- Email: [email protected]
Use of Quality of Life Measures in Clinical Trials
Use of Quality of Life Measures in Clinical Trials Christine Eiser, BSc, PhD To be useful in clinical trials, quality of life (QOL) measures must meet basic criteria related to psychometric properties (reliability, validity, and sensitivity), length and acceptability to patients, and availability of parallel child and proxy versions. This paper examines the adequacy of several current measures, with special reference to childhood cancer. We identified 10 such measures of varying length, age-appropriateness, and content. We also discerned a number of obstacles to inclusion of QOL measures in trials. These include clinician bias, financial cost, and time required to inform families and to obtain informed consent from parents. We noted some methodological difficulties associated with measuring QOL over time. These include allowing for maturation and accounting for response shift, or children’s propensity to change their internal standards for assessing QOL. KEY WORDS:
clinical trials; proxy ratings; quality of life; self-report measures
Ambulatory Pediatrics 2004;4:395 399
O
ne of the most hopeful developments in pediatrics has been the steady progress medical research has made in treating chronic diseases; with each passing decade, life expectancy has improved for children with conditions such as cancer and cystic fibrosis. A sobering corollary to this fact is, however, that an increasing number of children are living with the restrictions imposed by these conditions—restrictions that inevitably detract from their quality of life. Often, the very treatments that allow survival also produce adverse physical consequences during the long term: survivors of childhood cancer, for example, may experience deficits in heart and lung function, compromised fertility, or educational and behavioral problems.1 Quality of life is therefore a central issue in considering chronic disease today and as such cannot be ignored by researchers who are designing clinical trials.2 Health-related quality of life (QOL) has nevertheless proved relatively difficult to define and measure. Some of the earliest attempts to sharpen the concept of QOL relied on measures of anxiety, depression, or physical symptoms. Others used simple indicators of function, such as ability to attend school. Although the World Health Organization originally articulated the concept of QOL in 1947,3 only in recent years have researchers attempted to define and measure QOL rigorously, especially in a pediatric context. QOL measures have potential value in a number of settings. These include evaluations of changes in treatment delivery, assessment of the illness experience from the
child’s point of view, and clinical trials.4 This paper will review the current status of relevant research, with particular reference to use of QOL measures as outcomes in clinical trials. Traditionally, evaluation of trials has focused on objective indices including survival rates and reduction of physical symptoms. There are several reasons for also including QOL. First, in many chronic conditions, current trials are not expected to register major differences in survival rates. Second, there may not be a simple relationship between physical symptoms measured by a doctor and how the patient feels: for example, children with diabetes cannot tell when their blood sugar is higher than recommended. Third, the goals of some trials may be to improve a child’s behavior or concentration, and therefore appropriate end points must be included. Fourth, for many families, improvements in the child’s daily functioning count more than clinical indices: in the diabetes example, the child’s sense of well-being means more than absolute blood-sugar levels. Understanding how treatment affects patient QOL may lead to greater insight into patient adherence.5 If patients themselves do not notice an improvement, they are unlikely to adhere to protocols. In acknowledging the potential value of QOL as an outcome measure, a number of funding agencies now require assessment of QOL in clinical trials. This requirement poses problems for trial designers relating to choice of a measure. It is necessary to ask how ‘‘good’’ currently available measures are, and furthermore how suitable and sensitive they are for clinical-trials work. We therefore need first to enumerate the required attributes of a measure of QOL suitable for use in trials, then assess how far current measures fulfill these requirements.
From the CR-UK Child and Family Research Group, Department of Psychology, University of Sheffield, Sheffield, UK. Address correspondence to Christine Eiser, BSc, PhD, CR-UK Child and Family Research Group, Department of Psychology, University of Sheffield, Western Bank, Sheffield S10 2TP, UK (e-mail: [email protected]). Received for publication July 25, 2003; accepted January 28, 2004. AMBULATORY PEDIATRICS Copyright q 2004 by Ambulatory Pediatric Association
REQUIREMENTS OF A MEASURE OF QOL FOR CLINICAL-TRIALS WORK We shall discuss four considerations relevant to clinical-trial design: whether the measure probes a specific dis-
395
Volume 4, Number 4 Supplement July–August 2004
396
Eiser
AMBULATORY PEDIATRICS
The Main Domains Assessed by Cancer-Specific Measures of Quality of Life* Domain Measure BASES7 Miami Pediatric Quality of Life Questionnaire8 Pediatric Cancer Quality of Life Inventory9 PedsQL10 Pediatric Oncology Quality of Life Scale11 Play Performance Scale for Children12 HUI13 Quality of Well-Being Scale14 Cancer Quality of Life15 Minnesota-Manchester Quality of Life Instrument16
Physical
Social
Psychological
Cognitive
Treatment Related
X ... X X X X X X X X
X X X X ... ... ... X X X
X X X X X ... X ... X X
... ... X ... ... ... X ... X X
X ... X ... ... ... X ... ... ...
