something about HRQoL

本人做课题的时候找到的一种新的评估癌症患者生活质量的量表——QUICK FLIC，或许对做类似课题的朋友又一点帮助！
新手，第一次发帖，谢谢支持啊！
Quick-FLIC
A Short Questionnaire for Assessing Quality of Life of Cancer Patients
Yin-Bun Cheung, Kei-Siong Khoo, Zee-Wan Wong, Hui-Ti See, Han-Chong Toh,
Richard J. Epstein, Gim-Yew Ng and Say-Beng Tan
From the Division of Clinical Trials and Epidemiological Sciences (Y.-B. Cheung, G.-Y. Ng, S.-B. Tan) and
the Department of Medical Oncology (K.-S. Khoo, Z.-W. Wong, H.-T. See, H.-C. Toh, R.J. Epstein), National
Cancer Centre, Singapore
Correspondence to: Yin-Bun Cheung, Division of Clinical Trials and Epidemiological Sciences, National Cancer
Centre, 11 Hospital Drive, Singapore 169610. Tel: +65 6436 8208. Fax: +65 6225 0047. E-mail: ctecyb@
nccs.com.sg
Acta Oncologica Vol. 42, No. 1, pp. 36–42, 2003
Health-related quality of life instruments tend to include a great many items. This imposes a burden on the respondents as well as
undermining response rate and data quality. In this study we developed a shortened version of the Functional Living Index—Cancer
(FLIC), now called Quick-FLIC, and examined its measurement properties. A questionnaire package, self-administered by 140 patients,
included the FLIC and the Functional Assessment of Cancer Therapy—General. A factor analysis and clinical judgement were used to
shorten the FLIC, which included 22 items. Each subscale of FLIC was shortened to include two or three items only. The Quick-FLIC
included a total of only 11 items. Nevertheless, the measurement properties of the Quick-FLIC and its subscales were comparable to those
of the original FLIC. It is concluded that the shortening of established health-related quality of life instruments is viable in oncology
research.
Recei
ed 3 June 2002
Accepted 10 October 2002
ORIGINAL ARTICLE
Health-related quality of life (HRQoL) questionnaires tend
to include a great many items. This not only imposes a
burden on the respondents, but also undermines response
rate and data quality, since some respondents, especially
those suffering from a serious illness, may have difficulty
in completing a lengthy form. In reviewing the problems in
the evaluation of quality of life of cancer patients, Ballatori
(1) suggests that simpler HRQoL questionnaires that
can be completed quickly are needed, a sentiment echoed
by others (2). International experts also recommend that,
in general, short HRQoL questionnaires should be preferred
(3).
Since items in an HRQoL scale or subscale should be
highly correlated, and since some items may be more
informative than others from a clinical or statistical point
of view, it is reasonable to abbreviate an existing lengthy
questionnaire by excluding less-informative items (4, 5).
The shortening of composite measurement scales is relatively
common in the educational and psychological fields
in contrast to the health research field. In the area of
quality of life research, Ware et al. (6) streamlined a
general quality of life measure—the Short Form 36 Health
Survey—from 36 items to 12 items. In a study of the
psychosocial impact of diabetes a measurement instrument
was reduced from a 24-item measure to a 7-item measure
(7). Anne Moran et al. (8) showed that the number of
items in each of the Chronic Respiratory Questionnaire
subscales could be reduced to as few as two items without
substantially compromising the measurement properties of
the subscales. They demonstrated the practicability of
shortening HRQoL questionnaires.
In this paper we report the development of a shortened
version of the Functional Living Index—Cancer (FLIC)
(Chinese version), which we have termed Quick-FLIC. The
central aim of this study was to determine whether a
well-established HRQoL instrument for oncology studies
could be substantially shortened without damaging its
validity. The original and short versions were compared,
with reference to the Chinese version (V4) of the Functional
Assessment of Cancer Therapy—General (FACTG)
as well as patients’ clinical characteristics.
