| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Walaeus Library, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, the Netherlands
e-mail: j.w.schoones{at}lumc.nl
Editor:
In the March 2006 issue of Radiology, Dr Boks and colleagues (1) provide explicit details of the literature search they performed. They should be commended for doing so. The availability of literature search details improves the scientific value of published research because it offers a possibility to reproduce and judge results accordingly. However, the specific search performed by Dr Boks and colleagues appears not to be optimal. I would like to address three issues by which the search could be improved.
First, why was the search not performed in additional databases such as EMBASE or the Cochrane Library? If you do not search EMBASE, you will miss the content of around 2000 medical journals (2,3).
Second, it is not clear which version of MEDLINE was used for the search. There exist several differences between PubMed and other versions of MEDLINE (eg, OVID MEDLINE). It is practical to explicitly refer to the specific version of MEDLINE.
My most important issue is the semantic quality of the search. The retrieval of this search could be improved by giving more attention to three search techniques: (a) use of truncation, (b) use of natural language, and (c) identification and logical combination of concepts. A search consists of a combination of two or more concepts. In the article by Dr Boks and colleagues, four concepts are identified: (a) ligaments, (b) prediction (et cetera), (c) knee, and (d) magnetic resonance imaging. These four concepts are combined by using the Boolean operator AND. Because some terms used to represent the first concept are intrinsically part of the third concept, a further improvement can be made. An additional search can be formulated wherein only the first, second, and fourth concepts are combined.
The total retrieval would amount to 1022 references—an increase of 114 references compared with the 908 references retrieved by using the authors' original search (performed April 4, 2006; search details available on request).
In general, I would like to suggest as a rule for clinicians and researchers when performing a search to ask for the judgment and help of an information professional. A literature searcher could test the search strategy in order to maximize the quality of the literature search. The information specialist can offer a systematic search plan, which consists of the iterative identification of concepts, the suggestion of databases to be searched, and the translation of the search question in the language of the chosen databases.
Missing references equal missing evidence. Better searching will lead to better evidence.
 |
References |
|---|
 TOP References REFERENCE |
|---|
Department of General Practice, Erasmus Medical Center, Dr. Molewaterplein 50, Rotterdam 3025 GE, the Netherlands
e-mail: s.boks{at}erasmusmc.nl
Mr Schoones suggests that our search strategy (1) was not optimal. Although some points may be of value, we do not believe that his statement that "missing references equal missing evidence" applies.
The use of additional databases may be of value, but, for example, in the Cochrane Library in 2003 no articles of observational studies on this topic had been published. That the EMBASE database contains additional journals is well known (the overlap of journals with MEDLINE is about 60%). However, the overlap for search strategies shows a large variation depending on the subject of interest. Compared with MEDLINE, EMBASE better covers pharmacologic and medication studies, as well as studies on health care policy and management. Further, (national) European journals are better represented in the EMBASE database. For our topic, we did not expect to obtain additional studies to be included with an additional search in EMBASE, but we cannot state that for sure. An important problem of the EMBASE database is that it is not available for free, as is the MEDLINE database; such searches can therefore not be reproduced by every reader and/or referent.
Mr Schoones suggests that the search could be improved by giving more attention to the use of truncation, without mentioning how. We have already used truncation for most items to improve the search (presented in table 1 of our article). We did not use natural language for feasibility reasons but expect that this is not a major drawback.
In his final comment, Mr Schoones suggests to leave out "knee" as part of the search strategy. We do not understand why this should be done, since the aim of our review was to describe the natural course of posttraumatic knee lesions. More important, when performing the search as was suggested by Mr Schoones, we retrieved 1934 references (performed April 30, 2006) instead of the 1022 mentioned by Mr Schoones. In a quick overview of the 1015 "missed references," we found only studies on other topics, such as ankle, elbow, wrist, hip, spine lesions, cardiology, gynecology, and even psychiatry. This suggests that our strategy was more specific.
In general, when performing a systematic review, the aim is to combine high sensitivity of the search strategy (finding all available evidence), with reasonable specificity (restricting the amount of work). We invite Mr Schoones to perform the additional searches as he suggested and to show which relevant articles were missed in our study because of our search strategy.
|
REFERENCE |
|---|
|
TOPReferencesREFERENCE |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. de Jonge, C. L. Bockting, J. W. Schoones, P. T. Ninan, R. M. Poole, G. L. Stiles, E. H. Turner, and R. A. Tell Selective Publication of Antidepressant Trials N. Engl. J. Med., May 15, 2008; 358(20): 2180 - 2182. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
