|
|
||||||||
|
J Am Dent Assoc, Vol 135, No 4, 464-473.
© 2004 American Dental Association | ![]() |
PRACTICAL SCIENCE |
| ABSTRACT |
|---|
|
|
|---|
Methods. The authors searched MEDLINE and the Cochrane databases of systematic reviews and abstracts of reviews of effectiveness, as well as identified reviews that were known to the authors but not found in the searches. Systematic reviews included in this survey stated the intention to identify all relevant articles within predefined limitations, applied defined exclusion and inclusion criteria, and presented complete raw or synthesized data from included studies.
Results. This literature survey identified 131 systematic reviews, 96 of which had direct clinical relevance. During the past 14 years, clinically relevant systematic reviews have been published with increasing frequency. These reviews vary in the types of studies included and the assessment of those studies. The results of the reviews also varied in their definitiveness, with 17 percent finding the evidence to be insufficient to answer the key question. An additional 50 percent of the 96 reviews hedged in answering the key question, by noting that the supporting evidence was weak in quality or limited in quantity.
Conclusion. The number of systematic reviews that address clinical topics in dentistry is small but growing. However, the authors of more than one-half of these reviews believed that the evidence available to answer the key question was not strong.
Clinical Implications. As systematic reviews continue to grow, dentistry will become better informed about the adequacy and congruence of the scientific evidence underpinning clinical practice.
Evidence-based health care depends on application of the best knowledge a discipline can offer about the clinical course of a disease or condition and the effectiveness of alternative treatments. Like all health care professionals, dentists face the challenging task of keeping up with a constantly expanding knowledge base. Further, as front-line providers of oral health care, dentists are pressured regularly by manufacturers to use new materials and products. Dentists need to make everyday clinical decisions while attempting to sort the wheat from the chaff amid a virtually overwhelming volume of information.
During the 1990s, expert panels of clinical research methodologists and other specialists developed protocols to reduce bias in collecting and synthesizing findings from clinical scientific studies. The product of this process of reviewing the knowledge base related to a specific clinical question is known as a systematic review of the evidence. Today, systematic reviews are considered the preferred method for identifying all of the available knowledge, determining which information is "best" and summarizing it in a clinically useful manner.1
Extracting the current best evidence2 from the research literature is by no means easy, given the sheer quantity of information and the exponential growth in the scientific knowledge base. According to one estimate, more than 34,000 articles appear each month in the 4,000 journals indexed by MEDLINE, which is a mere fraction of the 100,000 scientific journals currently being published.3 In addition, not all scientific studies are equal, with some more likely to be free of bias than others.
Systematic reviews are designed not only to identify all relevant information contained in the literature, but also to evaluate the quality of the information, and then, if possible, to summarize the results from the strongest (or least biased) studies. In addition, a systematic review can help establish whether findings from individual studies are consistent and generalizable across patient types. Finally, under some circumstances, results of individual studies can be combined to increase the precision of effect estimates.
The shift toward systematic reviews has been a relatively recent phenomenon, with reviews of this type appearing with ever-increasing frequency during the past decade. The Cochrane Librarys Database of Abstracts of Reviews of Effectiveness, or DARE, which summarizes a subset of all systematic reviews that meet certain minimum criteria, lists only three reviews published in 1993, but increases exponentially to 484 reviews in 1999.4 The increasing prevalence of such reviews requires that clinicians understand the conceptual basis and structure of high-quality systematic reviews.
As with all scientific literature, the quality of evidence-based systematic reviews varies widely. Nevertheless, as the amount of information continues to increase in virtually every biomedical discipline, high-quality systematic reviews will be used further to summarize and synthesize the existing evidence. In this article, we describe the rationale for and methods used in systematic reviews, and survey the research literature to identify and characterize available systematic reviews of topics relevant to clinical dentistry.
Formulating the key question.
The steps of an evidence-based systematic review address these potential sources of bias (Box 1).5 The review is conducted much as is any scientific study, beginning with the development of a protocol that outlines all steps in the review process. The initial step of a systematic review is formulation of one or more related key questions, which usually are concisely stated and identify four crucial PICO elements (acronym derived from the first letters of the terms below):
As an example, a PICO question might be stated as follows: "What is the effectiveness of semiannual fluoride varnish in preventing dental caries in permanent teeth among children aged 6 through 12 years compared with semiannual fluoride gel treatment?" In a question of this type, not only is the questions clinical relevance immediately apparent, but its phrasing provides definitive guidance for subsequent steps in the review process.
Defining inclusion and exclusion criteria.