MATERIAL AND METHODS
The FLIC questionnaire includes 22 items within a total of
5 subscales (9, 10). Each item scores between 1 and 7 on a
visual analogue scale (VAS). The Psychological and Physi-
© Taylor & Francis 2003. ISSN 0284-186X Acta Oncologica
Acta Oncologica 42 (2003) Quick-FLIC 37
cal subscales include 7 items each; the Family and Symptoms
subscales include 3 items each, while the Social
subscale includes only 2 items. The index was translated
into Chinese and validated by the FLIC developers in
collaboration with the researchers in Singapore (11). The
researchers subsequently changed the VAS to a 7-point
ordered category scale as experience showed that some
patients have difficulties in understanding the VAS (12),
and that empirically the VAS is no different from an
ordered category scale (13, 14). Patients were recruited to
the present study from the outpatient clinics of the National
Cancer Centre, Singapore, from December 2001 to
February 2002. The patients were screened by oncologists
in the clinics according to the inclusion and exclusion
criteria. The inclusion criteria were: having the ability to
understand written Chinese and speak Mandarin (the offi-
cial Chinese language in Singapore); being age 18 years or
older; being physically fit enough to self-administer a
questionnaire. The exclusion criteria were: having no
known evidence of brain metastasis, psychosis, or severe
depression. Eligible patients were then referred to a research
psychologist who explained the study in detail,
obtained informed consent, and delivered and collected the
questionnaires. The questionnaires included the Chinese
versions of FLIC and FACT-G, as well as some demographic
variables. The referring oncologists rated Karnofsky
performance scores, treatment status and some other
clinical information. The FACT-G is a validated measure
of HRQoL that has been translated into many languages
(15) and is used as the main criterion of validity in this
study. The study was approved by the Ethics Committee
of the National Cancer Centre, Singapore.
Coste et al. (5) reviewed the approaches to the shortening
of composite measurement scales, and showed that the
major statistical methods available are a) correlation or
regression analysis and b) factor analysis. They pointed
out the analytic problems caused by correlated errors in
the former approach. Furthermore, they proposed that for
studies of subjective constructs, such as HRQoL, the
shortening procedure should not be totally reliant on
statistical criteria. Clinical judgement and statistical information
should be combined in determining which items
are deemed redundant and which are not. Statisticians
share similar views about the use of clinical judgement in
factor analysis (16). In order to maintain the factor structure
of the original version, we shortened the FLIC subscales
rather than directly shortening the FLIC. Taking
account of previous research findings that a subscale can
be shortened to include as few as two items (8), we paid
particular attention to finding two or three of the best
items within each subscale. The Social subscale of FLIC
only contains two items and was therefore not included in
the shortening procedures. On each of the Psychological,
Physical, Family and Symptoms subscales, we carried out
factor analysis with a single factor by maximum likelihood
(17). The items within each subscale were ranked according
to their factor loading. Items with a high factor
loading were highly likely to be retained, or vice versa. The
research team discussed and judged the face value of the
items in the above ranking order and made a final decision
on the order of priority of their being retained. The
subscale scores of the short version were computed by a
simple summation of the chosen items; the total score of
the short version was then computed by a simple summation
of the short subscale scores. We also proposed formulae
for converting the scores into the 0–100 scale.
Convergent and divergent validity evaluations were examined
by correlation analysis. For instance, a strong
correlation between the FLIC Physical subscale and the
FACT-G Physical subscale was taken as evidence of convergent
validity; a weak correlation between the FLIC
Physical subscale and the other FACT-G subscales was
taken as evidence of divergent validity. Relative efficiency
was examined by ratio of F-values in analysis of variance
(ANOVA) of FLIC scores by tertiles of FACT-G scores
(18). A higher efficiency means that a smaller sample size
is required to reflect a significant difference between patients
in tertiles of FACT-G scores. Reliability in terms of
Cronbach’s alpha was estimated (13, 18). In order to assess
how many items were required to give a satisfactory short
version, we presented the comparative results of using
between one and all the items of the original subscales to
form the short version.
Having developed the Quick-FLIC, we compared its
ability to differentiate patients with different clinical characteristics,
including performance status and treatment
status, with that of the original FLIC and FACT-G.
