Vol. 7, No.1 – Winter 2002
This issue of The Behavioral Measurement Letter contains a very large volume of material, so large, in fact, that it is a double issue. And although I was again assisted by our regular contributor, Fred Bryant, there was a related delay in its publication. I hope that both the quantity, and especially the quality of its contents will be adequate compensation for the wait.
In this issue of The BML, Dr. Frank Baker, Director of Research at the American Cancer Society, reviews instruments of various types used to measure the quality of life (QOL) of cancer survivors and research designs for evaluating cancer survivors’ QOL, and then discusses challenges and issues in QOL research. Due to the implementation of cancer screening programs and lifestyle changes (e.g., in diet, cigarette smoking, alcohol consumption), improved diagnostic methods, and the use of new and more effective cancer treatments in the last quarter of the 20th century, cancer, still associated with high levels of morbidity and mortality, is no longer the dreaded killer it once was. These advances have resulted in large and growing numbers of cancer survivors, many of whom live for years (even decades) beyond the time of initial diagnosis. Thus, instruments and methods are needed to assess the quality of life of cancer survivors in order to obtain baseline measures, to identify factors that contribute to a good QOL and those that contribute to a poor QOL for survivors, to design and improve ways and means for QOL enhancement, and to determine the effectiveness of attempts to improve cancer survivors’ QOL.
Racial, ethnic, and class bias are common, often substantial sources of error variance in measurement. Such bias may be introduced at any stage of instrument development and use, including definition and operationalization of variables to be measured, item construction, and instrument administration. These sources of measurement bias are discussed in a column by Drs. Mildred Ramirez, Marvella Ford, and Anita L. Stewart, from the Research Centers for Minority Aging Research — Measurement and Methods Cores. They point out that measures administered to various racial/ethnic groups and/or persons of low socio-economic status that do not account for racial/ethnic/class differences can produce results that are not generalizable to these groups. This, in turn, leads to flawed social policies and ineffective services designed using such research. The column strongly reminds us that 1) measurement bias of various types exists and must be addressed effectively to assure validity, and 2) the type of measurement bias due to insensitivity to racial, ethnic, and/or class differences has consequences not only for the corpus of research-based knowledge, but for applications of such invalid knowledge.
Also in this issue is the second of a two-part column, “Culture . . . One Step at a Time,” by cultural anthropologist John Gatewood. As the reader may recall, the first part (The BML (6) 2:5-10, Fall 1999) dealt with means to gain an understanding of a culture by learning how persons within the culture view their world. He presented three methods to discover how persons see similarities, differences, and relationships among products of a culture or items in its environment — free listing, pile sorting, and triadic comparison. In the second part of his column, Dr. Gatewood explores various ways to analyze data generated by these methods. Each is a type of consensus analysis and allows for analyses of either 1) similarities, differences, and relationships among items familiar to a culture, or 2) the level of knowledge each informant has about these items. (The reader is referred to part one for detailed discussion of the three methods and for references to software used to analyze data generated by these methods.)
Address comments and suggestions to The Editor, The Behavioral Measurement Letter, Behavioral Measurement Database Services, PO Box 110287, Pittsburgh, PA 15232-0787. If warranted and as space permits, your communication may appear as a letter to the editor. Whether published or not, your feedback will be attended to and appreciated.
We also accept short manuscripts for The BML. Submit, at any time, a brief article, opinion piece or book review on a BML-relevant topic to The Editor at the above address. Each submission will be given careful consideration for possible publication.
HaPI reading…
Al K. DeRoy, Editor
Assessing the Quality of Life of Cancer Survivors
Frank Baker
As a result of improvements in prevention, early detection, and treatment, more people are surviving cancer today than at any time in the past. It is estimated that there are at least 8.4 million people in the United States who are alive today after receiving a cancer diagnosis (American Cancer Society, 2000). The image of cancer has changed from a disease that was viewed as synonymous with death, to that of a chronic illness that can be survived but can inherently change various aspects of the life of the person who has the disease.
Even the definition of who is considered a cancer survivor has changed in recent years. The traditional definition of a cancer survivor was someone who was alive five years after diagnosis and was disease-free (Friedman, 1980). This definition of five-year survival was used because for many types of cancer, if a patient survives for five years, it is likely that he or she is cured. With the development of the Office of Survivorship at the National Cancer Institute, a new definition was offered that defined cancer survivors as people who were living two years postdiagnosis (Meadows et al., 1998). The American Cancer Society (ACS) has offered a definition of survivorship that is more inclusive, i.e., cancer survival begins with diagnosis and continues for the balance of the person’s life (H. Eyre, personal communication, April 24, 2000).
Cancer survivors have been described as going through a series of stages of survival (Mullan, 1985). First is an acute stage that begins at diagnosis and continues through the initial courses of medical treatment. At this stage, the survivor is commonly called a patient, and the primary focus is on physical survival. This period is followed by an extended, or intermediate, stage of survival in which survivors monitor their bodies for recurrence of disease. In this stage of recovery, quality of life becomes a major concern as survivors deal with the physical and emotional effects of treatment and their reentry into social roles. Most of the behavioral and psychosocial research with adult survivors has focused on the initial stage, when patients are in active treatment, and to some lesser extent on the second stage, when patients are likely to be seen at university-related hospitals and cancer centers where there is grant support for this kind of research.
The third major stage is called sustained remission, or long-term survival. Some long- term cancer survivors will do well. Others, however, must deal with chronic and late effects of treatment that is often toxic, that produces debilitating symptoms and impaired functioning, and that leads to short- and long- term changes in an individual’s ability to conduct a “normal” life. Thus, ever-improving technology and new treatments are saving cancer patients’ lives but at the same time have side effects and toxicities that can radically affect the quality of their lives.
Traditionally, medical evaluation of cancer treatment outcomes included tumor response, disease-free survival, and overall survival (US Department of Health and Human Services/ Public Health Service/National Institutes of Health, 1990). As length of survival increased, however, the value of cancer treatments began to be judged not only by which treatments contributed to increased survival, but also to the quality of survivors’ lives. This new criterion was called “quality of life” (QOL), a term that received its own heading in Index Medicus in 1977 (Frank-Stromborg, 1984). Quality-of-life research has increased dramatically in recent years, and there are numerous examples of activities on the national and international levels that indicate growing recognition of the importance of including QOL in assessments of patient outcomes. For example, the United States Food and Drug Administration uses QOL measurements in the process of approving drugs for the treatment of cancer (Johnson & Temple, 1985), and the National Cancer Institute includes assessments of QOL in endpoint evaluations of cancer clinical trials (Moinpour et al., 1989). The American Cancer Society (ACS) has set national challenge goals for 2015 that include improving the quality of life of cancer survivors in addition to decreasing cancer incidence and mortality (Beyers et al., 1999).
Assessing Cancer Patient’s Quality of Life
Early attempts to assess a cancer patient’s quality of life were based on ratings by doctors or other caregivers of the patient’s ability to function without special assistance. One of the first of these was the Karnofsky Performance Status Scale, which provides a rating of cancer patients’ physical functioning on a scale from 0% to 100% (moribund – normal) in increments of 10 (Karnofsky & Burchenal, 1949). Although this global clinician rating has been used widely by cooperative groups involved in clinical trials because it shows correlations with tumor response and survival, studies have shown that it has low interrater reliability and that such clinician-based ratings correlate poorly with patients’ ratings (Aaronson, 1990).
