Assessing Quality of Life in Cancer Patients
Assessing Quality of Life in Cancer Patients
, 1989
The value of measuring quality of life has been increasingly 1988)) but considered that they had not been calibrated in
recognised. It should identify and better describe the cancer patients. The clinical relevance of scores on these
damaging effects of the disease or its treatment. This should scales was, therefore, difficult to interpret and the instru-
help the physician and the patient to choose more easily ments are lengthy.
between alternative treatment options. In controlled clinical
trials, direct measurement of quality of life is an important
outcome measure. When there are no differences in survival 2. Assessment criteria
it will be possible to recommend a treatment which produces
the best quality of life. However, if one treatment produces We agreed on a set of key questions to permit detailed
satisfactory quality of life but unsatisfactory survival then assessment of available instruments.
very difficult value judgements are likely to remain whatever
methods for measurement are developed. (a) Function
Interest in measuring quality of life has increased during
the 1980s. In 1978, Bardelli and Saracci reported that less (i) Is it designed for use with specific or different types
than 5% of papers in major cancer journals measured any of cancer?
aspect of quality of life. Since then there has been a major (ii) Which dimensions of quality of life does it measure?
increase, perhaps reflecting the increasing recognition of the
limitations and potential hazards of modern cancer treat- (b) Format
ment and changes in social attitudes. Reviews of this field (i) How many items does it contain?
have been conducted by Fayers & Jones (1983), Holland (ii) How well laid out are they?
(1984), De Haes & Van Knippenberg (1985), Ventafridda et (iii) Does it employ a linear analogue, categorical or other
al. (1986), Clark & Fallowfield (1986), Selby & Robertson method of scaling?
(1987), McDowell & Newell (1987) and Walker & Rosser (iv) Are there alternative forms?
(1988). (v) What time period does it cover (e.g. the past few days
The widespread feeling that quality bf life should be or last month).
measured in clinical trials has led to requests for guidance (vi) Does it require patients to refer to their own baseline
from the Cancer Therapy Committee of the Medical (e.g. 'no more than usual' to 'very much more than
Research Council and other bodies as to the best available usual') or to absolute levels ('not at all' to 'very
methods at present. The Working Party was convened to much').
review existing measures of quality of life in patients with (c) Administration
cancer and advise the Cancer Therapy Committee about
their use within clinical trials and thus provide a source of (i) How easy is it to administer?
advice for all clinical trialists. (ii) Is training needed?
(iii) How long does it take to complete?
(iv) Are all the items easily understood by patients?
1. Background (v) Are all the items acceptable or do any cause distress?
(vi) Can the questionnaire/scale be administered by
We agreed to a pragmatic approach to quality of life and computer?
decided that assessment would cover these dimensions:
symptoms due to cancer; adverse effects of treatmeant; (d) Scoring
physical functioning; social interaction; psychological adjust- (i) How easy is it to score?
ment; sexual functioning; and body image. Any measure (ii) What scores does it generate (e.g. an overall score,
recommended must be capable of assessing one or more of scores on specific sub-scales, scores on each item)?
these dimensions in a reliable and valid way and be able to (iii) Do population norms exist?
be completed by the patient within 10 minutes, be easy to (iv) How easy is it to analyse the responses statistically?
administer, easy to score and reflect changes over time.
We recognised that no scale would fulfil the needs of all (e) Structure
clinical trials. So, we would try to recommend core question- (i) How were the items within the questionnaire/scale
naires which could be applicable to most trials. Additional obtained (e.g. did they come from interviews with
items could then be added to reflect specific diseases or patients, relatives, health professionals, or existing
treatments. We acknowledged that clinicians wanted scales scales)?
which were short and simple but were concerned that these (ii) Do the items form discrete sub-scales?
might be incapable of measuring quality of life in a valid (iii) Are the sub-scales based on factor analysis of the
and reliable way. We recognised that general instruments items?
already existed which purported to measure the impact of (iv) Has the factor structure been replicated? With what
physical illness (like the Nottingham Health Profile results?
(McEwen, 1988) and Sickness Impact Profile (Bergner, (v) Do the items which load highly on one sub-scale
load on others?
(vi) How much does each item contribute to the overall
The working party included: N. Bleehen, K. Calman, P. Fayers, L. score?
Freedman, S. Gore, T. McElwain, T. Priestman, A. Smith, H. (vii) Are shorter forms available?
Sutton and L. Wallace.
Correspondence: P. Maguire, University of Manchester, Department (f) Clinical usage
of Psychiatry, West Didbury, Manchester M20 8LR, UK.
Received 4 March 1989. (i) Which patient groups have completed the question-
naire?
438 P. MAGUIRE & P. SELBY
(ii) Has it been used to measure change over time? (At 5. Main comments
what time intervals? With what response rate?)
(iii) Has it been used in clinical trials? (a) Global measures
(g) Reliability (i) Gough's scale. Patients are required to rate a 10cm line
in terms of 'How would you rate your feeling of well-being
(i) Has test-retest reliability been evaluated? today?' The authors argue that it performs as well as more
(ii) Has the self-rating version been compared with an complex scales (The Spitzer QL index (objective and
observer rating version? subjective versions), Baum and Priestman LASA system).
(iii) Has it been compared with other quality of life This conclusion is erroneous since it is based on an over-
measures? reliance on correlation coefficients and a failure to examine
(iv) Are data on split-half reliability available? areas of significant disagreement between different measures.
No data are given about how well this scale measures
(h) Validity changes over time. Nor does this global rating of well-being
(i) Does the questionnaire/scale measure accurately: give information about the relative contributions of the
component dimensions. It was not considered a serious
Changes over time. contender despite its simplicity.
Differences between groups:
by stage of disease
by mode of treatment. (ii) Rosser and Kind's disability/distress matrix. This
(ii) How well does it correlate with the findings of trained method requires patients with physical or psychiatric illness
interviewers who administer standardised interviews and health care professionals to evaluate 29 combinations of
to assess physical, social and psychological disability (O=no disability to 8=unconscious) and distress
functioning (i.e. a 'gold standard' assessment)? (no to severe distress). It involves a lengthy interview of
between I' and 4 hours, which is acknowledged to be
'potentially distressing and traumatic'. Scores are presented
3. Instruments assessed as group medians and there is no interest in the scores of
individual patients since this measure seeks to determine a
These included: consensus view of the relative importance of differing states
of disability and distress. While data about such views could
(a) Global measures inform decisions about priorities within health care they do
not enable quality of life to be measured within clinical
(i) Gough's visual analogue scale (Gough et al., 1983). trials.
(ii) Rosser and Kind's distress/disability matrix (Rosser &
Kind, 1978). (iii) Quality of life adjusted years index (QUALYS). This
(iii) Quality of life adjusted years (QUALYS) (Williams, measure is designed to take account of the quality as well as
1985). the duration of survival in assessing the outcome of health
care procedures and services. Unfortunately, the judgements
(b) Performance indices on which it rests are drawn from small samples of respon-
(i) Karnofsky et al. (1948). dents who were chosen arbitrarily. There is no evidence that
(ii) ECOG, Eastern Co-operative Oncology Group these judgements bear much relationship to real judgements
(Zubrod et al., 1960). made by patients with cancer when facing decisions about
(iii) Katz activities of daily living (Katz & Akbom, 1976). the utility of different treatments set against the possible
(iv) World Health Organization scales (WHO, 1979). adverse effects on their lives. Thus, it is not known what the
trade-offs would need to be in terms of change in quality of
(c) Scales measuring several dimensions life before a patient or clinician elected to undergo or reject
a given treatment.
(i) Iszak and Medalie index (1971). While the QUALY index is important conceptually and
(ii) Priestman and Baum (1976) linear self assessment could be used to compare different treatments within a trial,
system. work is needed to determine the judgements of real patients
(iii) Functional living index for cancer (Schipper et al., in real predicaments.
1984). Meanwhile, the concept is already being misapplied and
(iv) Ontario Cancer Institute quality of life questionnaire abused. For example, clinicians are being asked to justify a
(Selby et al., 1984). new cancer drug on the basis of the QUALY index without
(v) Padilla QL questionnaire (Padilla et al., 1981). any attempt to obtain the relevant data from patients and
(vi) QL index (Spitzer et al., 1981). relatives.