Relative efficiency in differentiating groups was estimated
by the ratio of F-values in ANOVA (18). To facilitate easy
comparison, all three measures were normed to have a
range from 0 to 100.
RESULTS
Among the 152 patients approached, 6 refused to participate
and 6 were found to have insufficient knowledge of
written Chinese. A total of 140 patients were therefore
interviewed. Of these, 84 returned the questionnaires with
no missing values, 47 with one missing value and 9 with
two missing values. We applied the ‘‘half rule’’ for dealing
with missing values (19). That is, a missing value was
replaced by the mean of the answered questions in that
subscale if at least half of the questions in the subscale
were answered. There were no missing subscale scores after
this imputation procedure. The demographic and clinical
profiles of the patients are described in Table 1. The
patients were relatively young compared with western populations.
This was due to the fact that older Chinese
patients were less likely to meet the inclusion criteria as the
majority of them spoke dialects instead of Mandarin.
Y.-B. Cheung et al. Acta Oncologica 42 (2003) 38
Table 1
Descripti
e summary of participants : mean and SD (range ) for age ; frequency and percentage for category

ariables
Category Variable Frequency/Mean Percentage/SD (Range)
53 Age 11 (23 to 78)
77 55 Sex Female
63 45 Male
Primary incomplete Education 37 26
Primary complete 44 31
50 36 Secondary school
Post-secondary 9 6
109 78 Married Marital status
Single 14 10
Divorce/Separated 9 6
8 6 Widowed
Breast Tumour type 50 36
42 30 Colorectal
11 8 Stomach
16 11 Lung
Others 21 15
58 41 100 Karnofsky score
90 61 44
80 13 9
8 6 70 or below
50 36 Receiving chemotherapy Yes
90 64 No or RT
About one-third of the subjects were undergoing
chemotherapy or radiotherapy. Most participants had a
Karnofsky performance score of 100 (41%) or 90 (44%);
relatively few had a Karnofsky score of 80 (9%) and 70 or
below (6%).
The results of the factor analysis can be found in Table
2. The items within each subscale were ranked according
to their factor loadings, or equivalently, according to the
proportion of variation explained by the factor (R2). In the
Physical subscale, the items ‘Feel well’ and ‘Appear well’
had the highest factor loading. Statistically, these two
items should be the most important in representing this
subscale. However, in the judgement of the research team,
the two items look very similar on a questionnaire, although
they are conceptually distinct. Even the factor
loadings are identical up to two significant digits. This led
to some doubt concerning whether they are in fact measuring
the same thing. As such, we decided to give top
priority to ‘Feel well’ as a candidate for inclusion in the
short version, but gave lowest priority to ‘Appear well’.
So, after ‘Feel well’, the second item to be considered for
retention was ‘Feel uncomfortable’, and so on. Similarly,
we changed the order for inclusion in the Symptoms
subscale and Family subscale. Our clinical judgement was
in agreement with the other rankings based on the factor
analysis findings.
In order to assess convergent validity, we calculated the
Pearson’s correlation coefficient (r) between FLIC Physical
and FACT-G Physical scores, FLIC Psychological and
FACT-G Emotion scores, FLIC Symptoms and FACT-G
Physical scores, FLIC Family and FACT-G Social scores.
In Table 3 we present the results of retaining different
number of items in the short FLIC measures, based on the
order of priority for retention established above. For
comparison purposes, the correlations based on the original
FLIC subscale with all items retained are also included
in the table. A shortened version of the Physical subscale
with only two items gave a correlation of 0.69 with the
FACT-G Physical subscale. This was very close to that of
using all 7 items, which give a correlation of 0.71. If three
items were included in the short version, the correlation
was 0.75, even stronger than that of the 7-item subscale.
The short version of the Psychological subscale with two
items gave a correlation coefficient of 0.75 with the FACTG
Emotion subscale, which was almost the same as that
using all 7 items (r=0.76). Similarly, the Symptoms and
Family subscales with two items each gave a correlation
with their FACT-G counterparts, similar to that using all
three items in the original subscales.