Over the last several decades, the view that the patient should be the primary source of information regarding his or her quality of life has gained general acceptance (Aaronson, Beckmann, Bernheim, & Zittoun, 1987; Celia & Tulsky, 1990). The trend toward using cancer patient self-report was furthered by studies that demonstrated that proxy ratings, whether by health care providers or family members, showed little agreement with patients’ ratings and with each other as well (Epstein, Hall, Tognetti, Son, & Conant, 1989; Osoba, 1994). When physicians are used as the source of quality-of-life ratings, they tend to emphasize physiological data and focus on symptoms and physical functioning, whereas nurses, social workers, and family members tend to emphasize psychosocial dimensions (Schipper, 1990). However, no matter who is used as a proxy for the patient, they tend to produce ratings that underestimate the QOL of patients as rated by the patients themselves (Sprangers & Aaronson, 1992). Indeed, it has been suggested that one of the most important positive effects of the widespread acceptance of the quality-of-life concept in medical practice has been the recognition that the perspective of the patient is as valid as that of the clinician (Leplege & Hunt, 1997).
While no universally accepted definition for QOL has developed over time, a general consensus has grown that quality of life is best defined as a complex multidimensional construct, not just in terms of a single dimension such as level of physical functioning (Donovan, Sanson-Fisher, & Redman, 1989; Grant et al., 1992). The World Health Organization defines quality of life as “individuals’ perceptions of their position in the context of the culture and value system in which they live and in relation to their goals, standards, and concerns” (World Health Organization-Division of Mental Health, 1993). Six broad domains are included in their definition: physical health, psychological state, levels of independence, social relationships, environmental features, and spiritual concerns. Researchers focusing on cancer survivors have developed some consensus about a minimum of four dimensions to include in quality-of-life measures: physical functioning, emotional and psychological functioning, social functioning, and disease/ treatment-related symptoms (Aaronson, 1990; Moinpour, Hayden, Thompson, Feigl, & Metch, 1990; Ganz, 1994). In addition, a global measure of perceived health status may enhance QOL measurement (Haberman & Bush, 1998).
Two approaches have been employed in attempts to define operationally the multiple dimensions of quality of life. One approach focuses on developing a single multi-dimensional scale, while a second approach is to develop many separate scales to measure QOL. The second approach may involve using a battery of single-domain scales to assess QOL, or a combination of a multidimensional measure with one or more scales that measure other domains. Instruments developed using the first approach often include generic mental health measures of such variables as depression, anxiety, mood state, self-esteem, perceived meaningfulness of life, and the like. In most cases, these measures have been normed with populations other than cancer patients, such as college students and outpatient psychiatric patients, and thus little or no information is available on how well they measure QOL of cancer patients. To overcome these limitations, normative data, especially data on validity and reliability, have been developed for some measures based on their administration to groups of cancer patients.
There have also been attempts to deal with these limitations by developing cancer-specific and diagnosis-specific QOL measures. While there is little theoretical basis for deciding what to include in a cancer-specific measure of QOL, most include measures of the key dimensions of physical well-being and associated symptoms, of psychological well-being, and measures of social well-being (Cella & Tulsky, 1993). Some researchers, too, argue for the inclusion of spirituality as a core dimension of QOL, pointing out that in interviews concerning quality of life, cancer survivors often emphasize the role of spirituality (Ferrell, Grant, Padilla, & Vemuri, 1991). In this regard, specific tools for measuring the spiritual well being of cancer patients are being developed, and evidence of a strong relationship of spirituality to QOL has been reported recently (Brady, Peterman, Fitchett, Mo, & Cella, 1999).
As interest in the assessment of quality-of-life treatment outcomes for cancer and other chronic diseases grew, some existing generic measures of health status outcomes were used initially. They were called measures of “health-related” quality of life (HRQOL), even though they had not originally been developed as QOL measures. These generic HRQOL measures include: the Sickness Impact Profile (SIP; Bergner, Bobbitt, Carter, & Gilson, 1981), which covers a wide range of functioning and focuses on behaviors rather than subjective expressions; the Medical Outcomes Study Short Form-36 (MOS SF-36; Stewart, Hays, & Ware, 1988), which consistently distinguishes among groups of patients differing in severity of illness; and the Nottingham Health Profile (NHP; Hunt, McKenna, McEwen, Williams, & Papp, 1981), which has most widely been used in the UK. It should be noted here that although the NHP is less precise in detecting distress associated with mild illness than the other generic HRQOL measures, it seems to assess adequately differences between groups with moderately severe to severe illness.
