Registration puts basic information about a clinical trial in the public domain before it begins with the aim of providing reliable information about research in progress to the public, health care providers, researchers and funding bodies.
Dentistry has entered a new era of evidence-based practice, and society is demanding that clinicians deliver prevention and treatment that has been proven to be effective. Randomized clinical trials (RCTs) are the "gold standard" of evidence-based practice, and their results must be reported promptly and clearly so they can be rapidly translated into clinical practice and provide evidence for establishing public health policy. For this reason, I recommend that The Journal of the American Dental Association and other high-impact oral health journals adopt conditions for publication that are consistent with current standards of trial reporting.
 |
THE CONSORT STATEMENT
|
|---|
The Consolidated Standards of Reporting Clinical Trials (CONSORT) Statement was developed to improve the quality of clinical trial reporting.1 It has been adopted by high-impact medical journals2 and uses a 22-item checklist and a flow diagram to ensure transparency in reporting clinical trials. This information is essential to judge the reliability or relevance of the findings, and empirical evidence indicates that not reporting this information is associated with biased estimates of treatment effect.1 Systematic reviews have shown that the reporting of clinical trials in dentistry is poor and that the quality of reporting of trials improves in journals that have adopted CONSORT guidance.3 Accordingly, I recommend that all publications of clinical trials in JADA and other high-impact oral health journals conform to the CONSORT statement so that reporting of clinical trials can be improved and standardized.
|
MAIN PUBLICATION AND INTENTION-TO-TREAT AND SUBGROUP ANALYSES
|
|---|
Primary outcomes of an RCT should be reported in one article that focuses on the design of the trial and its primary outcomes. Secondary outcomes may be summarized in this article, but details should be reported in subsequent publications. In this regard, because clinical trials are not designed or powered to investigate these outcomes, secondary outcomes must be viewed with caution.
Missing data in a clinical trial are a serious issue because they may introduce unknown bias. Several methods of data imputation have been proposed to deal with missing data in the statistical analysis of trials. A common method is to use an intention-to-treat (ITT) analysis, in which data from all participants are included in the group to which they were originally assigned, regardless of whether they adhered to the protocol or completed the intervention.4 While often counterintuitive to clinicians, ITT analysis prevents bias caused by nonadherence to protocol or nonrandom loss of participants, both of which may disrupt baseline equivalence established by random assignment.
On the other hand, per-protocol analysis selectively analyzes data from participants who complied with the protocol. It may bias trial results because participants may drop out of a trial or not comply with the protocol on a nonrandom basis. For example, in a clinical trial comparing a new (test) mouthrinse with a placebo rinse, some subjects randomized to the test rinse may find its taste unpleasant and not use it according to the protocol, or some subjects in the placebo group may discontinue its use because they feel it is not helping them. In a per-protocol analysis, the results might be positively or negatively biased because only data from test subjects who found the taste acceptable or data from placebo subjects who felt that their rinse was effective would be analyzed. It is impossible to adjust the analysis for such bias because its direction and magnitude are unknown.
A per-protocol analysis is most appropriate for Phase I and Phase II studies, in which one is attempting to establish initial safety and "proof of principle" for an intervention. It is not appropriate for Phase III trials, in which the purpose is to obtain evidence of safety and efficacy or effectiveness under conditions that approximate real-world conditions. Indeed, the burden of proof is on those who claim a need to violate ITT; it has been argued that this burden should be extremely high, and plans for a per-protocol analysis should be stated in a public registry before randomization begins.5
It also must be emphasized that subgroup analyses, whether per-protocol or not, often result in small sample sizes that are unbalanced among treatment arms of the trial. For this reason, post-hoc subgroup analyses of clinical trials may be best described as "data dredging" and should be avoided. Unfortunately, such analyses are far too common, especially in clinical trials that show no effect of the main outcome.
|
TRIAL REGISTRATION
|
|---|
Registration puts basic information about a clinical trial in the public domain before it begins with the aim of providing reliable information about research in progress to the public, health care providers, researchers and funding bodies. It increases the likelihood that results eventually will reach the public domain. Unregistered and unreported trials cannot contribute to the evidence base for health care decisions.6 People who participate in clinical trials typically provide consent in the belief that they are contributing to knowledge—but if the knowledge gained is not reported, trust between participants, investigators and research ethics review boards is damaged.7
As summarized by Tonks,6 trial registration has many advantages. It helps prevent publication bias that results from underreporting of trials with disappointing, negative or inconclusive results. This bias misleads those who conduct systematic reviews and clinicians who rely on published evidence for making treatment decisions. Registration also helps prevent unnecessary duplication of research effort and encourages replication and confirmation of results. It alerts researchers to gaps in knowledge and provides reliable information about ongoing trials that helps funding agencies direct support to where it is needed most and provides investigators with an opportunity to inform the community about the basic design and analytic procedures of the trial. It also provides improved access to industry research and satisfies public demand for unbiased evidence about the effectiveness of treatments.
The International Committee of Medical Journal Editors has adopted a policy to encourage public trial registration as a condition of publication.8 There are three types of trial registries:
- – government registries such as that maintained by U.S. National Institutes of Health ("www.clinicaltrials.gov");
- – registries developed by publishing companies such as Current Controlled Trials ("www.controlled-trials.com");
- – registries maintained by pharmaceutical companies such as GlaxoSmithKline ("www.gsk-clinicalstudyregister.com/").
The World Health Organization also maintains an International Clinical Trials Registry Platform to provide information about clinical trials ("www.who.int/ictrp").
|
SUMMARY OF RECOMMENDATIONS FOR CONSIDERATION
|
|---|
As a condition for publication of clinical trials, I recommend that JADA and other high-impact oral health journals do the following:
- – require adherence to the CONSORT Statement recommendations;
- – unless specified otherwise in a public trial registry before enrolling subjects, require an ITT analysis or some other acceptable means of data imputation for missing data in RCTs;
- – require registration of Phase III RCTs on a public Web site before randomization.
Adoption of these conditions for publication will provide clear and transparent guidance to authors and reviewers and will give policymakers and clinicians the unambiguous information they need to make sound decisions based on evidence from RCTs.
TOP
THE CONSORT STATEMENT
MAIN PUBLICATION AND INTENTION...