(vii) EORTC QL questionnaire (Aaronson et al., 1986).
(viii) Rotterdam symptom check list (Haes et al., 1986).
(b) Indices of physical performance
(d) Psychological dimension (i) Karnofsky index. The Karnofsky index is a valid if
(i) Hospital anxiety and depression scale (Zigmong & crude predictor of survival. It emphasises physical perfor-
Snaith, 1983). mance and dependency. Scores are weighted heavily on the
(ii) General health questionnaire (Goldberg & Hillier, physical dimensions of quality of life. Consequently,
1979). Karnofsky scores can give a misleading picture of the social
and psychological dimensions. It has limited reliability in
routine clinical application. While its accuracy in measuring
4. Mode of assessment performance can be improved by training the assessors this
would still fail to measure quality of life adequately, particu-
Each questionnaire/scale was assessed independently by at larly variation in psychological status.
least two members of the working party using the agreed
criteria and relevant literature. These assessments were (ii) ECOG. This provides a five-point scale of performance
circulated to the remaining members of the working party in contrast to the 10-point Karnofsky scale. It is simple to
and then discussed to determine the relative advantages and use and has a precise but narrow function. It is less easy to
disadvantages of each measure. use and interpret than the comparable WHO scale.
ASSESSING QUALITY OF LIFE 439
(iii) WHO scale. This measures physical performance on a monitoring the impact of treatment regimes over time. They
five-point scale. It is easy to use and interpret but has a can tap all dimensions, are simple to administer and score.
limited function. However, the time periods covered are often unstated or
vague. It is difficult to interpret the clinical significance of
(c) Scales measuring several dimensions the scores. Parametric methods of statistical analyses are
(i) Iszack and Medalie ability index. This provides a series suspect. There is no evidence that they are superior to
of simple rating scales to measure physical, social and categorical rating methods.
psychological variables. These are tailored to specific cancers
and treatments and are not generalisable across different (vi) Spitzer QL index. This index measures five areas,
cancers or different treatments. They seem of most use in including physical activity, daily living, perception of own
helping clinicians assess rehabilitation needs and monitoring health, support from family and friends, and outlook on life.
outcome in relation to specific cancers. The scores are summed to produce an overall score of
quality of life. It has the attraction of being simple, easy and
(ii) Baum and Priestman LASA system. Their set of linear quick to administer. It is available in clinician and patient
analogue scales originally covered 10 aspects: well being, rated versions. It was carefully developed by reference to
mood, activity, pain, nausea, appetite, housework, social patients with cancer, patients suffering from other chronic
activity, anxiety and help from treatment. It has been used in diseases, their relatives, people who were free of disease,
patients with cancer of the breast and appears of most value doctors, nurses and social workers. Unfortunately, the reduc-
in picking up strong treatment effects over time. They tion to five aspects of quality of life reflects an over-reliance
expanded the scale to 25 items and included more relating to on correlations between the different items that go up to
psychological aspects. It has not been compared with other make each dimension. Consequently it does not do justice to
QL methods or with gold standard interviews. Consequently, the different aspects of quality of life. Nor is there evidence
its validity and reliability have still to be determined. But it that this index has yet been used successfully within clinical
seems to detect change over time and to distinguish between trials to reflect change over time. Recent studies have found
different treatments. a poor correlation between the clinician and patient rated
versions. They have suggested that the overall score is
(iii) Functional living index for cancer. The FLIC includes unduly affected by clinicians' judgements about the extent of
22 items covering physical symptoms, mood, physical the cancer and associated physical disability.