While generic measures are commonly used to assess outcomes in other chronic diseases, there has been significant focus on developing disease-specific measures for cancer outcomes. For example, quality-of-life research on the effects of cardiovascular disease has generally relied on generic measures, with some dimension-specific measures included as part of a “battery-of-measures approach” in which multiple distinct instruments are used, each measuring a specific QOL domain and each being scored separately (National Institutes of Health/National Heart Lung & Blood Institute, 1995). Additionally, since cancer is actually a number of diseases affecting different sites in the body, there has been a further refinement of cancer QOL measures to develop site-specific scales as well.
Measures of Quality of Life of Cancer Patients
Among the measures of quality of life developed specifically to use with cancer patients that have been shown to have adequate reliability and validity are the Cancer Rehabilitation Evaluation System (CARES; Schag, Heinrich, Aadland, & Ganz, 1990), the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC-QLQ-C30, Aaronson et al., 1993), the Functional Assessment of Cancer Therapy (FACT; Cella, et al., 1993), the Functional Living Index-Cancer (FLIC; Schipper, Clinch, McMurray, & Levitt, 1984), the Satisfaction with Life Domains Scale for Cancer (SLDS-C; Baker, Curbow, & Wingard, 1992), and the Spitzer Quality of Life Index (SQOL-Index; Spitzer et al., 1981). The original CARES included 139 problems that might be encountered by cancer patients on a daily basis as they attempted to cope with their disease and its treatment. A shorter form consisting of 59 items has been developed, the Cancer Rehabilitation Evaluation System-Short Form (CARES-SF; Schag, Ganz, & Heinrich, 1991). Unlike some of the other QOL measures for cancer, both forms of the CARES require permission for use and payment of fees, which may have limited their use. The European Organization for Research and Treatment of Cancer (EORTC), which involves 15 countries in one of the oldest and largest cancer clinical trials groups in Europe, developed the QLQ-C30 as a core instrument to which diagnosis-specific questionnaire modules could be added. The EORTC QLQ-C30 was developed specifically for assessment of the QOL of cancer patients participating in multinational clinical trials and has numerous parallel language versions. The FLIC and the Spitzer SQOL-Index currently appear not to be as popular as when they were first developed in the 1980s. The SLDS-C, consisting of only 17 items and employing a “smiley-frowney” faces response format to assess different life domains, has been shown to be a relatively less burdensome cancer QOL measure for patients to complete (Baker, Jodrey, Zabora, Douglas, & Fernandez-Kelly, 1996). The FACT-G, a general version of The Functional Assessment of Cancer Therapy scale that can be used with patients of any tumor type, has gone through four revisions (Cella, 1997). Disease-specific versions of the FACT have been developed for many of the major types of cancer (i.e., breast, bladder, colorectal, head and neck, lung, ovarian, and prostate), as well as for HIV infection, by adding items specifically relevant to different cancer sites to the core set of items in the FACT-G (Kornblith & Holland, 1994).
Several measures have been developed to evaluate the unique quality-of-life outcomes of bone marrow transplantation, a particularly demanding cancer treatment. Among these are the Quality of Life Scale-Bone Marrow Transplant, developed at the City of Hope by Ferrell and Grant (Ferrell et al., 1992; Grant et al., 1992), the Cella FACT-BMT (McQuellon et al., 1997), and the Bush Bone Marrow Transplant Symptom Inventory (Bush, Haberman, & Donaldson, 1995; Bush & Langer, 1998).
Research Designs for Evaluating Quality of Life
The most commonly used research designs in psychosocial and behavioral research on cancer have been cross-sectional in nature. Cross-sectional assessments are useful for describing the current effects that survivors are experiencing at a particular point in time in relation to other variables assessed at that time. However, this type of design has several noteworthy deficiencies. Retrospective reports are of only limited reliability regarding the respondents’ perceived quality of life at an earlier time, such as at the time of treatment or before cancer diagnosis. Retrospective assessments performed years after treatment also have a selection bias in that only those who survived to tha