Cross-cultural adaptation of the korean version of the minneapolis-manchester quality of life instrument-adolescent form.

We verified the reliability and validity of the Korean version of the Minneapolis-Manchester Quality of Life Instrument-Adolescent Form (KMMQL-AF) among Korean childhood cancer survivors. A total of 107 childhood cancer patients undergoing cancer treatment and 98 childhood cancer survivors who completed cancer treatment were recruited. To assess the internal structure of the KMMQL-AF, we performed multi-trait scaling analyses and exploratory factor analysis. Additionally, we compared each domains of the KMMQL-AF with those of the Karnofsky Performance Status Scale and the Revised Children's Manifest Anxiety Scale (RCMAS). Internal consistency of the KMMQL-AF was sufficient (Cronbach's alpha: 0.78-0.92). In multi-trait scaling analyses, the KMMQL-AF showed sufficient construct validity. The "physical functioning" domain showed moderate correlation with Karnofsky scores and the "psychological functioning" domain showed moderate-to-high correlation with the RCMAS. The KMMQL-AF discriminated between subgroups of different adolescent cancer survivors depending on treatment completion. The KMMQL-AF is a sufficiently reliable and valid instrument for measuring quality of life among Korean childhood cancer survivors.

INTRODUCTION

Because of remarkable improvements in childhood cancer treatment, the 5-yr survival rate of childhood cancer has reached nearly 80% in developed countries (1). In Korea, the 5-yr survival rate of childhood cancer reached as high as 76.7% for those who were diagnosed between 2006 and 2010 from only 54.6% for those who were diagnosed between 1993 and 1995 (2). Such outstanding improvement in the survival rate has resulted in a growing population of childhood cancer survivors and an increasing need to address their health related quality of life (HRQoL).

Although some instruments measuring HRQoL have been used in childhood and adolescent cancer survivors (3-12), most of them are only applicable to a specific cancer type (11, 12) or cannot be used simultaneously in childhood cancer patients on treatment and in those off treatment (4, 7, 9, 13). Moreover, some instruments have been used proxy assessments, completed by parents or physicians (14, 15). The Minneapolis-Manchester Quality of Life (MMQL) instrument is a reliable and well-validated self-reporting inventory (5, 16, 17), which has also been adapted for British (18, 19) and Swedish populations (20). Because childhood cancer survivors often cross developmental stages during cancer treatment and follow-up (21), longitudinal assessment of HRQoL in childhood cancer patients is difficult. To address this problem, 3 versions of the MMQL were developed. The Youth Form (YF) was used for children aged between 8 and 12 yr and was administered to the child by interview. The Adolescent Form (AF) was used for adolescents aged between 13 and 20 yr and was self-administered. The Young Adult Form was used for cancer survivors aged between 21 and 45 yr and was self-administered. These tools can reflect the changes in the developmental stages of the patients and be applicable to childhood cancer patients on treatment and those off treatment.

This study aimed at verifying the reliability and validity of the Korean version of the Minneapolis-Manchester Quality of Life Instrument-Adolescent Form (KMMQL-AF) among Korean childhood cancer survivors.

MATERIALS AND METHODS

Study design, subjects, and data collection

We recruited study participants from Seoul National University Children's Hospital and the Center for Pediatric Cancer, National Cancer Center in Korea between May 2008 and January 2010. The study interviewers recruited the childhood and adolescent cancer patients from either the outpatient clinic or the inpatient wards. The patients were asked to complete the self-administered questionnaire alone in the room when the interview was conducted. All subjects were between 13 and 20 yr of age and were able to read and understand Korean. We divided the participants into 2 groups: 1) 13- to 20-yr-old patients receiving cancer treatment for 2 or more months prior to study participation ('on treatment' cancer patients), and 2) the same-aged participants who had completed cancer treatment over a year ago ('off treatment' cancer survivors). We recruited 107 'on treatment' cancer patients and 98 'off treatment' cancer survivors.

Instruments

We obtained consent from the original author of the original English version of the MMQL-AF. Linguistic validation of the MMQL-AF was performed through a standard forward-backward translation process. A provisional version of the Korean MMQL-AF was pretested on 15 patients during their follow-up clinic visit at the National Cancer Center in Korea. Subjects were asked to comment on the comprehensiveness and clarity of the items in the KMMQL-AF and on the degree of difficulty encountered when answering the questionnaires.

The original version of the MMQL-AF was developed for subjects aged between 13 and 20 yr, and it comprises 46 items pertaining to 7 HRQoL domains: 1) physical functioning, 2) psychological functioning, 3) social functioning, 4) cognitive functioning, 5) body image, 6) outlook on life, and 7) intimate relations. Scoring on the MMQL-AF ranges from 1 (minimal HRQoL) to 5 (maximal HRQoL). However, items 20, 21, and 22 are scored on a 4-point Likert scale and their scoring advances by 1.25. Thus, in these cases, the lowest score is 1.25 and the highest is 5. Therefore, higher scores indicate minimal negative impact and thus greater HRQoL (5). We computed the domain-specific score by summing the scores for all items in each domain and dividing the value by the number of items in that domain. An overall QOL score is calculated by summing the scores for all items, and dividing the value by the number of items in the questionnaire. Because of a low missing rate, we excluded responses with missing values from the calculation.

For assessing concurrent validity, we used Karnofsky Performance Status Scale (KPS) (22) and the Revised Children's Manifest Anxiety Scale (RCMAS) (23, 24). KPS was developed to measure the level of patient activity and medical care requirements. It has been widely used in childhood cancer patients. Karnofsky score ranges from 0 (death) to 100 (perfect). The higher the Karnofsky score, the better the performance status. We hypothesized that the "physical functioning" domain of the KMMQL-AF might have moderate correlation with the Karnofsky score.

RCMAS is a self-report to assess the degree and nature of anxiety experienced by children and adolescents. RCMAS has 37 items that assess trait anxiety of school-aged children. A total anxiety score is computed based on 28 items, which are divided into 3 anxiety domains: physiological anxiety (10 items related to somatic manifestations of anxiety such as sleep difficulties, nausea, and fatigue), worry/oversensitivity (11 items measuring obsessive concerns about a variety of things, most of which are typically vague and ill-defined, as well as fears about being hurt or emotionally isolated), and social concerns/concentration (7 items measuring distracting thoughts and fears that have a social or interpersonal nature). The remaining 9 items on the RCMAS constitute the Lie domain. All the items use a simple "yes-or-no" response format, and each item is given a score of 1 for a "yes" response, yielding a total anxiety score. High scores on the subscales can represent different aspects of anxiety, which can be used to develop hypotheses about the origin and nature of a child's anxiety. We hypothesized that the "psychological functioning" domain of the KMMQL-AF might have moderate correlation with 3 anxiety domains of RCMAS.

Statistical analyses

Descriptive analyses were conducted to evaluate the characteristics of the study participants and to confirm the presence of the ceiling or floor effect in the questionnaires. Cronbach's alpha coefficients were calculated for each of the 7 domains to assess the internal consistency of the KMMQL-AF. A Cronbach's alpha value of 0.7 or more was considered satisfactory.

To assess the underlying factor structure of the KMMQL-AF, exploratory factor analysis with varimax rotation was done. We hypothesized that the KMMQL-AF would have a similar factor structure as the original English version. We extracted components with eigenvalues of more than 1.00. After rotation, individual items with loadings exceeding 0.40 were considered as significant.