*X indicates that the measure assesses the specific domain; ellipses indicate that the domain is not assessed; BASES, Behavioral, Affective, and Somatic Experience Scale; PedQL, Pediatric Quality of Life; and HUI, Health Utilities Index.
ease, what are the psychometric properties of the measure, how long it is, and who responds to the queries it poses. Disease-Specific versus Generic Measures Clinical researchers sometimes prefer disease-specific measures to generic measures, which are dismissed as lacking in sensitivity. Unfortunately, disease-specific measures tend to be less psychometrically sophisticated, and none may be available for the disease of interest. Development and validation of measures is time-consuming and costly, and measures for very rare conditions may prove impossible to develop. Psychometric Properties A measure must be reliable, valid, and sensitive.4 Especially for work in trials, the measure must be sufficiently sensitive to detect changes that might result from treatment. Length Length and acceptability of a measure to child and family are also important criteria. Measures need to be brief and also to address the issues of concern to families. Proxy Issues In choosing a measure of QOL, the researcher must select the most suitable respondent (parents, child, or both). In many ways, parents are inadequate respondents. Their perceptions of the child’s QOL are a reflection of their own well-being, with considerable evidence suggesting that parents who are depressed or have poor QOL project these views onto their child. Typically, parents’ ratings of the impact of illness are greater than those of the children themselves.6 Parents may also simply lack information. Children may deliberately not involve parents in some aspects of their lives. In other cases, parents are simply not there; they are unable to assess the effect of the illness on the child’s school life because they are not at school. Parent-completed measures are invaluable where children are too ill or young to answer for themselves.
DISEASE-SPECIFIC MEASURES Having reviewed many QOL measures for children, Eiser and Morse4 found that a greater number were available for cancer patients than for children with other conditions. To facilitate the choices of trial developers, we shall now devote some attention to these cancer-specific measures. Cancer-Specific Measures We have identified 9 measures developed specifically for assessment of QOL in children with cancer and one for use following bone-marrow transplant (Table).7 Of the 10 measures, 4 were developed for completion by parents only and had no parallel form for children to self-complete. Three measures included parallel forms for parents and child. The number of items ranged from 18 to 38.9 Most measures use standard Likert scales of 3, 4, or 5 points, but the test makers rarely justify their decisions concerning scale length. Some assessment of physical QOL is included in all measures. Social QOL is included in at least 39,10,15 as is cognitive functioning.9,13,15 In addition, some measures include unique domains: compliance,7 externalizing behavior,11 and body image.15 In the context of childhood cancer, trial developers have a relatively wide choice of measures available. Some may be less suitable given their length or complexity, whereas others lack parallel forms for child and parent completion. There are also differences in the content of measures, with resulting implications for the conclusions of any trial. Disease-Specific Measures for Diseases Other Than Cancer Disease-specific measures are also available for use in asthma,17 cystic fibrosis, and diabetes. Though the focus is typically on the child’s QOL, related measures have been introduced to assess the burden of care for parents.18 Such parent measures are of potential value in comparing different treatments. BARRIERS TO INCLUDING QOL MEASURES Although trials involving adult patients with cancer routinely include measures of QOL, trials involving pe-
AMBULATORY PEDIATRICS
diatric cancer patients do not. One may attribute such omissions in part to clinician prejudice against measures whose quality is suspect.19 Researchers have, however, advanced arguments for exclusion on other bases, including economics, maturation, and response shift. Economics Some researchers have protested that adding a QOL measure to the protocol would increase the monetary cost of the study. One could counter that modern techniques such as self-administered questionnaires whose responses can be scanned into a computer would minimize the cost associated with that part of the inquiry. A more fundamental point, however, is that omission of QOL measures often vitiates the meaning of the study, as one of the most obvious questions concerning an intervention is what effect it has on the patient’s sense of well-being. Indeed, this issue is so pervasive that the researcher may not recognize that they have made a commitment at the beginning of the study to report on quality of life and may accordingly forfeit coherence by leaving out the QOL measure in a misguided attempt at frugality. Furthermore, in studies where collaboration across research centers is necessary to achieve larger sample sizes required to enhance statistical power, omission of a measure by researchers at some centers would leave collaborators at other centers with too few data. As Pollock argues,19 these problems can be alleviated by raising the awareness of investigators concerning study objectives, pointing out the leverage that can be gained through collaboration across centers, and making sufficient funds available to ensure that trial results are consistent and meaningful. Another economic concern is the cost in time associated with obtaining informed consent. For informed participation, researchers should discuss with potential subjects the following considerations: 1) purpose of the research, 2) any personal benefits to the child of participating in the research, 3) the meaning of relevant research items, 4) the nature of the procedure, 5) potential harms and benefits, 6) the name of a researcher they can contact, 7) the name of the doctor responsible, and 8) procedures for withdrawing participation.20 Clearly, a shorter protocol would cut down on this time-consuming preparation; but we can only make the same reply to cost-in-time arguments as we did in the previous paragraph to cost-in-money arguments. Maturation Maturation poses both short- and long-term problems. In the short term, too-frequent administration of questionnaires can result in fatigue and learning. In the longer term, psychological and environmental changes can affect responses.21 Such changes wreak havoc with studies that intend to probe long-term effects but must rely on QOL measures applicable to only a narrow age range of children. Choosing a versatile measure and administering it sensitively is perhaps the best insurance against maturation effects; omitting the QOL measure altogether would reduce the value of the study and should therefore be done
