The term quality of life (QOL), or health-related quality of life, has emerged to organize and galva nize a collection of outcome evaluation activities over the past two decades in cancer treatment research. Prior to this, length of survival was considered to be the only primary outcome in oncology treatment research. Recently, however, progress in increasing survival has been slow, and at times has exacted considerable cost .
It is now widely accepted that in most circumstances quality of survival is as important as quantity of survival. This implies that a severely toxic treatment must be evaluated for its detrimental impact as well as its survival benefit. It also raises a less obvious point: that treatments can be considered efficacious if they improve the quality of life even in the absence of survival benefit. Thus, investigating the impact of cancer treatments on QOL is a two-tailed enterprise where treatment toxicity is traded not only with survival time but also with post-treatment function and well-being.
QOL evaluation entails a multidimensional quantification of patient functional status, usually as perceived by the patient. In the decades to come, treatment intensification strategies which increase toxicity are likely to continue, given the advent of hematopoietic growth factors and improved antiemetic regimens. This further increases the importance of evaluating toxicity, patient function, and patient preferences for treatment. QOL evaluation differs from classical toxicity ratings in two important ways:
1. It incorporates more aspects of function (eg, mood; affect; social well-being) than those which have typically been attributed to treatment.
2. It focuses on the patient's perspective.
The United States Food and Drug Administration has stated that benefit to quality of life (QOL) is one of two requirements for approval of new anticancer drugs . The other, of course, is improved survival. Given the incurability and increasing chronicity and prevalence of many forms of advanced cancer, the QOL endpoint has become very important. Industry has thus joined hands with the caring clinician in an unusual marriage, promoting supportive care and symptom relief in the name of quality of life.
Despite general acceptance of the value of assessing quality of life during cancer treatment, relatively few clinical trials actually include a QOL component. For example, fewer than 5% of clinical trials reviewed as of 1982 by the Department of Health and Human Services studied QOL . A 1986 survey of surgical trials revealed that only 3% had systematically evaluated QOL . In 1995, 15% of the currently active Eastern Cooperative Oncology Group (ECOG) trials include a QOL component. Although there are some prevailing attitudes which devalue the role of quality of life investigation in clinical trials, a larger obstacle to successful QOL research has to do with difficulty coordinating the social and medical sciences in a clinical setting.
Recently, however, developments in health-specific quality of life methodology have made accurate QOL evaluation a possibility. Dozens of measures, many of which are both practical and valid, have emerged over the past decade and are available for use. This paper discusses issues in the selection of patients and measures when studying quality of life during cancer treatment.
One of the purposes of this publication is to clarify the extent to which we can agree on a definition of quality of life as it applies to people with cancer. The closer we can come to agreement, the more likely we will be to prevent the use of inappropriate measures leading to inaccurate and confusing conclusions. Coming to agreement about a definition does not mean selecting one or a single set of measures; there is no "gold standard," and there cannot possibly be one until the construct as it applies to cancer is clarified. Even then, it would probably be unwise to name a gold standard; that would risk allowing the tail to wag the dog. As soon as a measure is accepted as complete, the investigator surrenders the opportunity to assess components not included in the scale even if they have major implications for QOL. For example, although most QOL scales measure common physical problems such as pain and nausea, most do not measure confusion. Confusion may be the linchpin of quality of life in a patient with a brain metastasis, or whose calcium level cannot be controlled. Should a QOL scale therefore measure confusion?
The same question could be asked of dozens of other clinical problems. If the decision about inclusion of an item into a quality of life scale was based upon the possibility that it could be important for any cancer patient, the gold standard scale would be very long indeed. Instead, the probability of occurrence and the relative importance of the item in the overall scheme must prevail. The modular approach, as described by the European Organization for Research and Treatment of Cancer (EORTC) QOL Working Group [5,6] and by Cella and colleagues [7-9], in which a core of general questions is supplemented with disease- and treatment-specific items, is a method for addressing this dilemma.
In their classic volume, Campbell and colleagues  describe quality of life as: "a vague and ethereal entity, something that many people talk about, but which nobody very clearly knows what to do about." While it may ring true, this description is a nightmare for the test developer. Some have suggested abandoning the term quality of life because it is too general to have meaning. Other less nihilistic observers have pointed out that because the current definition of the term is so vague, it has been exploited as a marketing tool . There is a consensus of opinion, at the very least, that QOL is multidimensional [5,11-14].
The integrity of the term quality of life has been justifiably challenged on the grounds that it cannot be validly measured because it means so many different things to so many different people. With respect to both content and construct validity, this is certainly true. Until one has a clear definition of the concept, including its component parts if applicable, one cannot determine whether a scale is validly measuring that construct. The first step toward successful assessment of QOL in the clinical research setting is to clarify its definition and component dimensions.
We had earlier developed a working definition of quality of life which laid the groundwork for measurement: "Quality of life refers to patients' appraisal of and satisfaction with their current level of functioning as compared to what they perceive to be possible or ideal."  This earlier definition was modified to explicitly incorporate the multidimensionality of QOL: "Health-related quality of life (QOL) refers to the extent to which one's usual or expected physical, emotional, and social well-being are affected by a medical condition or its treatment." 
As the initial definition implies, it is important to obtain an appraisal of the extent of dysfunction as well as a rating of how this appraisal matches expectations. The appraisal itself is important because it documents the patient's report of actual dysfunction. The expectation rating is important because it provides the patient's opinion as to whether that dysfunction is tolerable. Some patients with minimal actual disability are extremely dissatisfied, while others seem quite able to tolerate severe impairment and may even feel fortunate to be obtaining therapy. Many decisions about treatment are best made with this knowledge.
Patients' perceptions of their illness are extremely variable, and factors other than actual disability enter into that perception. For example, bedridden status may be more upsetting to an adolescent receiving bone marrow transplantation than to an older adult with a history of chronic arthritis. For the adolescent, bedridden status represents a 100% decrease in normal activity level. For the older adult who could never expect to be fully ambulatory because of preexisting arthritis, the bedridden status represents less than a complete loss of possible ability. To assume that the same actual activity level in these two individuals would reflect comparable quality of life would be an obvious error.
A more subtle example is the presence of sexual dysfunction in a couple with an active and unconflicted sexual history, compared to the same dysfunction in a couple with a premorbid history of marital conflict and sexual difficulties. To the former couple, the same level of actual dysfunction would likely be more disruptive because it deviates more dramatically from their history. For the couple with premorbid sexual dysfunction, it is unwise to assume their difficulty can be attributed to cancer treatment.