activity, work, and social interaction. Items are presented in
a linear analogue format. Each line is divided into seven and (vii) EORTC questionnaire. This questionnaire was
covers 'now' or the 'past two weeks'. The items derive from designed for heterogeneous groups of cancer patients. It
a pool suggested by a panel of patients and health pro- seeks to measure key dimensions, including symptoms of
fessionals involved in cancer care. The index consists of a disease, side-effects of treatment, physical functioning,
number of sub-scales which are derived from factor analysis psychological distress, social interaction, sexuality, body
and have been replicated. It is easy to use, administer and image and satisfaction with medical care. It is a self adminis-
score. It has been validated in a cross-sectional study of tered scale and the format varies according to the dimension
patients with different stages of cancer and on different being addressed. Items on physical activity comprise 'yes or
treatments. no' responses whereas questions on symptoms use a categori-
The word cancer is used in several items and has dis- cal form asking for answers ranging from 'not at all' to 'very
tressed patients. Validity has been tested only by comparing much'. The checklist of symptoms assessed varies according
different groups and by correlation with other self-rating to the type of cancer being studied. It is designed for
instruments. Performance in detecting changes over time has administration by a nurse within clinics and takes up to 10
yet to be assessed. Nor has it been assessed or calibrated minutes to complete. It is easy to score and produces both
against 'gold standard' interviews. Consequently, scores on an overall score and scores on each sub-scale. Unfortunately
the component sub-scales are difficult to interpret. There are norms do not yet exist but work is underway to establish
so few items on each dimension that its ability to measure norms and check reliability and validity.
significant changes over time may be poor. For example, few While the scale was carefully developed on the basis of
physical symptoms are included. existing measures, its reliability and validity have yet to be
These factors prevented the working party from recom- properly assessed. It has still to be checked against 'gold
mending its use in clinical trials despite the attractiveness of standard' interviews to see how well it performs. The sub-
simplicity and ease of use. scales and factor structure have yet to be replicated. Thus,
although it is promising it is still being developed and is
(iv) Ontario Cancer Institute quality of life scale. The OCI probably not ready for inclusion in large scale clinical trials.
scale comprises 32 linear analogue items which cover all the Initial feedback suggests that the acceptance rate is variable
required dimensions. They are derived from the sickness (ranging between 33 and 90%) but it appears to reflect
impact profile and items relating to specific cancer sites. It is changes over time as well as differences between different
easy to use, understand and score. It distinguishes well diseases and treatments. A major validation study is under-
between patients with breast cancer who are on different way to address the questions of how well it performs within
treatments and are at different stages of their disease. But major clinical trials.
the author feels that while it is useful in a research context it
is not yet ready for use in clinical trials. It has not been (viii) The Rotterdam symptom checklist. This was carefully
tested in other patients except those with cancer of the derived from a pool of items based on existing question-
ovary. The factor structure has still to be replicated. naires, together with checklists from interviews with cancer
patients.
(v) Padilla QL scale. This is a short 14 item scale contain- It comprises 38 items and each item is rated on a four-
ing three sub-scales. It covers the 'present time' and takes point scale (O=not at all, 3=very much). It is well laid out
less than 5 minutes to administer. It is easy to score. and asks respondents to indicate how much they have
Unfortunately, data on reliability and validity are lacking experienced particular symptoms over the last week. It is
and there are too few items to reflect changes on key readily administered by nurses and takes 5-10 minutes to
dimensions. It appears inferior to the FLIC and OCI Scales. complete. All items appear understandable and acceptable. It
is easy to score and produces distinct sub-scale scores.
Linear analogue scales in general can be used to map out It provides two main sub-scales measuring physical and
differences between different illnesses but are most valid in psychological dimensions. It is considered to be a good, clear
440 P. MAGUIRE & P. SELBY
References
AARONSON, N.K., BAKKER, W., STEWART, A.L. and 5 others McEWEN, J. (1988). The Nottingham Health Profile. In Quality of
(1986). A multidimensional approach to the measurement of Life: Assessment and Application, Walker, S.R. & Rosser, R.M.
quality of life in lung cancer clinical trials. In Quality of Life in (eds) p. 95. MTP Press: Lancaster.
Cancer, Aaronson, N.K. et al. (eds) p. 171. Raven Press: New PADILLA, G., PRESANT, C.A., GRANT, M. and 4 others (1981).
York. Assessment of quality of life in cancer patients. Proc. Am. Assoc.