Quality of Life Measures in Clinical Trials
397
only when there is no known way to include the measure without sacrificing scientific quality. Response Shift Whereas maturation factors emanate from the child’s physical and mental capabilities, which may relate overtly to the environment, response shift refers to more subtle, and also more private, changes in values and standards, including standards of functioning.22 More specifically, we may identify three types of response shift: 1) recalibration of scales comprising the yardsticks respondents use to gauge personal standards, 2) actual changes in values as measured by ranking of outcome domains, and 3) redefinition of the target construct.22 Interest in response shift has emerged from the frequent observation of discrepancies between clinical features of a disease and patients’ self-reports about their QOL. A potential explanation is that as a disease progresses, patients make a shift in choice of reference group. Thus, on diagnosis, the child with cancer is confronted with information about how life is going to be different (and more restricted compared with before). Consequently, self-ratings suggest poor QOL. In time, these children adjust to the difficulties or things seem more manageable than originally supposed, and also they meet others with the same or even more restricting illnesses. As their horizons widen, the children recognize that things could be worse, and against these new standards their self-ratings of QOL remain fairly stable. Nevertheless, these shifts in choice of comparison groups pose considerable problems for measuring QOL in trials. It is therefore important to determine how far any such response shift occurred and to assess its impact on ratings. Failure to recognize response shifts can lead to incomplete understanding of the outcome of a trial. USE OF QOL MEASURES IN CLINICAL TRIALS Although there seems to be more interest in developing QOL measures now, their application to clinical-trials research during the years has been anything but extensive. Indeed, only 3% of trials in pediatric oncology include a QOL measure.23 The way to ensure an increase in this percentage is to design QOL considerations into the study at its outset. For example, trials conducted to compare standard treatment with a new treatment or standard dosage with an altered dosage may seek primarily to detect a longer and more event-free survival with the new treatment or to establish that the new dosage reduces symptoms; but factoring QOL in as a secondary outcome would add valuable information to the comparison.24 Consider a case in which the new treatment or dosage is by hypothesis superior to standard remedies for some medical condition, such as diabetes, yet also gives rise to undesirable side effects, as when multiple daily insulin injections consume more time and care than twice-daily injections: How does one decide whether the results on balance favor the innovation? With the diabetes study, a QOL measure would facilitate comparison of the 2 frequencies of injection and thereby gauge whether the consequences of increased fre-
398
Eiser
AMBULATORY PEDIATRICS
quency were positive or negative.25 In another study, Harper et al26 compared continuous therapy with multiple short courses of cyclosporin over a 1-year period with respect to efficacy, safety, tolerability, and QOL. These investigators found that short-course treatment was adequate for some though not all patients; in most respects, including QOL, there was little difference between the samples after treatment. The researchers recommended tailoring the mode of treatment to the needs of the individual. A second type of experimental design compares a treatment with ‘‘watchful waiting’’ as opposed to an alternative medication or dosage. Here again, the inclusion of a QOL measure is crucial to help investigators distinguish effects attributable to the condition from those attributable to the treatment. The study by Rovers et al27 of ventilation tubes versus watchful waiting in children with persistent otitis media is an instance of this design. A third type of study sets out explicitly to enhance QOL. The controversy surrounding growth-hormone therapy (GHT) offers a case in point. Clinicians have administered GHT to children with idiopathic short stature28 and to children who have been treated for cancer or renal disease. Administration of GHT to children treated for brain tumors is intended to compensate for the low growth-hormone levels common in such patients while simultaneously improving QOL as a consequence of gains in height, muscle tone, body mass, and possibly also energy and vitality.29 On the other hand, the way children receive GHT is through injections, and this could pose problems: Many pediatric cancer patients have received extensive treatments already, and some will have developed needle fear. The adverse effects on QOL accruing from daily growth-hormone injections and frequent hospital visits will detract from the positive gains noted above; QOL measures will provide researchers with a best guess as to the net effects of the treatment and therefore an index to the value of the overall strategy.30 CONCLUSIONS The preceding text has indicated some of the difficulties researchers encounter in measuring children’s QOL. We have argued nevertheless that inclusion of QOL measures in clinical trials should be de rigueur—otherwise, a number of pressing issues concerning the effects of treatments can scarcely be discussed. Against this argument is the common retort that current QOL measures lack the versatility or specialized content necessary to probe children’s issues credibly. Unfortunately, this criticism is not without merit; hence the continuing need for new measures that offer parallel forms for children and adults and that assess change sensitively during the course of a trial. The current review will have achieved its end if it prompts new research initiatives into the effect of treatments on quality of life and into the optimal means of measuring quality of life in children. Mere length of survival of an experimental as opposed to a control group is a crude measure of medical efficacy. Today, researchers have, by thinking carefully and precisely about quality of
life, retrieved it from the ethereal realm of conjecture to place it on the relatively solid intellectual ground of detailed measurement and cogent inference; this breakthrough must now be exploited much more extensively in clinical trials involving pediatric subjects. ACKNOWLEDGMENT Christine Eiser was funded by Cancer Research-UK.