The second step of a systematic review is defining inclusion and exclusion criteria for the research literature. These predefined criteria commonly stem from the key question and other considerations, such as study designs, publication dates and languages, and details of treatment procedures (for example, sponge versus custom trays) and outcomes (for example, occlusal-surface caries versus smooth-surface caries). General classes of outcomes (for example, disease measures, pain, tooth retention, patient satisfaction, esthetics, tooth whiteness) usually are specified in the key question. The reviewers define the criteria in the systematic review protocol before identifying potentially eligible studies to minimize the influence of bias when making decisions to include or exclude studies.
Designing a search strategy.
The third step of the systematic review protocol is designing a search strategy. Because systematic reviews attempt to identify all studies relevant to the key question, the search for such studies characteristically is intense and commonly includes searching electronic indexes, such as MEDLINE, EMBASE, the Cochrane Library and, if appropriate given the topic, more specialized indexes. The reviewers develop and test keyword combinations to identify as many relevant studies as possible (sensitivity), while simultaneously including the smallest number of irrelevant studies (specificity).
Reviewers often manually search relevant current journals and back issues of relevant journals that are not electronically indexed, as well as review reference lists of all potentially eligible studies identified in the initial stages of the search. In addition, they often examine "gray literature" as well, including dissertations and theses, conference reports, abstracts and unpublished studies, many of which are identified through formal requests to, and informal networking among, professional colleagues.
Determining study eligibility.
The fourth step involves the application of the inclusion and exclusion criteria to determine eligibility for each study identified in the search. The review protocol usually specifies that two reviewers should develop independent lists of included studies and describes a procedure for resolving disagreements. This selection process is conducted in at least two stages. In the first stage, studies identified in the search that are clearly ineligible based on title or abstract are excluded. In the second stage, the reviewers apply the criteria to each remaining full report, and maintain a written record of reasons for exclusion.
Extracting information.
The fifth step of the systematic review protocol is extracting information in a standardized manner from each included study. The information extracted includes specific characteristics of the study design, subjects, methods and results, along with the information needed to assess the quality of the study. Also, as with the eligibility determination completed in the preceding step, the data extraction process usually is performed by two independent reviewers.
Analyzing and presenting results.
The sixth step includes analysis and presentation of results of the systematic review. All extracted data are presented in one or more evidence tables, which are prepared to facilitate comparison of the included studies. A qualitative summary of these studies, based directly on the evidence tables, usually is presented that describes the principal characteristics and findings of the studies. The summary also identifies the excluded studies and provides reasons for exclusion. In some instances, a qualitative narrative summary is the only analysis presented. However, in most instances, the reviewers evaluate the study results for heterogeneity or between-study differences.
Depending on the extent of heterogeneity present, study designs and data available in the published studies, the systematic review team also may conduct a meta-analysis of the outcome data. The American Dental Association defines a meta-analysis as a review that uses quantitative methods to combine the statistical measures from two or more studies and generates a weighted average of the effect of an intervention, degree of association between a risk factor and a disease, or accuracy of a diagnostic test.6
Interpreting the evidence.
The final step in the systematic reviewinterpreting the evidenceis the only step not guided by the review protocol, and the only one for which some subjectivity is permissible. Here, the reviews limitations and strength of the evidence are discussed, and the applicability of the study results to the clinicians work is considered. Equally important, the systematic reviewers may identify implications for future research.
In a report of a systematic review, the presentation of research findings depends primarily on the publication format. A free-standing report, generally published by sponsors of substantial reviews, is the most comprehensive medium. Such reports typically contain information about the methods included in the study protocol, complete reference lists of all included and excluded studies, evidence tables of all extracted data, detailed search strategies, a clear statement of what the evidence indicates with respect to the key question, and tables summarizing all analyses.
Depending on the proclivities of the study sponsor and authors, the interpretation may stress clinical or research implications. By necessity, reports of systematic reviews appearing in scientific journals are more abbreviated, but they still should indicate the key question, search strategy, inclusion and exclusion criteria, complete citation of each study and characteristics of included studies in a condensed evidence table, a statement of review results (including results of any meta-analyses) and clinical or research implications.
