nine items explore the extent to which a particular depressive symptom has bothered the respondent in the preceding 2 weeks. Responses are rated on a scale from 0 to 3, with total scores ranging from 0 to 27 [23,24].

Participant characteristics
The patients' age, sex, and primary cancer site from their medical records were recorded by the physician responsible for patient treatment at each of the participating institutions. We also asked bereaved family members for their age, sex, relationship to the patient, and educational background as part of the questionnaire.

Background of the diagnosis of CUP patients
Only the bereaved family members of CUP patients were asked for details regarding patients' chief complaint at the time of the rst hospital visit, length of time from the awareness of symptom to diagnosis, number of hospitals visited by patients until CUP diagnosis, type of hospital at which the patient received a con rmed CUP diagnosis, availability of a second opinion consultation, and history of any cancer other than CUP.

Sample size calculation
We initially calculated a sample size of 60 cases in 2 groups based on an effect size (ES) of 0.5, α of 0.05, and power of 0.8. From the rst survey (J-HOPE 2016), we obtained 39 responses. We added the J-HOPE4 survey assuming that the impact of medical changes was minimal during the study period (2016-2018).

Statistical analyses
For the background variables and item responses, descriptive statistics were calculated for the measured items in the CUP and common cancer groups, and between-group differences were examined using Student's t-test and one-way analysis of variance (ANOVA) for continuous variables and chi-squared test for categorical variables. Exploratory factor analysis was performed using the maximumlikelihood method. The repeated Promax rotation solution was used to con rm the identity of the underlying structure of the Burden scale.
The total and subscale scores were compared using multiple regression analysis to adjust for covariates between the groups, and Cohen's d was used as ES. Multiple regression analysis was performed using the GDI and PHQ-9 scores as dependent variables and the Burden scores and background variables as independent variables for a subset of the CUP data. Furthermore, the determinants relevant to each subscale of Burden were analyzed for a subset of CUP data. Univariate analyses were performed with chi-squared analyses, one-way ANOVA, and t-tests, where appropriate. For multivariate analysis, multiple regression analysis was performed using items that had P < 0.10 in univariate analysis. Variable selection was performed using the variable reduction method.
A P value < 0.05 was considered statistically signi cant. All analyses were performed using SPSS version 27.0 (IBM Japan Institute, Tokyo, Japan).

Results
The participating institutions were PCUs (258), home hospice services (14), and acute hospitals (14). The original dataset comprised a total of 27,591 respondents ( Figure 1). For the families of CUP patients, 237 questionnaires were sent, 115 were returned (response rate, 48.5%), and 97 were ultimately analyzed. For the families of non-CUP patients, 22,551 questionnaires were sent and 14,834 were returned (response rate, 65.8%). Of these, 1,877 responses corresponded to the families of common cancer (lung, colon, and stomach) patients, and 717 were ultimately analyzed. Table 1 shows the characteristics of patients and their bereaved family members. Variables that were signi cantly different between the groups were patient gender (P = 0.006) and number of days elapsed since death (P = 0.005). Frequency distribution and factor analysis of the Burden scale Table 2 shows the percentage of family members who answered" sometimes" or "often" to each Burden item. Among the families of CUP patients, the questions that received a "sometimes" or "often" response 50% of the time or more were patient exhaustion due to the large number of evaluations (73.9%), anxiety associated with the prolonged diagnosis period (70.2%), and uncertainty about the future progression of the cancer (56.3%). Exploratory factor analysis identi ed three underlying structures of the Burden scale ( Table 2): 1) Factor-1: "Di culty in accessing specialist doctors" (specialist subscale, α = 0.844), comprising three items; 2) Factor-2: "Uncertainty due to lack of information about the disease" (uncertainty subscale, α = 0.821), consisting three items; and 3) Factor-3: "Psychological burden associated with extensive testing and prolonged diagnosis period" (prolonged diagnosis subscale, α = 0.703), consisting two items. Cronbach's alpha across the eight items was 0.901. Table 3 shows the difference in the mean family-perceived Burden score associated with cancer testing and diagnosis between the groups, after adjusting for covariates, i.e., patient age, gender, and relationship to the patient. The scores for all three factors of Burden were signi cantly higher for the CUP group than the common cancers group: specialist subscale (ES = 0.30, AdjP = 0.010), uncertainty subscale (ES = 0.66, AdjP < 0.0001), and prolonged diagnosis subscale (ES = 0.69, AdjP < 0.0001).  Table 4 shows the correlation of GDI or PHQ-9 score with the three factors of Burden, after adjusting for patient age, gender, and relationship to the patient (only in the case of PHQ-9, the period after the patient's death was added). There was a signi cant association of all three factors of Burden with GDI and PHQ-9. Strongly associated factors included uncertainty subscale with GDI (Std β = −0.41, AdjP < 0.0001), specialist subscale with PHQ (Std β = 0.29, AdjP = 0.012), and prolonged diagnosis subscale with PHQ (Std β = 0.29, AdjP = 0.011). Factors related to family-perceived Burden associated with cancer testing and diagnosis (CUP data only)

Comparison of Burden between family members of CUP patients and those of common cancer patients
Factors associated with Burden are summarized in Table 5. Multiple regression analysis identi ed that the independent factors for total and subscale scores for Burden included relationship to the patient (spouse had higher Burden), duration of diagnosis (≥1 month was associated with a higher Burden), and second opinion consultation (families who reported a second opinion consultation had a higher Burden).

