We conducted a cross-sectional study from December 2013 to August 2014. We recruited family caregivers of patients with terminal cancer at three hospitals in Tokyo, Japan. We opted to conduct our survey in the gastroenterology ward in consideration of the high proportion (>50%) of cancer deaths from digestive cancer in Japan [8]. The following criteria were observed in the recruitment: a) having a relative who was a terminal cancer patient aged >20 years; b) being perceived by the patient as the primary caregiver (or perceived by the family as the primary caregiver if the patient was unable to answer because of poor health); c) aged >20 years; d) judged by the head nurse as physically and mentally capable of participating in the survey.

The head nurse in each ward chose the family caregivers who met the criteria. The nurse provided information to the admitted patients regarding the research objectives. For patients and family caregivers whom the researchers were allowed to see, the visit was conducted in the patient’s hospital room; the researcher offered written and verbal explanation of the study. Written consent was obtained from each patient and family caregiver. Coupon books worth JPY 500 were given to the family caregivers. Family caregivers were asked to return the anonymous questionnaire either by mail or by using a mailing box in the ward.

All procedures performed in studies involving human participants were approved by the institutional review boards of University of Tokyo, Kyoundo hospital, and Cancer Institute hospital (approval numbers 10297, 10211, 101013); and in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

We measured the QOL of family caregivers using the Japanese version of the Caregiver Quality of Life Index-Cancer (CQOLC) [9] [10]. The CQOLC contains 21 items that comprise four subscales: psychological burden, positive emotions, financial burden, and disruption of daily living. A higher total score reflects a higher QOL.

We used the Japanese version of the Family Relationship Index (FRI) to measure family functioning [11] [12] [13]. The FRI is a well-validated measure of an individual’s perception of family functioning. It contains 12 items that originated in the short form of the family environment scale [14]. The items comprise three aspects of family functioning: cohesion, expressiveness, and conflict. Each item is marked as true or false, with a total score ranging from 0 to 12. A total score of 9 or higher indicates good family functioning.

Demographic data of patients and caregivers were shown as n (%) or mean (standard deviation [SD] and range). For assessing family perceptions among patients and their families, the FRI and its three subscales were chosen to form the clusters. Following Ozono et al. [4], we used a hierarchical cluster analytic approach employing Ward’s method. Total CQOLC and FRI score were compared between clusters using analysis of variance (ANOVA) and following comparisons of a cluster with other clusters. We set a p-value less than 0.05 as statistically significant. Cluster analyses were performed using SPSS 23, and the other analyses were performed using SAS 9.4.

Table 1 summarizes the characteristics of the 51 caregivers and patients. In this study, caregivers’ mean age was 59.7 (SD = 15.0), and 43 caregivers (84%) were female. The patients’ mean age was 66.6 (SD = 11.5), and 37 patients (73%) were male. Common cancers were of the liver (45%), spleen (24%), and stomach (14%).

The average total FRI score among 51 caregivers was 8.5 (SD = 2.8). The average scores of the subscales were as follows: cohesion = 3.2 (SD = 1.2), expressiveness = 2.3 (SD = 1.4), and conflict = 1.2 (SD = 1.1). The average total FRI score was below the threshold for good family functioning (<9). Among 51 caregivers, 22 (44%) had a total FRI score of <9.

Table 2 shows the total and subscale scores for FRI, and the total score for CQOLC stratified by three clusters obtained by cluster analysis. Based on their subscale scores of FRI (cohesion, expressiveness, and conflict), we have defined

a. Mean ± SD (range).

Abbreviations: ANOVA, analysis of variance; FRI, Family Relationship Index; CQOLC, Caregiver Quality of Life Index-Cancer ^aPost-hoc analysis for difference between the conflictive type and the mean of the supportive and communicative types showed p = 0.036.

“supportive”, “communicative”, and “conflictive” types. The first type of family structure, which we named the “supportive” type (n = 12), was characterized by high cohesion (average of 3.7, SD = 0.5) and low conflict (average of 0.4, SD = 0.7). Supportive families reported the highest CQOLC total score (average of 77.2, SD = 9.5). The “communicative” group (n = 30) was characterized by high cohesion (average of 3.7, SD = 0.7) and high expressiveness (average of 3.4, SD = 0.7) as well as high conflict (average of 1.3, SD = 1.2). Communicative families reported the second-highest CQOLC total score (average of 74.9, SD = 13.5). The final type, “conflictive” type (n = 8), showed the highest conflict (average of 1.6, SD = 1.2) and lowest CQOLC total scores (average of 65.0, SD = 15.4).

ANOVA showed that the total and subscale FRI scores significantly differed across the three groups. On the other hand, ANOVA did not demonstrate that the CQOLC total score did not differ across three groups. However as shown in Table 2, the conflictive group had lower CQOLC total score than supportive and communicative groups. Therefore, we conducted post-hoc analysis. Post-hoc analysis for differences in CQOLC total scores between the conflictive type and the mean of supportive and communicative types showed statistical significance (p = 0.036). This means that when supportive and communicative groups were combined, it has higher CQOLC total score than conflictive group.