The relative efficiency (RE) of the shortened versions of
the FLIC subscales in relation to the full versions is
summarized in Table 4. In all four subscales, using only
one item resulted in a much lower level of efficiency. Using
two items in the Physical subscale was almost as efficient
as using all 7 items in differentiating FACT-G Physical
score tertiles (RE=0.96). A 3-item version was even more
efficient (RE=1.19) than the original subscale. Similarly,
a 2-item Symptoms subscale and a 2-item Family subscale
gave satisfactory results. Although a 2-item Psychological
subscale was as valid as the 7-item subscale, as indicated
Acta Oncologica 42 (2003) Quick-FLIC 39
Table 2
Results of factor analysis for each FLIC subscale
Factor loading R2 Order for inclusion in Subscale Item*
short subscale**
0.849 0.72 Physical 1 Feel well
Appear well 0.848 0.72 7
Feel uncomfortable 0.799 0.64 2
0.682 0.47 Pain or discomfort 3
Meal or minor housework 0.608 0.37 4
Non-work activities 0.522 0.72 5
0.440 0.19 Maintain usual recreation 6
0.792 0.63 1 Psychological Discouraged
0.762 0.58 Frightened of the future 2
Thinking about illness 0.682 0.47 3
0.634 0.40 4 Hardship on you
0.583 0.34 Feelings of depression 5
Coping with stress 0.506 0.26 6
Confidence in treatment 0.334 0.11 7
0.87 0.76 How much nausea 1 Symptoms
Nausea-affected functioning 0.80 0.64 3
Pain related to illness 0.60 0.36 2
0.83 0.69 Hardship on those closest to you 1 Family
Disruptive to those closest to you 0.82 0.67 3
0.71 0.50 2 Satisfaction with housework
Abbreviation: FLIC=Functional Living Index—Cancer.
* Items selected for the short FLIC are in bold type.
** Taking into account both factor analysis result and clinical judgement.
Table 3
Correlation between shortened
ersions of the FLIC subscales with
arious numbers of items retained and FACT-G
subscales
FLIC subscale Correlation with No. of items retained in short FLIC subscales
1 2 3 4 5 6 7
0.64** 0.69** 0.75** 0.75** FACT-G Physical 0.75** 0.71** 0.71** Physical
0.61** Psychological 0.75** 0.73** 0.74** 0.76** 0.76** 0.76** FACT-G Emotion
0.40** 0.61** 0.65** FACT-G Physical Symptoms
0.16 0.18* 0.17* Family FACT-G Social
** p
0.01; * p
0.05.
Abbreviation: FLIC=Functional Living Index—Cancer.
by the correlation with the FACT-G Emotion subscale
(r=0.75 vs. 0.76 in Table 3), its relative efficiency was not
high (RE=0.82). Using the sample size calculation
method for ANOVA (20), we calculated that the sample
size for rejecting the null hypothesis of no difference in the
2-item Psychological subscale score between tertiles of
FACT-G Emotion was only eight subjects per group
(power=0.8 and 2-sided p-value=0.05). This 2-item version
should be considered efficient enough for practical
purposes.
Based on the above findings, we intended to use a 3-item
version of the FLIC Physical subscale, and a 2-item version
of the other subscales in the short version. Table 5 presents
convergent and divergent validity data. As conceptually
expected, the short FLIC Physical score was correlated the
most with the FACT-G Physical score; the short FLIC
Psychological score was correlated most with the FACT-G
Emotion score; and the short FLIC Symptoms score with
the FACT-G Physical, followed by the FACT-G Functional
scores. The overall short FLIC score, which was an
unweighted sum of 11 items (including the two items in the
Social subscale), was correlated the most with the FACT-G
total score. The short FLIC Family score was only weakly
correlated with the FACT-G Social subscale score. The
Cronbach’s alpha suggests good reliability in the shortened
version of the Physical and Psychological subscales, as well
as the overall FLIC. Reliability of the short Symptoms and
Family subscales was moderate.