To assess the correlations between items and domains within the KMMQL-AF, we conducted multi-trait scaling analysis (25). Convergent validity for each domain was examined by assessing the correlation between each item and its own domain (corrected for overlap) with Pearson's correlation coefficients. Items with correlation values ≥0.4 were considered as valid. The discriminant validity of the KMMQL-AF was assessed by comparing the correlation of each item with its own domain (corrected for overlap) with the correlation of each item with the other domains of the KMMQL-AF. Scaling errors were defined as cases in which an item correlated significantly less with its own domain than with the other domains.

To assess concurrent validity, we analyzed Pearson's correlation coefficients among the KMMQL-AF, KPS, and RCMAS. Coefficients below 0.40 revealed weak correlation, and those between 0.40 and 0.60 revealed moderate correlations. Coefficients above 0.60 showed high correlation between the domains.

Known group validity was examined by comparing the 2 study participant groups ('on treatment' and 'off treatment' groups). Student's t-tests for each domain of the KMMQL-AF were conducted to determine which scales were able to differentiate between 'on treatment' and 'off treatment' groups.

All the statistical analyses were performed using STATA 12.1 (STATA Corp., Houston, TX, USA). Statistical significance was defined as P value≤0.05 on 2-tailed analyses.

Ethics statement

This study was approved by the institutional review boards of the National Cancer Center (NCCNCS-08-111) and Seoul National University Hospital (H-0803-046-238). The study interviewers explained the survey purpose and procedures to the patients and their parents. They also obtained informed consent from both the patients and parents.

RESULTS

Baseline characteristics of the study participants

The baseline characteristics of the study participants are shown in Table 1. The mean age of the participants was 17.9 yr (standard deviation [SD]=2.1 yr). Male participants were more common (n=120, 58.5%) than were female participants; and most of the participants had graduated elementary school (n=89, 43.4%) or middle school (n=75, 35.6%). Leukemia/lymphoma was the most common cancer type (n=88, 42.9%), followed by solid tumors/other types (n=82, 40.0%). All of the participants had received chemotherapy, and about 40% of the participants had undergone surgery (n=82, 40.0%) and radiotherapy (n=77, 37.6%). There was no significant difference in the baseline characteristics between 'on treatment' and 'off treatment' groups.

Table 1

Baseline characteristics of the study participants

Frequency distribution of responses

The frequency distribution of responses is shown in Table 2. All of the items showed low missing proportions below 2%, except item 2 (8.3%). Although all items showed the low proportions of the lowest score (below 40%), 11 items (items 7, 12, 13, 16, 27, 30, 31, 33, 35, 38, and 40) showed the high proportions of the highest score (over 40%) (Table 2).

Table 2

Distribution of responses

*Reversed item; †The highest scores of item reversed items 20, 21, and 22 are 4.

Internal consistency and multi-trait scaling analyses

Table 3 represents the internal consistency and multi-trait scaling analyses of the KMMQL-AF. All of the domains had satisfactory Cronbach's alpha coefficients ranging between 0.78 and 0.92. When we omitted items 4, 11, and 40 from the scales, the internal consistency improved (not shown in Table).

Table 3

Internal consistency and multi-trait scaling analysis of KMMQL-AF

*Cronbach's α values≥0.7 indicate adequate scale reliability; †Corrected for overlap.

Multi-trait scaling analyses of the KMMQL-AF confirmed the suggested structure with low scaling errors (6/322=1.9%). However, items in the "physical functioning" domain showed relatively low item-own scale correlations.

Exploratory factor analyses

The original version of the MMQL-AF has 46 items in 7 domains. They are as follows: "physical functioning" (items 2-8, 11, and 15); "cognitive functioning" (items 34-42); "psychological functioning" (items 9, 10, 12-19); "body image" (items 20-25); "social functioning" (items 28-33); "outlook on life" (items 45-47); and "intimate relations" (items 26, 27, 43, and 44).

