Measuring health-related quality of life in pediatric oncology patients: A brief commentary on the state of the art of measurement and application (discussion)

Historically, the conceptual framework for measuring health outcomes in children and adolescents can be traced to the late 1970s and early 1980s (Starfield, 1974, 1987; Grave and Pless, 1976; Eisen et al., 1979; Walker and Richmond, 1985). It was not until fairly recently, however, that important psychometric evidence concerning the underlying structure of a core set of comprehensive, generic health-assessment measures became available (Feeny et al., 1992, 1995; Starfield et al., 1993, 1995; Kurtin et al., 1994; Landgraf and Abetz, 1996, 1997, 1998; Landgraf et al., 1996, 1998). Despite these advances in instrument development, fundamental issues related to measuring and interpreting the health-related quality of life (HRQL) of children and adolescents and their families remain largely unanswered (Landgraf and Abetz, 1996). It is not surprising, then, that key issues which reflect the current state of the field emerged during the International Workshop on assessing the HRQL of children surviving cancer. My purpose is to address these broad principles briefly within the context of pediatric oncology and the reports presented at the workshop. Traditionally, clinicians have relied on personal data-gathering techniques, such as the history interview, to diagnosis and manage the health and well-being of patients. While this technique is still very useful, the changing marketplace has placed many constraints on the provider–patient relationship. Since its inception, the field of health outcomes has sought to augment this relationship by providing clinicians with a standardized way of collecting a comprehensive pool of information from patients that reflects how disease and illness may affect everyday functioning and well-being. The current standard is to solicit information from patients using paper-and-pencil instruments that have been developed carefully from focus groups, using both closed and open-ended interviews, reviewed by experts and members of the target population (e.g., cancer patients and their families) and evaluated using empirically grounded methods of analysis (e.g., confirmatory and exploratory factor analyses). Thus, both qualitative and quantitative methods are employed in the design and validation of instruments. Instrument development is an iterative process, and each new application provides evidence of its validity. At the outset, it is imperative that instruments for use with children and adolescents be developed using an approach grounded in the principles of child development. Other characteristics that should be identified include whether the instrument is designed for use (i) in a specific condition/disease group or general sample, (ii) with a broad or narrowly defined age range of children, (iii) in chronic or acute settings, (iv) by parent completion or child self-report or (v) with English speakers of a specified reading comprehension level. Additionally, it has been argued that examination of socio-demographic variables and their potential influence on the measurement characteristics of an instrument should be an integral part of the development and validation process (Landgraf and Abetz, 1998). Aside from potentially affecting measurement properties, latent variables such as age, gender, marital status, ethnicity and parental reports of their own health and well-being may influence health outcomes also. Crom et al. (1999) note, however, the emergence of conflicting evidence concerning the predictive validity of these salient variables among long-term survivors of pediatric solid tumors. Cancer-specific instruments were used in 3 of the 4 communications presented during the first panel of the workshop (Crom et al., 1999; Armstrong et al., 1999; Phipps et al., 1999). The single exception was work by van Schaik et al. (1999), who used generic health status measures, the Health Utilities Index (HUI) Mark 2 and Mark 3. Armstrong et al. (1999) note that a future directive for their development program is to examine the integrity of the factor structure of their cancer-specific measure in a “non-ill” population. This method of validation, referred to as the known-groups method, is less robust than other commonly available measurement techniques, such as construct or criterion validation. Yet the application of an instrument across both “healthy” and “ill” groups can better inform our understanding and interpretation of results and provide quantifiable evidence about the potential gains and losses in precision using disease/condition-specific instruments relative to generic ones. Similar work about relative precision has been conducted among children varying in severity of asthma (McGrath et al., 1999). Phipps et al. (1999) draw an important distinction concerning the recall period and application of their measure relative to those reported in the literature to date. Their tool, the BASES, is designed for use with hospitalized patients and utilizes a narrow recall window (8 to 12 hr, 24 hr). Unfortunately, the time frame for the instrument developed by Armstrong and the 2 measures employed by Crom et al. (1999) were not noted. The standard recall used in the design of most instruments is usually 2 or 4 weeks. This has allowed investigation of the chronic problems associated with long-term survivors. However, as Phipps et al. (1999) note, relatively few studies have examined the acute effects of treatment on the quality of life of children undergoing therapy. Future studies might be designed to capture both the short- and long-term effects on cancer survivors. In so doing, it may be possible to identify short-term effects in HRQL that predict future outcomes for children receiving therapy for cancer. Regarding age, Phipps et al. (1999) observed that, in their study, younger children reported