The structure of a systematic review facilitates, but does not guarantee, an objective summary of the evidence for a clinical question. Departing from accepted standards for conducting a systematic review will increase the likelihood that the results will be biased. The reader then must determine if the increased likelihood of bias is sufficient to render the review not useful. Existing instruments and guidelines purporting to assess the quality of systematic reviews generally focus on completion of the steps outlined in Box 1.7,8 One recent appraisal of systematic reviews of randomized controlled trials in medicine found generally low levels of adherence to these quality levels, which were attributed to both poor practice and incomplete reporting of methods.9
Whether the question addressed by the systematic review can be definitively answered by the review is not a measure of its overall quality. Surprisingly, the results of systematic reviews often are equivocal, because either the necessary studies have not been conducted or the reviewers judge the quality of the studies to be inadequate to address the clinical question without bias. Thus, from the standpoint of clinical applications, a primary advantage of the systematic review also is one of its greatest frustrations: it not only tells us what we do know, but also what we do not.
Our MEDLINE search strategy included the terms "systematic review" or "meta-analysis" paired with the terms "dentistry" or "dental" or "tooth" or "teeth" or "periodontal disease" or "stomatognathic diseases not pharyngeal diseases." The term "meta-analysis" was an important part of the search strategy because, in principle, a meta-analysis should be preceded by a systematic review to ensure an unbiased estimate.10
We identified 555 articles from the MEDLINE and PubMed searches, and added three DARE reviews, 23 Cochrane reviews and 14 other reviews, for a total of 595 reports. Using a subjective set of inclusion criteria, we identified what we considered to be reports of systematic reviews. Our inclusion criteria required the report to do the following:
We excluded studies not published in English and included all reports with any direct relevance to dental research or practice.
After we applied these criteria, the number of reports originally identified in the MEDLINE and PubMed searches dropped to 91, but we did not exclude any other identified reports. Thus, with the 40 additional reports (DARE, Cochrane and others) listed above, the total number of reports included was 131. We applied a further inclusion criterion to limit the identified reviews to topics with direct clinical relevance (that is, the results of the review would be useful to dentists or their patients in making decisions about clinical procedures).
Reviews addressing survival, treatment effectiveness and other treatment outcomes and side effects also were eligible, as were studies examining prognosis (that is, risk factors for disease and factors affecting treatment outcomes, if these factors were open to intervention). We excluded systematic reviews of epidemiologic studies, community-level interventions, emerging treatments or diagnostic methods and benchmarking reviews. Application of this criterion resulted in identifying 96 clinically relevant systematic reviews. We emphasize that the inclusion criteria we used were subjective, and thus may not be exactly duplicable.
Publication dates.
To characterize the status of systematic reviews in dentistry, we analyzed the set of identified reviews in four ways. First, we examined the publication dates to determine if these reviews mirrored the pattern reported in the medical literature of increasing frequency of systematic reviews over time. As the figureAs systematic reviews continue to grow, dentistry will become better informed about the scientific evidence underpinning clinical practice.
![]()
RATIONALE AND STRUCTURE OF SYSTEMATIC REVIEWS
TOP
ABSTRACT
RATIONALE AND STRUCTURE OF...
SURVEY OF SYSTEMATIC REVIEWS...
EVIDENCE FOR EVIDENCE-BASED...
CONCLUSION
REFERENCES
A thorough review of the literature generally is considered the key step in assessing the evidence that must underlie any evidence-based care decision.1 However, the traditional or narrative literature review may be anything but thorough, and can be plagued by bias and errors. Just as the design of the randomized clinical trial minimizes bias and threats to the studys internal validity, a systematic review is designed to minimize bias when summarizing the evidence for a particular clinical issue. Bias, however, can be introduced into the review process in numerous ways, from incomplete identification of existing studies to subjective decisions to include or exclude studies, or from failure to objectively appraise the strength of the included studies to subjective synthesis of the results of those studies.
Inclusion and exclusion criteria commonly stem from the key question and other considerations.
The reviews limitations and strength of the evidence are discussed in the final step.
![]()
SURVEY OF SYSTEMATIC REVIEWS IN DENTISTRY
TOP
ABSTRACT
RATIONALE AND STRUCTURE OF...
SURVEY OF SYSTEMATIC REVIEWS...
EVIDENCE FOR EVIDENCE-BASED...