When exploratory factor analysis with principal-component factor extraction was performed using our data, 7 factors were extracted from the data, explaining 84.5% of the total variance (Table 4). Factor 1 was composed of all the items in the "social functioning" and "intimate relations" domains. Factor 2 was composed of all the items in the "outlook on life" domain and some items from the "body image" domain. Other items from the "body image" domain were combined into Factor 6. Factor 3 was similar to the "cognitive functioning" domain, and Factor 4 shared most items with the "physical functioning" domain. Items in the "psychological functioning" domain were included in Factors 5 and 7.

Table 4

Result of exploratory factor analysis of KMMQL-AF

*Item with factor loading under 0.4.

Items 4 ("Need time to rest during the day"), 8 ("Prefer to watch rather than to take part in games and sports"), and 11 ("Feeling tired during the day") had no significant factor loadings on their own domain, i.e., the "physical functioning" domain. Items 16 ("Worried about dying"), 19 ("Feeling inferior to most people"), 20 ("Being satisfied with their weight"), and 40 ("Difficulty with reading or writing") also had no significant factor loadings on their own domain. Item 18 ("Worried about things in general") showed significant overlapping factor loadings between Factors 5 and 7. Item 22 ("Feeling about their body development") also showed significant overlapping factor loadings between Factors 2 and 6 (Table 4).

Concurrent validity

Table 5 shows the concurrent validity of the KMMQL-AF. The "physical functioning" domain showed moderate correlation with the Karnofsky scores. The "psychological functioning" and "outlook on life" domains also demonstrated moderate correlation with the Karnofsky scores. The "psychological functioning" domain showed moderate-to-high correlations with all domains of the RCMAS (0.43-0.64). Furthermore, the "cognitive functioning" domain showed moderate and positive correlations with the "social concerns/concentration" domain (0.45) of the RCMAS, and the "intimate relations" domain showed moderate and positive correlations with the "worry/oversensitivity" (0.40) and "social concerns/concentration" (0.51) domains of the RCMAS.

Table 5

Concurrent validity of KMMQL-AF

<0.40 (weak correlation), 0.40-0.60 (moderate correlation), >0.60 (high correlation). *Domain measured by Revised Children's Manifest Anxiety Scale (RCMAS).

Known group validity

We used Student's t-test to assess known group validity (Table 6). According to the KMMQL-AF, 'off treatment' group had significantly higher scores in the "physical functioning" (3.1 vs 3.6; P<0.001), "psychological functioning" (4.0 vs 4.2; P=0.01), "outlook on life" (3.2 vs 3.9; P<0.001), and "intimate relations" (3.5 vs 3.8; P=0.04) domains than did the 'on treatment' group. However, the MMQL-AF could not discriminate between the 2 groups when considering the cognitive functioning, body image, and social functioning domains.

Table 6

Known group validity of KMMQL-AF

*P values are calculated by Student's t-tests.

DISCUSSION

To our knowledge, this is the first study that has cross-culturally adapted the MMQL-AF to the diverse childhood cancer survivors in Asia. We validated the KMMQL-AF in Korean children and adolescents with various cancer types. The KMMQL-AF had sufficient reliability and demonstrated correlation with other inventories, as we predicted. Furthermore, it distinguished between 'on treatment' cancer patients and 'off treatment' cancer survivors. Most of the participants completed the questionnaires with the missing rates below 2%. This low missing rate reveals that the KMMQL-AF is a feasible tool. Therefore, we were able to confirm that the KMMQL-AF is valid and reliable.

The factor structure of the KMMQL-AF was similar to that of the original version, although some differences were observed. In multi-trait scaling analyses, the KMMQL-AF had few scaling errors (6/322=1.9%; items 4, 8, 11, 19, 23, and 40). However, exploratory factor analysis revealed some potential differences in the underlying factor structure of the KMMQL-AF from that of the original version of the MMQL-AF. Although the "cognitive functioning," "social functioning," "outlook on life," and "intimate relations" domains demonstrated similar internal structure as the original version, the others had some differences. Seven of the 46 items in the KMMQL-AF (items 4, 8, 11, 16, 19, 20, and 40) had factor loadings under 0.4, which means that these items do not strongly belong to one specific domain. For example, items 4 ("Need time to rest during the day") and 8 ("Prefer to watch rather than to take part in games and sports") could not be strong indicators of poor physical functioning, although factor loading was the strongest for this domain than for any other domain. In another example, the factor loading for item 19 ("Feeling inferior to most people") was shared by Factor 1 (corresponding to the "social functioning" domain) and Factor 3 (corresponding to the "cognitive functioning" domain), indicating that this item is not just related to psychological functioning in our population.