less somatic distress than did adolescents. It would be important to uncover whether this finding holds across children, irrespective of cancer type, or whether age is a confounding variable. Finally, Armstrong et al. (1999) mention that focus groups were conducted across a broad age spectrum so as to derive an “item pool reflective of the full range of experiences”. Analyzing the properties of their instrument across the conventional age groups would strengthen this important design feature. Determining the most appropriate respondent to complete a quality-of-life instrument is complex. It should be guided by study-specific issues including protocol design, projected age range to be assessed, the reading and comprehension level of the reporters and the specific hypotheses to be tested. In the absence of definitive findings about the validity of child self-report for some ages and concepts, it has become an accepted practice to make inferences about children's ability to self-report by estimating the degree of association between the response from paired groups of parents and their children. In these analyses, it is often assumed (directly or indirectly) that parents are the “gold standard”. Since disease may impact each party differently, it stands to reason that differences in mean scores between parents and their children would be observed. As has been argued elsewhere (Landgraf and Abetz, 1996), high agreement among reporters, whether between parents or across parents and children, does not suggest that reports mirror one another, nor does low agreement suggest that one reporter is more “truthful” than the other. Findings to date suggest that, under the appropriate study conditions, children can provide useful and provocative information. Thus, investigations may be improved if the focus is shifted from “who is better” to assessing whether valuable information is lost by relying exclusively on one reporter vs. another. As noted by others (Landgraf and Abetz, 1996, 1998) and argued in the report of Parsons et al. (1999), understanding the issue of proxy reporting has less to do with “truth” and more to do with the potential influences of parents concerning access, direction and priorities related to children's medical care. Understanding the roles that fathers and mothers play, in this regard, is essential. Challinor et al. (1999) and Watson et al. (1999) highlight the potential influences of maternal well-being on proxy reports of children's HRQL. Challinor et al. (1999) report that parents of children with cancer rate themselves as more depressed than they rate their children. Watson et al.(1999) report cautiously that, based on their small sample, the mental health of the parent does not contribute to reporting bias. Taken together, these findings underscore the compelling need for further study concerning potential influences on reports of children's HRQL by proxy. Levi and Drotar (1999) and Sawyer et al. (1999) advance the field by extending beyond the issue of discrepancy among reporters to explore potential explanations for these differences. These investigators examined differences between paired parent and child responses at the item level for several of the scales that constitute the Child Health Questionnaire (CHQ). Although the CHQ is designed to be reported at the scale level (i.e., the sum of a given set of items), their objective was to better understand the overall pattern of discrepancies, where they occurred and if they could be explained by the presence/absence of disease in the child (cancer vs. healthy). Sawyer et al. (1999) provide further insight about adolescent self-report. While it does present information about parent–child correlations, it extends this analysis by examining the potential influence of independent demographic and treatment variables on the different domains used to define the HRQL concepts in the child-report version of the CHQ. Seid et al. (1999) focus exclusively on child self-report and, like others, present compelling empirical psychometric evidence that adolescents, at least 12 years of age, can provide information that is reliable, valid and defensible when compared to relatively objective criteria such as missed days from school. Using trained interviewers, Glaser et al. (1999) demonstrated that patients as young as 5 years can answer questions about their health status. Eiser et al. (1999) depart from the conventional models used to define the conceptual framework for most HRQL instruments and present compelling evidence concerning discrepancy between one's aspirations and experience as an alternative theoretical approach to instrument design. Their evidence further supports self-report by children. Finally, the European publications review conducted by Calaminus and Kiebert (1999) suggests that, in general, much of the ground-breaking work regarding HRQL in pediatrics represents “work in progress” and, as such, is not available through traditional searches like MEDLINE or ERIC. Many of the available HRQL measures developed specifically for children, such as the CHQ and HUI, are self-completed and easy to use and score. They have been translated and standardized rigorously to facilitate easy interpretation of results. As such, they are potential candidates for use across a variety of applications beyond the traditional clinical investigation. Their availability is an exciting development that has yielded opportunities for further study concerning relative strengths and weaknesses in performance characteristics. A principal example is provided by the work of Nixon Speechley et al. (1999), wherein both the CHQ and HUI were fielded in a study designed to assess the physical and psycho-social late effects in childhood cancer survivors. Information regarding the performance characteristics of the measures is based on preliminary data and, therefore, inconclusive. Nonetheless, the work lays an important foundation