The Need for and Characteristics of Randomized, Phase III Trials to Evaluate Symptom Management in Patients With Cancer

The paradigm of cancer research has evolved from goals restricted to tumor response and survival end points to include goals relating to optimal control of cancer-related symptoms. The initial declaration of the War on Cancer set the target goal at the total eradication of cancer. However, the gradual change in cancer being increasingly treated as a chronic rather than an acute condition has warranted a shift toward examining progress in the management of disease and treatment-related symptoms in phase III clinical trials. The challenge inherent in studying cancer-related symptoms is to sustain the integrity, interpretability, and reproducibility of the scientific process in trials involving novel end points, such as quality of life or a reduction in symptoms. Research must be held to standards of methodology and reporting that are as rigorous as those to which other treatment trials are held (1). Fortunately, guidelines for designing trials incorporating cancer-related symptoms and quality-of-life end points are beginning to appear in the literature (2,3). The first requirement of symptom management trials is that the measures be consistent with the concepts to be evaluated in the research question. The measures may be physiologic. Alternatively, they may be self-report measures, necessitated by the subjective nature of many symptoms. Whether the measures used are physiologic or self-reported, it is important that they assess the concept of interest, not merely a closely related surrogate. The self-report measures of symptom severity or impact of the symptom on quality of life can be equated to biologic markers of tumor progression (4). In this context, the amount of change that will be considered to be a clinically meaningful change or a minimally clinically significant change must be set a priori. This value must be justified and made on the basis of previous clinical observations or psychometric properties of the measure. Similar to the study of cancer treatment agents, the analysis and interpretation of the data are paramount in judging the validity of the study and whether it should be considered as a basis for change in good clinical practice. The impact of missing data on the analysis and interpretation of the findings is best assessed by way of a sensitivity analysis. Such an analysis typically incorporates many alternative methods of imputation to take into consideration the impact of various underlying assumptions on the findings. The attrition rate and the reasons for noncompletion of the assessments can identify sources of bias. It is important to present the power analysis that justifies the sample size in symptom assessment trials, just as it is in therapy trials. It is imperative that researchers present the complete data and put them in the proper context so that the readers have information on which to judge the merits of the study (5). Complete data presentations with supportive assumptions permit the reader to judge the quality of the findings as a basis for making changes in clinical practice. Complete data, which include the number of patients entered into the trial, the number of patients included in the analysis, the distribution of baseline values, changes in scores over time, and how missing data are handled, must be explicitly presented in the paper. Presenting results for individual patients via simple graphics such as scatter plots can be as illustrative as presenting profiles that use group averages. Simply put: “Reporting quality is associated with methodologic quality, but similar quality of reporting may hide important differences in methodologic quality and well-conducted trials may be reported badly. A clear distinction should therefore be made between reporting and methodologic quality of trials” (6). Furthermore, the results need to be presented in an objective and scientific tone and not as a marketing campaign for a particular treatment or pharmaceutical agent. There are encouraging signs that rigorous study of cancer-related symptoms can be undertaken successfully, including the article in this issue of the Journal by Vansteenkiste et al. (7), reporting the results of a randomized phase III trial of darbepoetin alfa on cancer-related symptoms in lung cancer patients receiving chemotherapy. There is a recognized need for better assessment and measurement of symptoms in research, which is illustrated by some of the recommendations of the State of the Science Conference on Symptom Management in Cancer: Pain, Fatigue, and Depression convened by the National Institutes of Health in July 2002 (8). The conference identified areas that have now been targeted for improvement in methods used in the clinical assessment of symptoms throughout the course of cancer. Some of the areas identified include: Lack of uniformity in measurement and methodology of symptom management, Lack of well-defined study populations, Failure to adequately describe study settings and study designs and a lack of standardization of study procedures, Lack of appropriate comparison groups, and Lack of knowledge about clinically useful cutoff scores and assessment of meaningful changes over the course of illness. Clinical trials of various approaches to the management of symptoms related to cancer or its treatments are needed to promote the well-being of persons with cancer, especially in the context of cancer as a chronic condition. These clinical trials must be as rigorous as those that evaluate cancer treatments.