The "social functioning" and "intimate relations" domains were combined into 1 factor. This result suggests that both the "social functioning" and "intimate relations" domains ask similar issues on interpersonal relationships; therefore, in exploratory factor analysis, they were combined into 1 factor. The "body image" domain was divided into 2 factors, namely, satisfaction with their own body (items 21 "Being happy about the way they look" and 23 "Liking their body the way it is") and cognition about development status (items 22 "Feeling about their body development," 24 "Feeling that others think that their body is poorly developed," and 25 "Feeling uncomfortable about the way their body is developing"). The "psychological functioning" domain was also divided into 2 factors, grouped as emotional status (items 9 "feeling sad," 10 "Feeling angry," 12 "Feeling lonely," 13 "Feeling frightened," and 14 "Feeling nervous or anxious") and worry (items 17 "Worried about their health" and 18 "Worried about things in general"). While it may be possible to modify the MMQL-AF to reflect the potential differences in factor structure revealed by the exploratory factor analysis, we decided to retain the original factor structure based on acceptable multi-trait scaling analysis results and to ensure international comparison.

In concurrent validation, the "physical functioning" domain showed moderate correlation with the Karnofsky performance scale scores, as expected. This result suggests that the "physical functioning" domain reveals the performance status of cancer survivors. The "psychosocial functioning" domain also showed moderate-to-high correlation with all the domains in the RCMAS. Therefore, we can expect that the KMMQL-AF can simultaneously measure physical and psychological aspects of HRQL in childhood cancer patients. The "cognitive functioning" and "intimate relations" domains also showed moderate correlation with the "social concerns/concentration" domain in the RCMAS. The "social concerns/concentration" domain in the RCMAS suggests that the child is likely to feel that he or she is unable to meet the expectations of other important people and is inadequate and unable to concentrate on tasks.

In known-group validity analysis of the KMMQL-AF, the "physical functioning," "psychological functioning," "outlook on life," and "intimate relations" domains could discriminate between 'on treatment' and 'off treatment' groups. 'Off treatment' cancer survivors showed significantly higher scores than 'on treatment' cancer patients. This suggests that various physical and sociopsychological consequences of childhood cancer in childhood cancer survivors during the course of cancer treatments generally improve after cancer treatment. However, there were no significant differences in the "cognitive functioning," "body image," and "social functioning" domains between the 2 groups, indicating that the experience of appearance change and social isolation during cancer treatment lasts even after cancer treatment.

Our study has some limitations. First, we could not compare the KMMQL-AF with other instruments assessing QOL in childhood cancer survivors. At the beginning of our study, there was no validated tool for QOL assessment in adolescent cancer patients in Korea. Instead, we used various instruments for assessing concurrent validity of physical performance and anxiety in adolescent cancer patients. Second, because our study population had no control children without cancer, we could not compare childhood cancer survivors with children without cancer history. However, Bhatia et al. (5) reported that the MMQL-AF discriminated between the 2 groups, and cancer patients showed a significantly increased risk of poor QOL.

In conclusion, the KMMQL-AF appears to be a reliable and valid instrument for HRQoL assessment in childhood cancer patients aged between 13 and 20 yr. We believe that studies on HRQoL assessment by KMMQL-AF will help to develop interventional strategies for improving the HRQoL of childhood cancer survivors in the Republic of Korea.