and sends a noteworthy message. Namely, it is more constructive and advantageous for the field to focus on the comparative strengths and weaknesses of individual instruments than to focus attention on a “futile quest for the perfect instrument”. In the same way, the work of Le Galès et al. (1999) highlights an important yet somewhat overlooked point. As the field matures and measures become more sophisticated in both presentation and interpretation, receptivity/acceptability by clinicians and patients, within the clinical care context, will become increasingly important. In the United States, where “HRQL” and “outcomes” are fast becoming standard terms in the medical marketplace, attention has shifted from measurement and research to the practical application and interpretation of HRQL instruments in busy clinic settings. While establishing reliability and validity are necessary prerequisites, the ultimate quest for health-care providers is access to practical, scientifically robust measures that provide information that has clinical relevance for both them and their patients. The message of usefulness and feasibility in real-world applications is articulated further by the work of Mulhern (1999). He argues that, although neuropsychological testing provides important data about the degree of neurocognitive deficits in children surviving medulloblastoma, a more practical and valid alternative to assess cognitive impairment is needed. His preliminary results suggest that the cognitive attribute of the HIU 2 is sensitive and may be an appropriate “proxy measure” for estimating IQ and achievement among children receiving treatment for medulloblastoma. In a similar vein, Kennedy and Leyland (1999) provide evidence that a generic HRQL instrument may be an appropriate and practical method for screening disability and psycho-social morbidity in children surviving brain tumors. Finally, the work by Szecket et al. (1999) reminds us that, while the field of measurement and its application in the pediatric oncology context is increasing in North America and much of Europe, comparable information for children surviving cancer in non-English-speaking countries is relatively unavailable. Whenever feasible, culturally and conceptually appropriate studies should utilize the availability of well-validated and rigorously translated measures developed specifically for pediatric populations. If cultural distinctions make adoption of available instruments problematic, at a minimum, developers should utilize standardized methods of design and analysis as described in the published reports. Whereas earlier papers in the conference focused on key measurement issues, work by Osoba (1999) and by Hudson et al. (1999) shifts attention from the background to the forefront by addressing issues related to interpretation of findings and the impact of education and intervention on the HRQL of adolescents surviving cancer. As mentioned previously, increased sophistication, in methods and measures to assess HRQL in pediatric populations, leads rather naturally to questions about the “usefulness” of these tools beyond the clinical trial context. The point is underscored in particular by the exposition of “quantification of meaningfulness” by Osoba (1999). For example, differences in HRQL scores across subjects or in an individual patient from time 1 to time 2 may indicate a statistically meaningful change in the clinical trial context, but the same change may not have the same relevance for a clinician and /or a patient. As with traditional clinical data, such as laboratory reports or sphygmomanometer readings, clinicians will gain confidence and develop their own internal benchmarks the more they are exposed to patient-based reports of HRQL within the clinical care context. Osoba (1999) describes an interesting empirically based method for studying the subjective significance of change from the patient's perspective. The work of Hudson et al. (1999) has broad implications for the field of HRQL in general and oncological care in particular. It highlights the important theoretical distinction between disease prevention and health promotion by exploring the potential usefulness of a promotion-based intervention for use with adolescents surviving cancer and its impact on their HRQL. The work of Haase et al. (1999) provides insight about the personal characteristics, such as resiliency and courageous coping, among adolescents surviving cancer and, like the report of Eiser et al. (1999), presents an alternative theoretical model for framing and defining the multi-dimensionality of HRQL. By studying related issues such as risk-taking assessment, adherence, barriers to behavioral change and family support, these reports further inform our general understanding of health perceptions among adolescents and their families. The International Workshop brought together a diverse panel of discussants and investigators representing the fields of measurement, clinical trials and academic research. As anticipated, the lively discussions raised more questions than answers. The communications collected here represent exciting work and demonstrate the potential for advancing our understanding of the HRQL of children surviving cancer. They also highlight the need for studies that focus on the implementation of scientifically robust instruments in settings outside the traditional context of clinical trials, the expansion of HRQL studies in countries outside North America and Europe, further validation of existing instruments and the continued exploration of alternative conceptual models for defining HRQL among children and adolescents. Finally, the work presented herein and mirrored in discussions throughout the workshop indicates that significant and timely advances in the field, in terms of both measurement and clinical practice, might be enhanced by increased and formalized collaborations across disciplines and professions.