BERGNER, M. (1988). Development, testing and use of the sickness Cancer Res., 22, 397.
impact profile. In Quality of Life: Assessment and Application, PRIESTMAN, T.J. & BAUM, M. (1976). Evaluation of quality of life in
Walker, S.R. & Rosser, R.M. (eds) p. 79. MTP Press: Lancaster. patients receiving treatment for advanced cancer. Lancet, i, 899.
CLARK, A. & FALLOWFIELD, L.J. (1986). Quality of life measure- ROSSER, R.M. & KIND, P. (1978). A scale of valuations of states of
ments in patients with malignant disease: a review. J. R. Soc. illness; is there a social consensus? Int. J. Epidemiol., 7, 347.
Med., 79, 165. SCHIPPER, H., CLINCH, J. McMURRAY, A. & LEVITT, M. (1984).
DE HAES, J.C.J.M. & VAN KNIPPENBERG, F.C.E. (1985). The quality Measuring the quality of life of cancer patients. The functional
of life of cancer patients: a review of the literature. Soc. Sci. living index-cancer: development and validation. J. Clin. Oncol.,
Med., 20, 809. 2, 472.
DE HAES, J.C.J.M., VAN OSTROM, M.A. & WELVAART, K. (1986). The SELBY, P.J., CHAPMAN, J.A., ETAZADI, AMOLI J., DALLEY, D. &
effect of radical and conserving surgery on the quality of life of BOYD, N.F. (1984). The development of a method for assessing
early breast cancer patients. Eur. J. Surg. Oncol., 12, 337. the quality of life of cancer patients. Br. J. Cancer, 50, 13.
FAYERS, P.M. & JONES, D.R. (1983). Measuring and analysing SELBY, P. & ROBERTSON, B. (1987). Measurement of quality of life
quality of life in cancer clinical trials. Stat. Med., 2, 429. in patients with cancer. Cancer Surveys, 6, 521.
GOLDBERG, D. & HILLIER, V.F. (1979). A scaled version of the SPITZER, W.O., DOBSON, A.J., HALL, J. and 5 others (1981). Measur-
General Health Questionnaire. Psychol. Med., 31, 139. ing the quality of life of cancer patients. J. Chron. Dis., 34, 595.
GOUGH, I.R., FURNIVAL, C.M., SCHILDER, L. & GROVE, N. (1983). VENTAFRIDDA, V., VAN DAM, F.S.A.M., YANCIK, R. &
Assessment of the quality of life of patients with advanced TAMBURINI, M. (eds) (1986). Assessment of Quality of Life and
cancer. Eur. J. Cancer Clin. Oncol., 19, 1161. Cancer Treatment. Elsevier: Amsterdam.
HOLLAND, J.C.B. (ed.) (1984). Proceedings of, Conference on WALKER, S.R. & ROSSER, R.M. (eds) (1988). Quality of Life:
Research Methodology in Psychological Oncology. Cancer, 53, Assessment and Application. MTP Press: Lancaster.
suppl., 2217. WILLIAMS, A.H. (1985). Economics of coronary artery by-pass
ISZACK, F. & MEDALIE, J.H. (1971). Comprehensive follow up of grafts. Br. Med. J., 291, 326.
carcinoma patients. J. Chron. Dis., 24, 179. WORLD HEALTH ORGANIZATION (1979). Handbook for Reporting
KARNOFSKY, D.A., ABELMANN, W.H., CRAVER, L.F. & Results of Cancer Treatment. WHO: Geneva.
BURCHENAL, J.H. (1948). The use of nitrogen mustards in the ZIGMOND, A.S. & SNAITH, R.P. (1983). The hospital anxiety and
palliative treatment of carcinoma. Cancer, 1, 634. depression scale. Acta Psychiatr. Scand., 67, 361.
KATZ, S. & AKPOM, C.A. (1976). A measure of primary sociobio- ZUBROD, C.G., SCHNEIDERMAN, M., FREI, E. and 16 others (1960).
logical functions. Int. J. Health Serv., 6, 493. Appraisal of methods for the study of chemotherapy in man. J.
McDOWELL, I. & NEWELL, C. (1987). Measuring Health. A Guide to Chron. Dis., 11, 7.
Rating Scales and Questionnaires. Oxford University Press:
Oxford.