diagnostic tests, screening procedures, devices, and other forms of medical advice or patient care.

The definition of clinical trial used in this report encompasses some studies involving only a single intervention group. Such studies are included because they are often carried out before the definitive, comparative study to gather specific pieces of information about the intervention before the comparative study can proceed. To a great extent, these "single-arm," early phase clinical trials have been defined by the regulations governing approval of new drugs and are also a prominent feature of trials in cancer treatment (see Table 1-1). Their value lies in setting the stage for definitive, randomized trials. The committee believes that at least some patient care costs incurred in these trials should be eligible for Medicare reimbursement.

TABLE 1-1. Phases of Clinical Trials (usually applied to drugs and devices)

Phase 1

First studies in people, to evaluate chemical action, appropriate dosage, and safety. Usually enrolls small numbers of participants and typically has no comparison group.

Phase 2

Provides preliminary information about how well the new drug works and generates more information about safety and benefit. Usually includes comparison group; patients may be assigned to groups by randomization.

Phase 3

Compares intervention with the current standard or placebo to assess dosage effects, effectiveness, and safety. Almost always uses random allocation to assign treatment. Typically involves many people (hundreds or thousands) but may be smaller.

Phase 4

"Post-marketing surveillance," evaluates long-term safety (and sometimes effectiveness) for a given indication, usually after approval for marketing has been granted by FDA.

Brief History of Clinical Trials

The formal evolution of the clinical trial dates from the eighteenth century, but the concept of comparing how well people fare after being "assigned" to different "interventions" (e.g., diets or medical treatments) has ancient historical origins. A considerable body of literature traces major developments along the way. (e.g., Bull, 1959; Lilienfeld, 1982; Meinert, 1986).

The practice of randomization—randomly assigning study participants to either an experimental or a control group—was introduced by the statistician Ronald Fisher in horticultural research in 1926 and was described in his 1935 book (Fisher, 1926, 1935). Fisher asked the elemental question: How does one determine whether an observed difference in yield between fields is due to the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement