| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Health Services Research and Development, VA Puget Sound Health Care System, Seattle, Washington; 2 Division of Pulmonary and Critical Care Medicine, University of Washington, Seattle, Washington; 3 Division of Pulmonary and Critical Care Medicine, Washington University, St. Louis, Missouri; and 4 Section of Pulmonary and Critical Care, Department of Medicine, and 5 Department of Health Studies, University of Chicago, Chicago, Illinois
Correspondence and requests for reprints should be addressed to Jerry A. Krishnan, M.D., Ph.D., Associate Professor of Medicine and Health Studies (Epidemiology), Director, Asthma Center, Section of Pulmonary and Critical Care Medicine, University of Chicago, 5841 South Maryland Avenue, MC 6076, Chicago, IL 60637. E-mail: jkrishna{at}bsdad.uchicago.edu
ABSTRACT
Understanding the strengths and weaknesses of the control group used in a randomized clinical trial is essential because comparisons of outcomes between the investigational and control groups form the basis of inferences regarding the safety and efficacy of the investigational treatment. For its 2007 International Conference, the American Thoracic Society (ATS) sponsored a scientific symposium in which the strengths and limitations of different types of control groups in clinical trials of pharmacotherapy, procedures, devices, and behavioral interventions were discussed. In this section, the co-chairs of this ATS symposium provide an overview of the presentations, including a brief historical perspective on the use of control groups in clinical trials.
Key Words: clinical trials • control groups • ethics
Inferences from findings of randomized controlled clinical trials (RCTs) provide the strongest level of evidence regarding the safety and efficacy of investigational treatments. The advantage of RCTs is largely derived from the increased likelihood that measured and unmeasured sources of confounding and selection bias are equally distributed between treatment groups under evaluation (1). RCTs are, by definition, comparative studies that evaluate differences in experience between two or more groups (1). Although the interpretation of RCTs usually focuses on the design of an investigational treatment group(s), an equally important issue is the design and outcomes of those randomized to the control group (2–4). Understanding the strengths and weaknesses of the control group used in an RCT is essential, because comparisons of outcomes between the investigational and control groups form the basis of inferences regarding the safety and efficacy of the investigational treatment(s) (Figure 1). Recent controversies have drawn attention to the experience of control participants in clinical trials (5). For its 2007 International Conference, the American Thoracic Society (ATS) sponsored a symposium on the selection of controls in clinical trials. In this issue of PATS, a number of the speakers from this symposium have provided an overview of methodologic considerations that need to be addressed, and at times balanced, when selecting control groups for clinical trials. In this introduction, we provide a brief historical perspective and a summary of the discussions presented at this ATS-sponsored symposium. Three of the presenters at the symposium (Drs. M. Castro, M.D., M.P.H., T. Thompson, M.D., and R. Sutherland, M.D., M.S.) have written individual manuscripts based on their presentations to provide greater detail and these are included as part of this issue of PATS.
|
James Lind, M.D. (1716–1794), a Scottish physician and naval surgeon, is credited with conducting the first documented controlled clinical trial in 1747 when he simultaneously tested several potential treatments for scurvy, a major cause of debility and death among sailors at the time. As discussed in a review by Thomas (6), Lind describes his clinical trial in a 1753 book, A Treatise of the Scurvy:
On the 20th May, 1747, I took twelve patients in the scurvy, on board the Salisbury at sea. Their cases were as similar as I could have them. They all in general had putrid gums, the spots and lassitude, with weakness of their knees. They lay together in one place, being a proper apartment for the sick in the forehold; and had one diet common to all. Two of these were ordered each a quart of cider a-day. Two others took twenty-five gutts of elixir of vitriol . ... Two others took two spoonfuls of vinegar . ... Two of the worst patients, with the tendons in the ham rigid, ... were put under a course of sea-water . ... Two others had each two oranges and one lemon given them every day . ... The two remaining patients ... took ... nutmeg . ... The consequence was, that the most sudden and visible good effects were perceived from the use of oranges and lemons; one of those who had taken them, being at the end of six days fit for duty . ... The other was the best recovered of any in his condition; and being now deemed pretty well, was appointed nurse to the rest of the sick.
Since this seminal publication, there have been significant improvements to Lind's original design, including the recognition of the need for randomization, masking, and prespecified endpoints. Nevertheless, Lind's study, using concurrent control groups to more clearly establish the superiority of citrus against all other remedies at the time, was a critical first step in the evolution of clinical trial study designs. A recent query using the search term "randomized clinical trial" using PubMed (7) revealed more than 280,000 citations, and a query of clinicaltrials.gov (8) using the same search term indicates there are at least 4,362 active studies recruiting patients. Lind would have been proud of his legacy!
SELECTION OF CONTROLS IN CLINICAL TRIALS: PRESENTATION SUMMARIESEthical Considerations in the Selection of Control Groups
Henry Silverman, M.D., M.A.
University of Maryland
Dr. Silverman provided a discussion of the ethical considerations in the selection of controls in clinical trials and discussed the distinction between the delivery of care provided in a medical setting compared with research. The goals and justification of risk in each situation are different and independent of one another. In the context of medical care, the goal is patient focused and the risks are assessed in relation to the benefit for the individual patient. In a research study, the goal is to advance medical science for society and the risks need to be balanced with the knowledge to be gained from the study. Because these represents two separate constructs, the use of placebos in lieu of "usual care" in clinical trials is not unethical so long as the use of placebos does not subject the participant to undue risk. Dr. Silverman also characterized potential ethical pitfalls when protocolized care in a control group deviates substantially from the care typically provided in clinical practice. In this situation, comparisons of outcomes between the control and investigational treatment group(s) could lead investigators to report differences more extreme than would occur when that same investigational treatment is later adopted into clinical practice; results of such trials, therefore, may not be readily generalizable and may not be ethically acceptable.
Setting Patient Expectations in Placebo-controlled Asthma Trials
Robert A. Wise, M.D.
Johns Hopkins University
Dr. Wise also reviewed the benefit and limitations of using placebo-controlled studies. About one-third of patients with a wide range of disorders improve after they are given a placebo (9). There are many reasons that patients entered into a clinical trial might improve after being administered a biologically inactive medication. These include the natural variation of disease severity combined with selection of the most symptomatic patients (i.e., regression to the mean), the self-limited nature of many disorders, adherence to other aspects of disease management (e.g., lifestyle or adherence with other treatments), reduction of anxiety through interaction with the research team, or the simple desire of patients to please their caregivers by reporting positive outcomes. Furthermore, response to placebo, particularly analgesics, can be manipulated by altering the expectation of the subject about the effectiveness of the proposed treatment. These mechanisms, however, may not require invocation of the "mind–body" physiological interaction that is often considered to be the mechanism of the "true placebo response." Ascertainment of the true placebo response requires comparison of a group treated with placebo with one that is selected, treated, and followed identically, but that does not take the inactive treatment. A recent analysis of trials that have made this comparison has suggested that most disorders, except for pain, do not show a true placebo response (10). Therefore, it is possible that, in many circumstances, our devotion to placebo-controlled trials may not be necessary to evaluate an active treatment effect. This is an important practical consideration for respiratory disease researchers, insofar as access to placebo formulations of inhaled medications may be limited.
"Placebo" versus "Best Available Therapy" as the Control Group in Clinical Trials of Pharmacologic Interventions: Which Is Better?
Mario Castro, M.D. M.P.H.
Washington University
Dr. Castro discussed the ethics and rationale of when it would be considered reasonable to use a a placebo control group and when it would be appropriate to use a "best available therapy" control group (see article by Castro in this issue). Dr. Castro advocates that using a placebo is ethically justifiable when it does not expose subjects to undue risk and when best-available-therapy alternatives are not available. Best-available-therapy control groups may be most appropriate when evaluating whether new interventions are superior to current practice.
"Usual Care" versus "Sham" as the Control in Clinical Trials of Devices: Which Is Better?
Rand Sutherland, M.D. M.S.
University of Colorado
Sham procedures for control subjects are used infrequently in clinical trials, but are of particular interest given the recent focus of invasive therapies for emphysema (e.g., endobronchial valves) and asthma (e.g., bronchial thermoplasty) (11, 12). Dr. Sutherland discussed the history and use of these devices in pulmonary medicine and outlined the need for such interventions because of the history of strong placebo effects associated with invasive procedures (see article by Sutherland in this issue) (13). These include a number of widely adopted procedures that were later found to be ineffective when compared with control arms that included a sham procedure (14, 15). In studies evaluating devices, sham procedures for the control group are desirable because they provide advantages of masking participants (and possibly researchers) to treatment allocation, including minimizing biases resulting from the measurement of some outcomes (e.g., patient-reported symptoms, outcomes requiring researcher judgment) and/or motivation of study participants and researchers to remain adherent to the study protocol. Of course, the use of a sham control group should consider the overall risk of the sham procedure as well.
"Usual Care" versus "Best Practice" as the Control Group in Clinical Trials of Nonpharmacologic Interventions: Which Is Better?
B. Taylor Thompson, M.D.
Harvard Medical School
Given the wide range of practices in use for most aspects of care and the paucity of research data supporting a single "best" approach for an individual patient, Dr. Thompson discussed difficulties in assigning a control group to "best practice" (see article by Thompson in this issue). Surveys of clinicians using clinical scenarios are recommended to determine the range of existing care practices during the development of a clinical trial. Several challenges, however, exist in designing a usual-care group in nonpharmaceutical interventions, including the possibility of secular changes in practice (before and during the course of the clinical trial) and the effect of different usual-care groups on our ability to detect a difference in outcomes at the end of the study (assuming a true difference exists; statistical power) and the extent to which study findings might apply to clinical settings (generalizability). In addition, the ethical and statistical implications of including multiple usual-care arms are discussed.
Control Groups in Behavioral Research
Cynthia Rand, Ph.D.
Johns Hopkins University
Behavioral intervention studies often test different approaches to improve patient or clinician behavior, such as increasing patients' and clinicians' adherence with recommended therapies. Control groups commonly used in behavioral studies include "no contact," "wait lists," or "attention" controls. In each of these situations, the selection of controls presents unique challenges. For example, attention controls may receive interventions such as educational pamphlets or alternative therapies that themselves change patient or clinician behaviors and outcomes. In addition, masking of treatment assignment when control groups are used in behavioral studies can be difficult. Furthermore, it is often difficult to develop an "inactive" behavioral control arm that is credible and equally preferable to patients. These considerations suggest the need for a compromise in which study conditions in the control group are sufficient to retain subjects while not being so involved that it significantly changes participant behavior.
Trials with Active Therapies as Controls: Superiority versus Equivalency Designs
R. Graham Barr, M.D., Dr.P.H.
Columbia University
Dr. Barr contrasted the principles of hypothesis testing in superiority and equivalency designs. In an equivalency design, the hypothesis is that the investigational treatment is neither superior nor inferior to the current gold standard. Thus, in equivalency designs, comparisons of interventions with a nonactive control arm (placebo, sham) would be meaningless, whereas it may be preferred for superiority designs, which are developed to test the hypothesis that an investigational treatment is superior to a control treatment. In addition to the rigor that is applied to clinical trials, such as masking and randomization to minimize the risk of imbalance in measured and unmeasured confounding factors between treatment groups, there are additional factors that must be addressed in a study designed to establish equivalency, including determining the equivalence margins (% difference from the gold-standard group that would be considered clinically equivalent). It is critical to use assays for endpoints with high sensitivity for the study outcomes and develop procedures to maximize adherence to the treatment protocol, because problems with either of these principles would lead one to erroneously conclude in favor of the null hypothesis of "no difference" between treatment groups.
SUMMARY
Inferences derived from RCTs provide the highest level of evidence from clinical research, and depend as much on the design of the control group as the intervention under investigation. As highlighted above, the control group may be assigned to a placebo, best-available therapy, usual- or protocol-driven care, or sham procedure. The governing principles in designing the appropriate control group for RCTs include minimizing the risk to participants, understanding the relationship between the control treatment and clinical practice, and assuring that the control group will provide the appropriate comparison to answer the primary study question.
FOOTNOTES
* All authors contributed equally to this article. 
The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.
Conflict of Interest Statement: D.H.A. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. M.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. J.A.K. has received $156,399 from Hill Rom, Inc., in research grants for an investigator-initiated clinical trial (2004–2008).
(Received in original form July 15, 2007; accepted in final form August 8, 2007)
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