REFERENCES 1. Hawkins MM, Stevens MCG. The long term survivors. BMJ. 1996;52:898–923. 2. Craft A. Childhood cancer—mainly curable so where next? Acta Paediatrica. 2000;89:386–392. 3. WHO. World Health Organization Constitution. Geneva: World Health Organization; 1947. 4. Eiser C, Morse R. Quality of life measures in chronic diseases of childhood. Health Technol Assess Monographs. 2001:5(4): 1–156. 5. Abbott J, Dodd M, Bilton D, Webb AK. Treatment compliance in adults with cystic fibrosis. Thorax. 1994;49:115–20. 6. Ennett ST, DeVellis BM, Earp JA, et al. Disease experience and psychosocial adjustment in children with juvenile rheumatoid arthritis: children’s versus mothers’ reports. J Ped Psy. 1991; 16:557–568. 7. Phipps S, Hinds PS, Channell S, Bell GL. Measurement of behavioral, affective, and somatic responses to pediatric bone marrow transplantation: development of the BASES scale. J Ped Oncol Nursing. 1994;11:109–117. 8. Armstrong FD, Toledano SR, Miloslavich K, et al. The Miami Pediatric Quality of Life Questionnaire: Parent Scale. Int J Cancer. 1999;(suppl 12):11–17. 9. Varni JW, Katz ER, Seid M, et al. The Pediatric Cancer Quality of Life Inventory (PCQL). I. Instrument development, descriptive statistics, and cross-informant variance. J Behav Med. 1998; 21:179–204. 10. Varni JW, Seid M, Rode CA. The PedsQL: measurement model for the pediatric quality of life Inventory. Medical Care. 1999; 37:126–139. 11. Goodwin DAJ, Boggs SR, Graham-Pole J. Development and validation of the Pediatric Oncology Quality of Life Scale. Psychol Assess. 1994;6:321–328. 12. Lansky LL, List MA, Lansky SB, et al. Toward the development of a Play Performance scale for children (PPSC). Cancer. 1985; 56:1837. 13. Barr RD, Pai MR, Weitzman S, et al. A multiattributional approach to health-status measurement and clinical management illustrated by an application to brain-tumors in childhood. Int J Oncol. 1994;4:639–648. 14. Bradlyn AS, Harris CV, Warner JE, et al. An investigation of the validity of the quality of Well-Being Scale with pediatric oncology patients. Health Psychology. 1993;12:246–250. 15. Calaminus G, Weispach S, Teske C, Gobel U. Quality of life in children and adolescents with cancer. Klin Padiatr. 2000;212: 211–215. 16. Bhatia S, Jenney ME, Bogue M, et al. The Minneapolis-Manchester Quality of Life instrument: reliability and validity of the adolescent form. JCO. 2002;29:4692–4698. 17. French DJ, Christie MJ, Sowden A. The reproducibility of the Childhood Asthma questionnaires: measures of quality of life for children aged 4–16 years. Qual Life Res. 1994;3:215–224. 18. Juniper EF, Guyatt GH, Feeny DH, et al. Measuring quality of life in the parents of children with asthma. Qual Life Res. 1996; 5:27–34. 19. Pollock BH. Obstacles and opportunities for the use of healthrelated quality-of-life assessment in pediatric cancer clinical trials (discussion). Int J Cancer. 1999;12(suppl):151–153. 20. McIntosh N, Bates P, Brykczynska G, et al. Guidelines for the ethical conduct of medical research involving children. Royal
AMBULATORY PEDIATRICS
21.
22. 23.
24.
25.
College of Paediatrics, Child health: Ethics Advisory Committee. Arch Dis Child. 2000:82:177–182. Brossart DF, Clay DL, Willson VL. Methodological and statistical considerations for threats to internal validity in pediatric outcome data: response shift in self-report measures. J Ped Psy. 2002;27:97–107. Sprangers M, Hoogstren J. Pretesting effects in retrospective pretest-posttest designs. J Applied Psychol. 1989;74:265–272. Bradlyn AS, Harris CV, Spieth LE. Quality of life assessment in pediatric oncology: a retrospective review of phase III reports. Soc Sci Med. 1995;41:1463–1465. Equi A, Balfour-Lynn IM, Bush A, Rosenthal M. Long term azithromycin in children with cystic fibrosis: a randomised, placebo-controlled trial. Lancet. 2002;360:978–984. DAFNE Study Group. Training in flexible insulin management to enable dietary freedom in people with type 1 diabetes: dose
Quality of Life Measures in Clinical Trials
26.
27.
28.
29.
30.
399
adjustment for normal eating (DAFNE) randomised controlled trial. Br Med J. 2002;325:746. Harper JI, Ahmed I, Barclay G, et al. Cyclosporin for severe childhood atopic dermatitis: short course versus continuous therapy. Br J Dermatol. 2000;142:52–58. Rovers MM, Krabbe PFM, Straatman H, et al. Randomised controlled trial of the effect of ventilation tubes (grommets) on quality of life at age 1–2 years. Arc Dis Child. 2001;84:45–49. Theunissen NC, Kamp GA, Koopman HM, et al. Quality of life and self-esteem in children treated for idiopathic short stature. J Ped. 2003;140:493–495. Ogilivy-Stuart AL, Shalet SM. Growth and puberty after growth hormone treatment after irradiation for brain tumors. Arch Dis Child. 1995;73:141–146. Lustig RH, Hinds PS, Ringwald-Smith K, et al. Octreotide therapy of pediatric hypothalamic obesity: a double blind, placebo controlled trial. J Clin Endocrinol Metab. 2003;88:2587–2592.
clinical trials; proxy ratings; quality of life; self-report measures
Ambulatory Pediatrics 2004;4:395 399
O
ne of the most hopeful developments in pediatrics has been the steady progress medical research has made in treating chronic diseases; with each passing decade, life expectancy has improved for children with conditions such as cancer and cystic fibrosis. A sobering corollary to this fact is, however, that an increasing number of children are living with the restrictions imposed by these conditions—restrictions that inevitably detract from their quality of life. Often, the very treatments that allow survival also produce adverse physical consequences during the long term: survivors of childhood cancer, for example, may experience deficits in heart and lung function, compromised fertility, or educational and behavioral problems.1 Quality of life is therefore a central issue in considering chronic disease today and as such cannot be ignored by researchers who are designing clinical trials.2 Health-related quality of life (QOL) has nevertheless proved relatively difficult to define and measure. Some of the earliest attempts to sharpen the concept of QOL relied on measures of anxiety, depression, or physical symptoms. Others used simple indicators of function, such as ability to attend school. Although the World Health Organization originally articulated the concept of QOL in 1947,3 only in recent years have researchers attempted to define and measure QOL rigorously, especially in a pediatric context. QOL measures have potential value in a number of settings. These include evaluations of changes in treatment delivery, assessment of the illness experience from the
child’s point of view, and clinical trials.4 This paper will review the current status of relevant research, with particular reference to use of QOL measures as outcomes in clinical trials. Traditionally, evaluation of trials has focused on objective indices including survival rates and reduction of physical symptoms. There are several reasons for also including QOL. First, in many chronic conditions, current trials are not expected to register major differences in survival rates. Second, there may not be a simple relationship between physical symptoms measured by a doctor and how the patient feels: for example, children with diabetes cannot tell when their blood sugar is higher than recommended. Third, the goals of some trials may be to improve a child’s behavior or concentration, and therefore appropriate end points must be included. Fourth, for many families, improvements in the child’s daily functioning count more than clinical indices: in the diabetes example, the child’s sense of well-being means more than absolute blood-sugar levels. Understanding how treatment affects patient QOL may lead to greater insight into patient adherence.5 If patients themselves do not notice an improvement, they are unlikely to adhere to protocols. In acknowledging the potential value of QOL as an outcome measure, a number of funding agencies now require assessment of QOL in clinical trials. This requirement poses problems for trial designers relating to choice of a measure. It is necessary to ask how ‘‘good’’ currently available measures are, and furthermore how suitable and sensitive they are for clinical-trials work. We therefore need first to enumerate the required attributes of a measure of QOL suitable for use in trials, then assess how far current measures fulfill these requirements.
From the CR-UK Child and Family Research Group, Department of Psychology, University of Sheffield, Sheffield, UK. Address correspondence to Christine Eiser, BSc, PhD, CR-UK Child and Family Research Group, Department of Psychology, University of Sheffield, Western Bank, Sheffield S10 2TP, UK (e-mail: [email protected]). Received for publication July 25, 2003; accepted January 28, 2004. AMBULATORY PEDIATRICS Copyright q 2004 by Ambulatory Pediatric Association
REQUIREMENTS OF A MEASURE OF QOL FOR CLINICAL-TRIALS WORK We shall discuss four considerations relevant to clinical-trial design: whether the measure probes a specific dis-
395
Volume 4, Number 4 Supplement July–August 2004
396
Eiser
AMBULATORY PEDIATRICS
The Main Domains Assessed by Cancer-Specific Measures of Quality of Life* Domain Measure BASES7 Miami Pediatric Quality of Life Questionnaire8 Pediatric Cancer Quality of Life Inventory9 PedsQL10 Pediatric Oncology Quality of Life Scale11 Play Performance Scale for Children12 HUI13 Quality of Well-Being Scale14 Cancer Quality of Life15 Minnesota-Manchester Quality of Life Instrument16
Physical
Social
Psychological
Cognitive
Treatment Related
X ... X X X X X X X X
X X X X ... ... ... X X X
X X X X X ... X ... X X
... ... X ... ... ... X ... X X
X ... X ... ... ... X ... ... ...