CONCLUSION
REFERENCES
We surveyed the English language literature to characterize the available systematic reviews of topics relevant to dentistry. Because this search was not exhaustive and relied on a single reviewer (J.B.), we do not consider it to be a systematic review. This search focused only on MEDLINE from 1966 to Dec. 31, 2002, and our results were supplemented with PubMed entries awaiting MEDLINE indexing, unduplicated entries in DARE, reviews in the Cochrane Library and additional reviews known to us personally.
indicates, there was a pattern of increasing numbers of systematic reviews published over the years. The small spike in the number of reviews appearing in 2001 is explained in part by the appearance of the proceedings of the 2001 National Institutes of Health Consensus Development Conference on the Diagnosis and Management of Dental Caries Throughout Life. This was the first NIH dental consensus conference to base its recommendations on a set of systematic reviews. This approach has become part of the standard background preparation for such proceedings.
|
More than half (54 percent) of the systematic reviews assessed the quality of the included studies. Such an assessment is a desirable feature of all systematic reviews, because it speaks to the strength of the evidence available to answer the clinical question being addressed. Typically, the assessment is made either through use of one of several existing quality scales, or by the explicit determination of the presence of key features of a study that safeguard against bias.8 Thirty percent of the reviews included only studies reported in English. The remainder either explicitly stated that studies in languages other than English were included (22 percent), or did not specify any language criterion (48 percent).
Review topics. Our third analysis grouped the systematic reviews by the general topics addressed. Box 2 (page 468 and 469) shows this distribution and indicates the specific topic for each review.11106 We provided citations to facilitate reader access. The breadth of topics addressed by single reviews ranges from the narrow (for example, treatment of keratocysts) to the broad (for example, treatment of periodontal disease). In some instances, two or more reviews address the same topic. The distribution of topics across clinical disciplines appears to be reasonably balanced, with dental caries the most frequently addressed topic and endodontic treatment the least frequently addressed topic.
Review conclusions. In our fourth analysis, we examined the conclusions of these reviews to determine if the key question had been answered. Some reviews stated no question explicitly, and for those reviews we assessed whether the stated purpose of the review had been fulfilled. Eighty (83 percent) of the 96 reviews presented results that answered the key question or otherwise satisfied the reviews purpose. In the 16 instances in which the question was not answered, the reviewers found the available literature to be inadequate, equivocal or both.
When the key question was answered or the stated purpose fulfilled, we examined the conclusions and the abstract to determine whether the report hedged the findings by noting that the literature supporting the finding was low in quality or limited in quantity. We did not consider hedging to have occurred when the review acknowledged weaknesses in the discussion section. We only considered a finding to have been hedged when the authors noted these concerns in conjunction with the results (that is, in the abstract, conclusion or both). We determined that of the 80 reviews that answered the key question, 48 (60 percent) hedged their answers. Thus, 64 (67 percent) of the 96 systematic reviews with clinical relevance to dentistry either could not answer the key question or answered the question with reservations about the strength of the evidence.
| EVIDENCE FOR EVIDENCE-BASED DENTISTRY |
|---|
|
|
|---|
To date, 96 systematic reviews address a smaller number of questions in clinical dentistry. These questions address only a small proportion of the clinical decisions that practicing dentists must make daily as they care for patients. Also, some of these reviews found inadequate evidence in the literature upon which to base a definitive finding, and a larger proportion tempered the definitiveness of the findings.
These problems may explain in part why there are relatively few systematic reviews of topics relevant to dentistrynamely, that the number of studies addressing a given clinical question, the design of these studies and their power often are inadequate. Dentistry has tended to be an empirical science,107 and only recently has begun to examine treatment decisions long assumed to be cut and dried. The growing number of systematic reviews is part of this new attention to effectiveness in dentistry; however, before the number of useful reviews can be greatly expanded, the evidence itself (that is, the pool of individual studies) also must be enlarged.
Another problem, however, could hinder the improvement of patient care even with the increased attention given to effectiveness of care. As the citations in the table indicate, most of the reviews appeared in journals with limited circulation to general dentists. Thus, the findings of systematic reviews are relatively inaccessible, thus impeding their dissemination to the profession. Fortunately, two mechanisms for information transfer are available to facilitate the distribution of findings.
First, two new journals have appeared in paper form and on the Internet: Evidence-Based Dentistry108 and The Journal of Evidence-Based Dental Practice.109 These publications present summaries of new clinical trials and systematic reviews with implications for dental practice. Second, practice guidelines in dentistry are slowly increasing in number and improving in quality,110 and the availability of these guidelines on the Internet has improved markedly through the creation of the National Guideline Clearinghouse.111 The ADA also has approved a policy statement and a program to introduce EBD to the dental profession, and The Journal of the American Dental Association may be another major source of EBD reviews and clinical recommendations.
| CONCLUSION |
|---|
|
|
|---|
|
|
| FOOTNOTES |
|---|
| REFERENCES |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
A.-M. Bollen Effects of Malocclusions and Orthodontics on Periodontal Health: Evidence from a Systematic Review J Dent Educ., August 1, 2008; 72(8): 912 - 918. [Abstract] [Full Text] [PDF] |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |