Clinical trials, outcome measures, and response criteria.

Pillemer and Tilley: Editorial 407 After a patient has completed a clinical trial, the physician is anxious to know: “Did my patient get better?” Usually, the next question is: “Is the new treatment any good?” The emphasis of the first question tends to be whether the patient showed meaningful clinical improvement by the end of the study. For the second question, the emphasis is more toward whether the response in the group receiving the treatment of interest was superior to that in the control or placebo group. A positive answer for one question does not necessarily imply a “yes” answer for the other. For example, the patient’s physician may think: “I regret that the trial ended up with a negative result, but my patient got better.” The clinical trialist may think: “I regret that your patient didn’t get better, but the treatment that we tested was clearly superior to the placebo.” Here, we will look into how these questions have been approached, with an emphasis on clinical trials for rheumatoid arthritis (RA). We will also address how an alternative analytic approach may be used in RA as well as in conditions for which, in contrast to RA, widely agreed-upon response criteria may not yet be available. Over an extended period of time, investigators in many countries have sought the optimal outcome meas res for clinical trials in various rheumatic diseases. For RA, the efforts included the examination of a number of individual measures to find the best one1. These efforts have included exercises on how to count and score joints, and which joints to evaluate2,3. When considerable data became available after decades of work, an international effort was launched to determine which measures should be included in a core set to best describe the disease activity. After the Outcome Measures in Rheumatoid Arthritis Clinical Trials (OMERACT) meeting, the American College of Rheumatology (ACR) published a core set of criteria that were recommended for inclusion in clinical trials of RA4-6. This core set was subsequently ratified by the World Health Organization (WHO) and the International League of Associations for Rheumatology (ILAR) as the WHO/ILAR core set7. However, the question of how to optimally combine these outcome measures remained8,9. Paulus had developed rules-based criteria for response in RA trials based on data from the Cooperative Systemic Studies of Rheumatic Diseases (CSSRD) program (Table 1)10. Further activities by the ACR followed a heuristic or rules-based approach. The ACR analyzed data from CSSRD trials and applied rules that were essentially different ways of weighting and summing up the data across the core set of outcome measures11. An optimal rule would give the greatest difference between the active treatment and the placebo group, while giving a low placebo response rate. The goal was to achieve the greatest power to detect a treatment with the smallest number of patients. Different response criteria proposed by the European League Against Rheumatism (EULAR) and others were tested in analyses of the same CSSRD trials11. These analyses resulted in the selection of a rule that could be applied by clinicians to determine whether a patient was or was not a responder for a given therapy. The EULAR criteria use a more complex approach than the ACR, where each outcome measure is divided into categories based on percentage change from baseline12,13. As shown in Table 1, the EULAR criteria are based on both an improvement and the achievement of a low disease activity state, as measured by the Disease Activity Score (DAS). In the DAS, the weights for summing across different measures have been determined using a multivariate approach. This results in a more complicated equation that may have more precision for the set of clinical trial data on which it was originally based than, say, the ACR response criteria. There are problems with the heuristic approach. The first of these problems is inherent in extrapolating data from a single set of clinical trials to a later period. We may think of the conduct of clinical trials as the process of following cohorts of subjects who have been exposed or not exposed to the treatment of interest over a limited period of time. Editorial