*X indicates that the measure assesses the specific domain; ellipses indicate that the domain is not assessed; BASES, Behavioral, Affective, and Somatic Experience Scale; PedQL, Pediatric Quality of Life; and HUI, Health Utilities Index.
ease, what are the psychometric properties of the measure, how long it is, and who responds to the queries it poses. Disease-Specific versus Generic Measures Clinical researchers sometimes prefer disease-specific measures to generic measures, which are dismissed as lacking in sensitivity. Unfortunately, disease-specific measures tend to be less psychometrically sophisticated, and none may be available for the disease of interest. Development and validation of measures is time-consuming and costly, and measures for very rare conditions may prove impossible to develop. Psychometric Properties A measure must be reliable, valid, and sensitive.4 Especially for work in trials, the measure must be sufficiently sensitive to detect changes that might result from treatment. Length Length and acceptability of a measure to child and family are also important criteria. Measures need to be brief and also to address the issues of concern to families. Proxy Issues In choosing a measure of QOL, the researcher must select the most suitable respondent (parents, child, or both). In many ways, parents are inadequate respondents. Their perceptions of the child’s QOL are a reflection of their own well-being, with considerable evidence suggesting that parents who are depressed or have poor QOL project these views onto their child. Typically, parents’ ratings of the impact of illness are greater than those of the children themselves.6 Parents may also simply lack information. Children may deliberately not involve parents in some aspects of their lives. In other cases, parents are simply not there; they are unable to assess the effect of the illness on the child’s school life because they are not at school. Parent-completed measures are invaluable where children are too ill or young to answer for themselves.
DISEASE-SPECIFIC MEASURES Having reviewed many QOL measures for children, Eiser and Morse4 found that a greater number were available for cancer patients than for children with other conditions. To facilitate the choices of trial developers, we shall now devote some attention to these cancer-specific measures. Cancer-Specific Measures We have identified 9 measures developed specifically for assessment of QOL in children with cancer and one for use following bone-marrow transplant (Table).7 Of the 10 measures, 4 were developed for completion by parents only and had no parallel form for children to self-complete. Three measures included parallel forms for parents and child. The number of items ranged from 18 to 38.9 Most measures use standard Likert scales of 3, 4, or 5 points, but the test makers rarely justify their decisions concerning scale length. Some assessment of physical QOL is included in all measures. Social QOL is included in at least 39,10,15 as is cognitive functioning.9,13,15 In addition, some measures include unique domains: compliance,7 externalizing behavior,11 and body image.15 In the context of childhood cancer, trial developers have a relatively wide choice of measures available. Some may be less suitable given their length or complexity, whereas others lack parallel forms for child and parent completion. There are also differences in the content of measures, with resulting implications for the conclusions of any trial. Disease-Specific Measures for Diseases Other Than Cancer Disease-specific measures are also available for use in asthma,17 cystic fibrosis, and diabetes. Though the focus is typically on the child’s QOL, related measures have been introduced to assess the burden of care for parents.18 Such parent measures are of potential value in comparing different treatments. BARRIERS TO INCLUDING QOL MEASURES Although trials involving adult patients with cancer routinely include measures of QOL, trials involving pe-
AMBULATORY PEDIATRICS
diatric cancer patients do not. One may attribute such omissions in part to clinician prejudice against measures whose quality is suspect.19 Researchers have, however, advanced arguments for exclusion on other bases, including economics, maturation, and response shift. Economics Some researchers have protested that adding a QOL measure to the protocol would increase the monetary cost of the study. One could counter that modern techniques such as self-administered questionnaires whose responses can be scanned into a computer would minimize the cost associated with that part of the inquiry. A more fundamental point, however, is that omission of QOL measures often vitiates the meaning of the study, as one of the most obvious questions concerning an intervention is what effect it has on the patient’s sense of well-being. Indeed, this issue is so pervasive that the researcher may not recognize that they have made a commitment at the beginning of the study to report on quality of life and may accordingly forfeit coherence by leaving out the QOL measure in a misguided attempt at frugality. Furthermore, in studies where collaboration across research centers is necessary to achieve larger sample sizes required to enhance statistical power, omission of a measure by researchers at some centers would leave collaborators at other centers with too few data. As Pollock argues,19 these problems can be alleviated by raising the awareness of investigators concerning study objectives, pointing out the leverage that can be gained through collaboration across centers, and making sufficient funds available to ensure that trial results are consistent and meaningful. Another economic concern is the cost in time associated with obtaining informed consent. For informed participation, researchers should discuss with potential subjects the following considerations: 1) purpose of the research, 2) any personal benefits to the child of participating in the research, 3) the meaning of relevant research items, 4) the nature of the procedure, 5) potential harms and benefits, 6) the name of a researcher they can contact, 7) the name of the doctor responsible, and 8) procedures for withdrawing participation.20 Clearly, a shorter protocol would cut down on this time-consuming preparation; but we can only make the same reply to cost-in-time arguments as we did in the previous paragraph to cost-in-money arguments. Maturation Maturation poses both short- and long-term problems. In the short term, too-frequent administration of questionnaires can result in fatigue and learning. In the longer term, psychological and environmental changes can affect responses.21 Such changes wreak havoc with studies that intend to probe long-term effects but must rely on QOL measures applicable to only a narrow age range of children. Choosing a versatile measure and administering it sensitively is perhaps the best insurance against maturation effects; omitting the QOL measure altogether would reduce the value of the study and should therefore be done
