We obtained and analyzed the data from questionnaires answered by children and their families in the ALL-B12. The induction therapy of ALL-B12 comprises approximately a month of combined chemotherapy, including steroid pre-phase and the induction multidrug therapy . The following inclusion criteria were used to select relevant child–family pairs: (1) children with B-cell Precursor ALL and their family had entered ALL-B12 from December 2012 to November 2017; (2) children with B-cell Precursor ALL were aged 5–18 years; and (3) both children and their families returned all and completed questionnaires to the Center for Quality of Life Research after the induction therapy (about 6 weeks from the start of the therapy).
Children and their families were given questionnaires and a self-addressed stamped envelope by their attending physicians before and after 2 weeks of the scheduled date of the end of the induction therapy. Children and their families answered the questionnaires within 4 weeks from receiving them, and then sent them to the Research Center using the self-addressed stamped envelope.
We used the Pediatric Quality of Life Inventory Japanese Version (PedsQL) [24, 25] to measure pediatric HRQOL. PedsQL was developed using surveys with many children, including healthy children and children with various types of disease (e.g., cancer), their families, and healthcare professionals. It was designed to measure pediatric HRQOL in the past month [9, 10, 26]. PedsQL Generic Core Scales (PedsQL-G)  measure general HRQOL using the World Health Organization’s (WHO) definition of health—“Health is a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”—with the added dimensions of role/school functioning . It comprises 21 items for children aged between 5 and 7 and 23 items for those between 8 and 18 in four functioning subscales (i.e., physical, emotional, social, and school functioning). We also used the PedsQL Cancer Module (PedsQL-C) , which focuses on the dimensions of health affected by pediatric cancer and its treatment. It comprises 26 items for those aged from 5 to 12 and 27 items for those from 13 to 18 in eight subscales (i.e., level of pain, nausea, procedural anxiety, treatment anxiety, worry, cognitive problems, perceived physical appearance, and communication). Children and families answered each item on a 5-point Likert scale, where 0 = no problem, 1 = almost never, 2 = sometimes, 3 = often, and 4 = almost always. However, children aged from 5 to 7 answered each item on a 3-point Likert scale adopting faces corresponding to frequencies: a smiley face representing “0 = no problem,” a neutral face representing “2 = sometimes,” and a frowning face representing “4 = almost always.” The children answered the questionnaires with the help of their families or medical staff, if necessary. In such cases, we requested family members to first answer their questionnaires before helping children. Based on the PedsQL scoring algorithm , we calculated the average score for each item in the subscales of both the child self-reports and family proxy-reports, and then transformed them to a 0–100 scale, where high scores indicated high HRQOL. If participants answered fewer than 50% of items in a subscale, the subscale was considered to have missing scores. For all domains, Cronbach’s alphas exceeded .80 for both children and family members. These scales were, therefore, adequately valid in target participants.
Family attendance ratio during children’s hospitalization
Family members, more precisely, the person who mostly took care of the child during hospitalization, reported the number of days of hospitalization and their own attendance during this period in the previous month. From this information, we calculated the family attendance ratio by dividing the number of days of family attendance in the previous month by the number of days of hospitalization in that month.
Social relationships and nursery/school characteristics
The children reported the number of friends they thought to have (1 = none, 2 = one friend, 3 = 2–3 friends, 4 = over 4 friends), whereas their family answered whether children were attending a nursery or school and how present they were (1 = almost always absent, 2 = present one-third of the time, 3 = present two-thirds of the time, 4 = almost always present).
The children reported their age and gender, whereas family members answered their age and relationship to the child.
We used R ver. 3.5.0  and set the significance level to 5% (two-tailed test). As a first step in the analysis, we calculated descriptive statistics classifying all children by their age group, as follows: young children (aged between 5 and 7), school-age children (aged between 8 and 12), and adolescents (aged between 13 and 18).
To verify the agreement between the child self-reports and family proxy-reports, we calculated the differences in each PedsQL subscale between the types of reports using t-statistics and their 95% confidence intervals (95% CI). We also calculated intraclass correlation coefficients (ICCs) between child self-reports and family proxy-reports, and their 95% CIs based on a two-way random effects model . Using Landis et al.’s criteria, ICCs were categorized as weak (≤ .40), moderate (.41–.60), substantial (.61–.80), and almost perfect (≥ .81) .
To explore the factors influencing the agreement between child self-reports and family proxy-reports for each family, we conducted multiple regression analyses for all participant pairs and age groups. The explanatory variables were the seven items previously mentioned, including (1) child’s age, (2) child’s gender, (3) family member’s age, (4) family member’s relationship to the child, (5) number of friends, (6) attendance at nursery/school, and (7) family attendance ratio; all were simultaneously entered into the regression analyses.
For outcome variables, we used several indicators of agreement, according to previous studies, including the absolute difference between PedsQL scores and the ICCs for all subscales between child self-reports and family proxy-reports. Previous studies have used both the difference , and absolute difference  in scores as indicators of agreement. In this study, the raw difference between reports was deemed inappropriate for multiple regression because larger numerical values would indicate less agreement when the difference was positive, but greater agreement when the difference was negative. Thus, we used instead the absolute values of the difference for the subscales of the PedsQL. Finally, we calculated the ICCs for child self-reports and family proxy-reports of all 4 PedsQL-G subscales (ICC-G), which was defined as the agreement for overall pediatric HRQOL. Similarly, we calculated the ICCs of child self-reports and family proxy-reports of all 8 PedsQL-C subscales (ICC-C), which we defined as the agreement of pediatric HRQOL specific to cancer and its treatment.
The ALL-B12 was approved by the institutional review board of the Japanese Society of Pediatric Hematology/Oncology and the Graduate School of Medicine, University of Tokyo. It also received approval from each participating medical center and hospital. We obtained permission to use the anonymized data of participants from the Japanese Society of Pediatric Hematology/Oncology ALL committee. We made sure to protect participants’ personal information by conducting all de-identified data handling and analyses in the Research Center.