As the highest and lowest scores for each item are 7 and
1, respectively, the item score can be normed to the 0–100
interval by the equations (19): normed item score=(item
score–1)×16.67.
Y.-B. Cheung et al. Acta Oncologica 42 (2003) 40
Table 4
Relati
e efficiency (F-
alue ) of shortened
ersions of FLIC subscales, with
arious numbers of items retained in differentiating tertiles of
FACT-G subscales
FLIC ANOVA by No. of items retained in short FLIC measures
sub-scale tertiles of
1 2 3 4 5 6 7
FACT-G 0.72 (34.2**) 0.96 (45.4**) 1.19 (56.7**) Physical 1.10 (52.1**) 1.11 (53.1**) 0.95 (45.4**) 1 (47.5**)
Physical
FACT-G Psychological 0.41 (30.3**) 0.82 (61.1**) 0.78 (57.6**) 0.88 (65.4**) 0.99 (73.0**) 1.01 (75.1**) 1 (74.1**)
Emotion
Symptoms 0.31 (11.6**) FACT-G 0.94 (34.8**) 1 (37.1**)
Physical
FACT-G 0.57 (2.7) 1.14 (5.5**) 1 (4.8**) Family
Social
** p
0.01; * p
0.05 in ANOVA.
Each with numerator/denominator degree of freedom 2/137.
Table 5
Con
ergent and di
ergent
alidity, reliability, and uni
ariate statistics of shortened FLIC subscales
Scale value Shortened FLIC scales Cronbach’s alpha Correlation with FACT-G
Functional Emotional Social Total Mean Physical SD
0.75** 0.49** 0.49** 0.21* 0.62** 76 25 0.82 Physical
0.29** 0.75** 0.21* 0.50** 74 Psychological 26 0.77 0.35**
0.41** 0.28** 0.15 0.47** 0.64** 73 Symptoms 25 0.49
0.48** Family 0.42** 0.46** 0.17* 0.51** 76 22 0.54
0.51** 0.64** 0.27** 0.67** Total 76 18 0.86 0.67**
Abbreviation: FLIC=Functional Living Index—Cancer; FACT-G=Functional Assessment of Cancer Therapy—General.
** p
0.01; * p
0.05.
In applying this formula, the normed values should be
rounded to one decimal point. The normed subscale score
is then a simple average of the normed item scores in the
subscale, and the normed overall scores a simple average
of the 11 normed subscale scores. The means and SDs of
the shortened measures normed to the 0–100 range are
also presented in Table 5.
As the correlation between the short Family subscale
and the FACT-G Social subscale was weaker than expected
(Table 5), we decided to further assess the knowngroup
validity of these two subscales. We used ANOVA
models to assess the mean scores by treatment status (on
chemotherapy/radiotherapy or neither) and Karnofsky
performance scores. For the FLIC Family subscale, the
F-values in the two ANOVA models were 14.63 (p
0.01)
and 2.88 (p=0.04), respectively. In contrast, the FACT-G
Social subscale gave F-values of 0.06 (p=0.81) and 1.29
(p=0.28), but failed to show known-group validity.
Table 6 shows ANOVA of the short version of FLIC,
the original FLIC and FACT-G by treatment status. All
three measures indicated a lower level of HRQoL in
patients undergoing chemotherapy or RT. Both the shortened
FLIC and the original FLIC reached a level of
statistical significance (p
0.01 each); and the mean values
were similar. Mean FACT-G scores were not significantly
different between the two groups (p=0.21). The relative
levels of efficiency of the short FLIC and FLIC in relation
to FACT-G in differentiating the groups were 7.0 and 6.9
(each p
0.05), respectively. In Table 7 the three measures
are compared using the Karnofsky performance status. All
of them showed a clear trend towards HRQoL according
to performance status (each p
0.01). Again, the mean
values of the short and original versions of FLIC were
similar. No significant difference in relative efficiency was
found between any of them (each p0.05).
DISCUSSION
Some studies elsewhere have demonstrated the practicability
of reducing the number of items in composite measurement
scales by 60–70% (6–8). In the present study, we
shortened the 22-item FLIC into an 11-item version without
modifying its five-factor structure. Our findings agree
broadly with those of Anne Moran et al. (8), who reported
little loss of accuracy on reducing the number of HRQoL
subscale items to as few as two. The 2-item versions of the
short subscales showed a level of validity similar to that of
their full versions. Their efficiency in differentiating groups
was also acceptable. We used a 3-item Physical subscale,
Acta Oncologica 42 (2003) Quick-FLIC 41
Table 6
Mean quality of life scores by treatment status, tested by analysis
of
ariance
Receiving Shortened FLIC FACT-G
chemotherapy version of
FLIC or radiotherapy
69 Yes 69 70
No 80 78 74
F-value 11.2 11.0 1.6
P-value 0.001 0.001 0.212
* All three scales normed to 0–100 interval.
The Quick-FLIC and the original FLIC performed comparably
in differentiating groups with different clinical
characteristics. The present study is a first step towards
shortening an HRQoL questionnaire for cancer patients.
As suggested by previous researchers (5), the new short
instrument should be further tested for its measurement
properties in future studies, e.g. sensitivity in detecting
change over time.
Review articles published recently in this journal point
out some of the advances in quality of life assessment that
have taken place in the past two decades (3, 22). Nevertheless,
there are still some practical problems to be resolved
and there continues to be a need for further development.
Major problems include the inability of some patients to
complete HRQoL questionnaires because of poor health
(22) and missing data during follow-up (3). There is not
yet an established and proven approach to deal with these
problems. The strategies that have been suggested include
using proxy respondents (22) and using short instead of
lengthy questionnaires (23, 24). Our study presents the
preliminary findings showing that it is possible to abridge
an HRQoL instrument in oncological research. A short
questionnaire with 11 items, Quick-FLIC, is proposed for
further research. Since the patients in this study had a
relatively high performance status (partly because of the
requirement of being able to self-administer a lengthy
questionnaire package), a future step may be to examine
the properties of Quick-FLIC in less robust patient
groups. If the response rate is to be increased in such
groups, we submit that shorter questionnaires, such as
Quick-FLIC, could become more widely applicable across
the clinical spectrum.
ACKNOWLEDGEMENTS
This work was supported by an individual research grant
(cNMRC/0546/2001) from the National Medical Research
Council, Singapore. Dr Cynthia Goh and Dr Harvey Schipper
kindly allowed us to modify and use the FLIC (Chinese version).
REFERENCES
1. Ballatori E. Unsolved problems in evaluating the quality of
life of cancer patients. Ann Oncol 2001; 12 (Suppl 3): S11–3.
2. Higginson IJ, Carr AJ. Using quality of life measures in the
clinical setting. Br Med J 2001; 322: 1297–300.
3. Gunnars B, Nygren P, Glimelius B, for the SBU-group.
Assessment of quality of life during chemotherapy. Acta
Oncol 2001; 40: 175–84.
4. Chow SC, Ki FYC. On statistical characteristics of quality of
life assessment. J Biopharm Stat 1994; 4: 1–7.
5. Coste J, Guillemin F, Pouchot J, Fermanian J. Methodological
approaches to shortening composite measurement scales. J
Clin Epidemiol 1997; 50: 247–52.
6. Ware JE, Kosinski M, Keller SD. A 12-item Short-Form
Health Survey (SF-12®): construction of scales and preliminary
tests of reliability and validity. Med Care 1996; 32:
220–33.
not because the 2-item version was inadequate, but because
the 3-item version appeared to be superior to the
original one in terms of both convergent validity and
efficiency. In view of such advantages, we considered it
worthwhile to include the third item. The 2-item Symptoms
subscale had a relatively low level of reliability
(a=0.49). Symptom checklists similar to this are sometimes
developed or included in HRQoL instruments under
the concept of ‘clinimetrics’ (18, 21). It is accepted that for
clinimetric items, the evaluation of psychometric properties
is less important.
The short version of the FLIC Family subscale did not
correlate strongly with the FACT-G Social subscale (r=
0.18), but neither did the full version (r=0.17). This is
because the FACT-G Social subscale items are more oriented
to the impact of family life on patients’ well-being
(e.g. ‘I get support from my family’), while the FLIC
Family subscale is more oriented to the impact of patients’
disease on the family (e.g. ‘Rate the degree to which your
illness has imposed a hardship on those closest to you’). In
an ANOVA, the short FLIC Family subscale showed a
difference between groups defined according to performance
status and treatment status, thus supporting its
validity. In addition, the analyses cast doubt on the validity
of the FACT-G Social subscale, since it failed to
differentiate groups characterized by varying treatment
and performance status.
Table 7
Mean quality of life scores by le
el of Karnofsky performance
score, tested by analysis of
ariance
FLIC Performance FACT-G Short version of
FLIC score
100 83 82 78
72 73 90 74
80 65 64 66
59 70 52 63
8.7 9.5 7.2 F-value

0.001
0.001
0.001 P-value
* All three scales normed to 0–100 interval.
Y.-B. Cheung et al. Acta Oncologica 42 (2003) 42
7. Cheung YB, Mok PH, Lee SH. The psychosocial impact of
diabetes mellitus: validation and abridgement of the Chronicity
Impact Scale. Hong Kong J Mental Health 1997; 26:
36–44.
8. Anne Moran L, Guyatt GH, Norman GR. Establishing the
minimal number of items for a responsive, valid, health-related
quality of life instrument. J Clin Epidemiol 2001; 54:
571–9.
9. Ruckdeschel JC, Piantadosi S. Quality of life in lung cancer
surgical adjuvant trials. Chest 1994; 1: 287–96.
10. Schipper H, Clinch J, McMurray A, Levitt M. Measuring the
quality of life of cancer patients: the Functional Living Index—
Cancer. J Clin Oncol 1984; 2: 472–83.
11. Goh CR, Lee KS, Tan TC, et al. Measuring quality of life in
different cultures: translation of the Functional Living Index
for Cancer (FLIC) into Chinese and Malay in Singapore. Ann
Acad Med Sing 1996; 25: 323–34.
12. Cheung YB, Ng GY, Wong LC, et al. Measuring quality of
life in Chinese cancer patients: a new version of the Functional
Living Index—Cancer (Chinese). Ann Acad Med Sing
(in press).
13. Streiner DL, Norman GR. Health measurement scales: a
practical guide to their development and use. Oxford: Oxford
University Press, 1989.
14. McCormack HM, Horne DJ, Sheather S. Clinical application
of visual analogue scales: a critical review. Psycho Med 1988;
18: 1007–19.
15. Cella D. FACIT manual: manual of the functional assessment
of chronic illness therapy (FACIT) measurement system.
Evanston: CORE, 1997.
16. Byrne BM. Structural equation modeling with EQS and EQS/
Windows. Thousand Oaks: Sage, 1994.
17. Long JS. Confirmatory factor analysis. In: Lewis-Beck MS,
ed. Factor analysis and related techniques. London: Sage,
1994.
18. Fayers PM, Machin D. Quality of life: assessment, analysis
and interpretation. Chichester: Wiley, 2000.
19. Fairclough DL. Design and analysis of quality of life studies
in clinical trials. London: Chapman & Hall, 2002.
20. Chow SC. Sample size calculation. In: Karlberg J, Tsang K,
eds. Introduction to clinical trials. Hong Kong: Clinical Trials
Centre, 1998: 267–86.
21. Feinstein AR. Multi-item ‘instruments’ vs Virginia Apgar’s
principles of clinimetrics. Arch Intern Med 1999; 159:
125–8.
22. Sprangers MAG. Quality-of-life assessment in oncology:
achievements and challenges. Acta Oncol 2002; 41: 229–37.
23. Bernhard J, Cella DF, Coates AS, et al. Missing quality of life
data in cancer clinical trials: serious problems and challenges.
Stat Med 1998; 17: 517–32.
24. Axelsson B, Sjoden PO. Assessment of quality of life in
palliative care: psychometric properties of a short questionnaire.
Acta Oncol 1999; 38: 229–37.