Quality of Life Measures in Clinical Trials
397
only when there is no known way to include the measure without sacrificing scientific quality. Response Shift Whereas maturation factors emanate from the child’s physical and mental capabilities, which may relate overtly to the environment, response shift refers to more subtle, and also more private, changes in values and standards, including standards of functioning.22 More specifically, we may identify three types of response shift: 1) recalibration of scales comprising the yardsticks respondents use to gauge personal standards, 2) actual changes in values as measured by ranking of outcome domains, and 3) redefinition of the target construct.22 Interest in response shift has emerged from the frequent observation of discrepancies between clinical features of a disease and patients’ self-reports about their QOL. A potential explanation is that as a disease progresses, patients make a shift in choice of reference group. Thus, on diagnosis, the child with cancer is confronted with information about how life is going to be different (and more restricted compared with before). Consequently, self-ratings suggest poor QOL. In time, these children adjust to the difficulties or things seem more manageable than originally supposed, and also they meet others with the same or even more restricting illnesses. As their horizons widen, the children recognize that things could be worse, and against these new standards their self-ratings of QOL remain fairly stable. Nevertheless, these shifts in choice of comparison groups pose considerable problems for measuring QOL in trials. It is therefore important to determine how far any such response shift occurred and to assess its impact on ratings. Failure to recognize response shifts can lead to incomplete understanding of the outcome of a trial. USE OF QOL MEASURES IN CLINICAL TRIALS Although there seems to be more interest in developing QOL measures now, their application to clinical-trials research during the years has been anything but extensive. Indeed, only 3% of trials in pediatric oncology include a QOL measure.23 The way to ensure an increase in this percentage is to design QOL considerations into the study at its outset. For example, trials conducted to compare standard treatment with a new treatment or standard dosage with an altered dosage may seek primarily to detect a longer and more event-free survival with the new treatment or to establish that the new dosage reduces symptoms; but factoring QOL in as a secondary outcome would add valuable information to the comparison.24 Consider a case in which the new treatment or dosage is by hypothesis superior to standard remedies for some medical condition, such as diabetes, yet also gives rise to undesirable side effects, as when multiple daily insulin injections consume more time and care than twice-daily injections: How does one decide whether the results on balance favor the innovation? With the diabetes study, a QOL measure would facilitate comparison of the 2 frequencies of injection and thereby gauge whether the consequences of increased fre-
398
Eiser
AMBULATORY PEDIATRICS
quency were positive or negative.25 In another study, Harper et al26 compared continuous therapy with multiple short courses of cyclosporin over a 1-year period with respect to efficacy, safety, tolerability, and QOL. These investigators found that short-course treatment was adequate for some though not all patients; in most respects, including QOL, there was little difference between the samples after treatment. The researchers recommended tailoring the mode of treatment to the needs of the individual. A second type of experimental design compares a treatment with ‘‘watchful waiting’’ as opposed to an alternative medication or dosage. Here again, the inclusion of a QOL measure is crucial to help investigators distinguish effects attributable to the condition from those attributable to the treatment. The study by Rovers et al27 of ventilation tubes versus watchful waiting in children with persistent otitis media is an instance of this design. A third type of study sets out explicitly to enhance QOL. The controversy surrounding growth-hormone therapy (GHT) offers a case in point. Clinicians have administered GHT to children with idiopathic short stature28 and to children who have been treated for cancer or renal disease. Administration of GHT to children treated for brain tumors is intended to compensate for the low growth-hormone levels common in such patients while simultaneously improving QOL as a consequence of gains in height, muscle tone, body mass, and possibly also energy and vitality.29 On the other hand, the way children receive GHT is through injections, and this could pose problems: Many pediatric cancer patients have received extensive treatments already, and some will have developed needle fear. The adverse effects on QOL accruing from daily growth-hormone injections and frequent hospital visits will detract from the positive gains noted above; QOL measures will provide researchers with a best guess as to the net effects of the treatment and therefore an index to the value of the overall strategy.30 CONCLUSIONS The preceding text has indicated some of the difficulties researchers encounter in measuring children’s QOL. We have argued nevertheless that inclusion of QOL measures in clinical trials should be de rigueur—otherwise, a number of pressing issues concerning the effects of treatments can scarcely be discussed. Against this argument is the common retort that current QOL measures lack the versatility or specialized content necessary to probe children’s issues credibly. Unfortunately, this criticism is not without merit; hence the continuing need for new measures that offer parallel forms for children and adults and that assess change sensitively during the course of a trial. The current review will have achieved its end if it prompts new research initiatives into the effect of treatments on quality of life and into the optimal means of measuring quality of life in children. Mere length of survival of an experimental as opposed to a control group is a crude measure of medical efficacy. Today, researchers have, by thinking carefully and precisely about quality of
life, retrieved it from the ethereal realm of conjecture to place it on the relatively solid intellectual ground of detailed measurement and cogent inference; this breakthrough must now be exploited much more extensively in clinical trials involving pediatric subjects. ACKNOWLEDGMENT Christine Eiser was funded by Cancer Research-UK.
REFERENCES 1. Hawkins MM, Stevens MCG. The long term survivors. BMJ. 1996;52:898–923. 2. Craft A. Childhood cancer—mainly curable so where next? Acta Paediatrica. 2000;89:386–392. 3. WHO. World Health Organization Constitution. Geneva: World Health Organization; 1947. 4. Eiser C, Morse R. Quality of life measures in chronic diseases of childhood. Health Technol Assess Monographs. 2001:5(4): 1–156. 5. Abbott J, Dodd M, Bilton D, Webb AK. Treatment compliance in adults with cystic fibrosis. Thorax. 1994;49:115–20. 6. Ennett ST, DeVellis BM, Earp JA, et al. Disease experience and psychosocial adjustment in children with juvenile rheumatoid arthritis: children’s versus mothers’ reports. J Ped Psy. 1991; 16:557–568. 7. Phipps S, Hinds PS, Channell S, Bell GL. Measurement of behavioral, affective, and somatic responses to pediatric bone marrow transplantation: development of the BASES scale. J Ped Oncol Nursing. 1994;11:109–117. 8. Armstrong FD, Toledano SR, Miloslavich K, et al. The Miami Pediatric Quality of Life Questionnaire: Parent Scale. Int J Cancer. 1999;(suppl 12):11–17. 9. Varni JW, Katz ER, Seid M, et al. The Pediatric Cancer Quality of Life Inventory (PCQL). I. Instrument development, descriptive statistics, and cross-informant variance. J Behav Med. 1998; 21:179–204. 10. Varni JW, Seid M, Rode CA. The PedsQL: measurement model for the pediatric quality of life Inventory. Medical Care. 1999; 37:126–139. 11. Goodwin DAJ, Boggs SR, Graham-Pole J. Development and validation of the Pediatric Oncology Quality of Life Scale. Psychol Assess. 1994;6:321–328. 12. Lansky LL, List MA, Lansky SB, et al. Toward the development of a Play Performance scale for children (PPSC). Cancer. 1985; 56:1837. 13. Barr RD, Pai MR, Weitzman S, et al. A multiattributional approach to health-status measurement and clinical management illustrated by an application to brain-tumors in childhood. Int J Oncol. 1994;4:639–648. 14. Bradlyn AS, Harris CV, Warner JE, et al. An investigation of the validity of the quality of Well-Being Scale with pediatric oncology patients. Health Psychology. 1993;12:246–250. 15. Calaminus G, Weispach S, Teske C, Gobel U. Quality of life in children and adolescents with cancer. Klin Padiatr. 2000;212: 211–215. 16. Bhatia S, Jenney ME, Bogue M, et al. The Minneapolis-Manchester Quality of Life instrument: reliability and validity of the adolescent form. JCO. 2002;29:4692–4698. 17. French DJ, Christie MJ, Sowden A. The reproducibility of the Childhood Asthma questionnaires: measures of quality of life for children aged 4–16 years. Qual Life Res. 1994;3:215–224. 18. Juniper EF, Guyatt GH, Feeny DH, et al. Measuring quality of life in the parents of children with asthma. Qual Life Res. 1996; 5:27–34. 19. Pollock BH. Obstacles and opportunities for the use of healthrelated quality-of-life assessment in pediatric cancer clinical trials (discussion). Int J Cancer. 1999;12(suppl):151–153. 20. McIntosh N, Bates P, Brykczynska G, et al. Guidelines for the ethical conduct of medical research involving children. Royal
AMBULATORY PEDIATRICS
21.
22. 23.
24.
25.
College of Paediatrics, Child health: Ethics Advisory Committee. Arch Dis Child. 2000:82:177–182. Brossart DF, Clay DL, Willson VL. Methodological and statistical considerations for threats to internal validity in pediatric outcome data: response shift in self-report measures. J Ped Psy. 2002;27:97–107. Sprangers M, Hoogstren J. Pretesting effects in retrospective pretest-posttest designs. J Applied Psychol. 1989;74:265–272. Bradlyn AS, Harris CV, Spieth LE. Quality of life assessment in pediatric oncology: a retrospective review of phase III reports. Soc Sci Med. 1995;41:1463–1465. Equi A, Balfour-Lynn IM, Bush A, Rosenthal M. Long term azithromycin in children with cystic fibrosis: a randomised, placebo-controlled trial. Lancet. 2002;360:978–984. DAFNE Study Group. Training in flexible insulin management to enable dietary freedom in people with type 1 diabetes: dose
Quality of Life Measures in Clinical Trials
26.
27.
28.
29.
30.
399
adjustment for normal eating (DAFNE) randomised controlled trial. Br Med J. 2002;325:746. Harper JI, Ahmed I, Barclay G, et al. Cyclosporin for severe childhood atopic dermatitis: short course versus continuous therapy. Br J Dermatol. 2000;142:52–58. Rovers MM, Krabbe PFM, Straatman H, et al. Randomised controlled trial of the effect of ventilation tubes (grommets) on quality of life at age 1–2 years. Arc Dis Child. 2001;84:45–49. Theunissen NC, Kamp GA, Koopman HM, et al. Quality of life and self-esteem in children treated for idiopathic short stature. J Ped. 2003;140:493–495. Ogilivy-Stuart AL, Shalet SM. Growth and puberty after growth hormone treatment after irradiation for brain tumors. Arch Dis Child. 1995;73:141–146. Lustig RH, Hinds PS, Ringwald-Smith K, et al. Octreotide therapy of pediatric hypothalamic obesity: a double blind, placebo controlled trial. J Clin Endocrinol Metab. 2003;88:2587–2592.