Discussion
To the best of our knowledge, this is the rst study comparing family experiences between CUP and common cancer patients. Familyperceived burden associated with cancer examination and diagnosis was signi cantly higher in the families of CUP patients than in those of common cancer patients, and their perceived burden was signi cantly associated with patients' QOL at end-of-life and family depression. In addition, a diagnosis that required ≥1 month and receiving a second opinion consultation were signi cantly associated with burden.
Family-perceived burden comprises three components, namely specialist access, uncertainty, and prolonged diagnosis. All three of these subscales were signi cantly higher in the families of CUP patients than in those of common cancer patients. This difference may be because CUP usually presents with atypical symptoms; thus, physicians are often unfamiliar with diagnosing CUP [6]. This is analogous to rare cancers requiring a longer time for diagnosis than common cancers [27,28]. In particular, the uncertainty and prolonged diagnosis subscales showed signi cant and large differences between CUP and common cancers. These ndings con rm the results of previous qualitative studies [9][10][11], suggesting that CUP patients undergo prolonged examinations with high uncertainty.
In CUP, a comprehensive strategy to establish access to specialists, obtain an early diagnosis, and reduce the uncertainty experienced by patients and their families is urgently required.
Another important nding is that a higher Burden score was signi cantly correlated with lower patient QOL at end-of-life (GDI) and higher depression among bereaved family members. CUP patients are often symptomatic before the de nitive diagnosis [29], are typically diagnosed during an unplanned hospital admission, and have very poor survival [30]. Owing to these characteristics, the family-perceived burden during the diagnostic process affects not only the CUP patients at the end-of-life but also their bereaved families. Establishing a comprehensive care strategy for CUP patients and their families may improve patients' QOL at end-of-life and post-bereavement depression.
A notable result of our analysis is the identi cation of a reasonable diagnosis period from the perspective of a healthcare recipient as being "within 1 month"; this is in agreement with the guidelines [31] for the appropriate diagnosis of CUP, which is typically associated with a prolonged diagnostic period [32][33][34]. Clarifying speci c periods in diagnosing CUP may be helpful for clinicians during the initial diagnostic phase.
Of note, the difference in the specialist subscale was relatively lower than that in the other two subscales. This may be due to the presence of a national health insurance system covered by the government that is available to the Japanese population [35]. In the Japanese healthcare system, all patients can freely choose medical institutions and receive medical treatment anytime within the upper limits of expenses (~1000 USD per month); therefore, it is rare for anyone to have no access to healthcare at all [35,36]. Additionally, most Japanese, particularly the elderly, prefer the traditional culture of "leave it to the doctor" [37]. Although there are indications of inequalities in access to services in CUP [32][33][34], it is possible that this cultural background in Japan is re ected in our results.

Strengths and limitations
This study has multiple strengths, including the use of a nationwide sample, use of common cancer patients as a reference group to make comparisons, and use of validated measures of good death (GDI). However, this study also has several limitations. First, the response rate was moderate (48.5%) and we could not compare the backgrounds between responding and nonresponding families; thus, there may be some selection bias. Second, although the responses of the 97 bereaved families of CUP patients were from nationwide surveys, they are not fully representative because the patients were mainly recruited from a convenient sample of specialized palliative care services. Third, recall bias may have occurred owing to retrospective assessment by bereaved family members. Fourth, although the outcome measures had good psychometric properties, including factor validity and internal consistency, reliability and validity was not formally tested.

Conclusions
Family-perceived burden associated with cancer testing and diagnosis was higher among the families of CUP patients than those with common cancers, and the higher burden was correlated with lower patient QOL at end-of-life and higher depression among bereaved family members. Burden comprised three factors: specialist access, uncertainty, and prolonged diagnosis; it was signi cantly associated with the relationship to the patient (spouse), prolonged diagnosis (>1 month), and second opinion consultation. CUP patients and their families should be cared for from the initial stages to establish access to specialists, obtain an early diagnosis, and reduce uncertainty.

Declarations
Funding This study was funded by the Japan Hospice Palliative Care Foundation (No grant number).
Competing Interests Figure 1 Legend not included with this version.