Describe the major available electronic resources
Describe how to build a search strategy
Describe some alternate sources for finding trials
Describe what to do once you get your search results
1. Searching for Trials for a
Systematic Review
US Cochrane Eyes and Vision Group
Completing a Cochrane Systematic Review Workshop
July 18, 2013
2. Describe the major available electronic
resources
Describe how to build a search strategy
Describe some alternate sources for finding
trials
Describe what to do once you get your
search results
Objectives
3. OK, I Have a Title-
What next?
If you are doing a Cochrane review, work with the
Trials Search Coordinator for the relevant Cochrane
review group.
Work with an information specialist trained in searching
for systematic reviews
Use Chapter 6 in the Cochrane Handbook as a guide
Include the search protocol in your protocol (strategy
and sources)
Document everything you do and when you do it!!!!
4. Identification of evidence
Develop your protocol for conducting the search:
What sources
• Databases
• Hand searching
How – search strategies
How to make decisions (duplicate screening?)
How to document – dates, numbers included,
excluded
5. Cochrane Central Register of Controlled Trials
(CENTRAL) (www.thecochranelibrary.com)
Includes about 650,000 citations
Includes MEDLINE and EMBASE searches
Includes Cochrane group specialized registers
MEDLINE/PubMED (www.ncbi.nlm.nih.gov/sites/entrez)
Includes 22 million citations; > 5,600 journals in 39 languages
Citations fully indexed from 1966 forward
PubMed has most recent – but non-indexed articles
EMBASE (www.embase.com)
Includes 25 million citations > 7,600 journals
Includes MEDLINE
Includes > 6 million conference abstracts
Major electronic databases used to
search for clinical trials
6. Other important databases
National and regional databases (often local language)
LILACS (bases.bireme.br)
Subject-specific databases
CINAHL (www.cinahl.com)
PsycINFO (www.apa.org/psycinfo/)
OTSeeker (www.otseeker.com)
Citation databases
Web of Science (www.thomsonreuters.com)
Scopus (www.scopus.com)
Dissertations, thesis databases
ProQuest (www.Proquest.com)
Grey literature databases
Opengrey (www.opengrey.eu) – formerly SIGLE
7. Will review be limited to RCTs or will it include other
study designs?
Will adverse events be included?
Is there any geographic consideration?
Is there a limited time period when the intervention
was used?
Other issues to consider
8. Add to search strategy
Plurals - (acuity, acuities)
Abbreviations - (e.g., CNV for choroidal neovascularization)
Synonyms -
Lucentis, ranibiuzumab
Avastin, bevacizumab
Spelling variations - (randomized/randomised)
Truncation – (antibod* for antibody or antibodies)
9. Medical Subject Headings (MeSH)
- Descriptors (with thesaurus)
- Most specific term used
Textword searching
- titles for all records, abstracts since 1975
- truncation useful
Publication type
Sensitivity (recall) vs Specificity (precision)
- no gold standard available
Searching trials on PubMed
10. 1. Start with simple search strategy
2. Run search and retrieve reports
3. Analyze MeSH and text words of studies
fitting your criteria
4. Re-run search with revised strategy
5. Repeat steps 2 – 4 if necessary
6. Run optimal search strategy
7. Retrieve reports identified with optimal
search strategy
Development of
MEDLINE search strategy
11. To develop simple search strategy,
start with your question
Population
Intervention(s)
Comparison(s)
Outcome
12. P
I
C
O
For patients with choroidal neovascularization
associated with age-related macular degeneration,
do intravitreal injections of Lucentis, when compared
with Avastin, prevent vision loss?
Individuals with choroidal neovascularization
associated with age-related macular degeneration
Lucentis, ranibizumab
Avastin, bevacizumab
Change in visual acuity or visual field
13. Searching PubMED
Start with the general format:
(Population OR synonym# 1 OR synonym #2) AND
(Intervention OR synonym # 1 OR synonym#2) AND
(Comparator OR synonym#1 OR synonym#2) AND
(Outcome OR synonym# 1 OR synonym # 2) AND
Add study type filter terms
14. How do we find the
terms to use in the
search strategy?
28. Completed search strategy
(this is added to the Cochrane “highly sensitive search strategy”)
#1 exp macular degeneration/
#2 exp retinal degeneration/
#3 exp retinal neovascularization/
#4 exp choroidal neovascularization/
#5 exp macula lutea/
#6 maculopath$.tw.
#7 ((macul$ or retina$ or choroid$*) adj3 degener$).tw.)
#8 ((macul$ or retina$ or choroid$*) adj3 neovasc$).tw.)
#9 ((macula$ adj2 lutea.tw
#10 or/1-9
#11 exp angiogenesis inhibitors/
#12 exp angiogenesis inducing agents/
#13 exp vascular endothelial growth factors/
#14 (lucentis$ or avastin$ or ranibizumab$ or bevacizumab$).tw
#15 (anti adj2 VEGF$).tw.
#16 (endothelial adj2 growth adj2 factor$).tw.
#17 or/ 11-16
#18 10 AND 17
P
I,
C
Add study filter terms to identify study type
29. Common errors in search
strategies
Spelling errors
Missed spelling variant (randomized instead of random)
Truncation error (methods* instead of method*)
Logical operator error (using NOT instead of AND)
Wrong line number (when combining lines)
MeSH and free text terms on same line
Irrelevant MeSH term
Missed MeSH term
Unwarranted explosion of MeSH terms
Redundancy without rationale
Search strategy not tailored for other databases
Sampson et al, J Clin Epi 2006
30. Use of all databases may be necessary to identify all reports
Lawrence et al Inj Prev 2008
31. Registers of ongoing or completed trials
www.clinicaltrials.govwww.clinicaltrials.gov
www.controlled-trials.comwww.controlled-trials.com
www.who.int/ictrp/enwww.who.int/ictrp/en (WHO portal)
Clinical Trial registers
32. Google Scholar vs PubMed
Search engine No citations
Mean (SD) Median (IQR)
Google scholar 2211 (3999) 1040 (339,1958)
PubMed 44 (47) 35 (20,53)
Identified 22 Drug Information Rounds review articles
Searched PubMed or Google Scholar with 2 keywords
(MeSH) or text
Google presented “older” literature first compared with
PubMed - Freeman et al Pharmacotherapy 2009
Similar results reported using clinical query (PubMed)
and advanced scholar search (Google scholar)
- Anders et al (Resp. Care 2010)
33. Journals in topic area
Conference proceedings in area
Bibliographies of related systematic reviews
More efficient and more accurate to use SCOPUS to
download all references in a review than to review them
manually - Chapman et al, Health Information and Libraries J 2010)
Handsearching
34. Other sources of trial results
FDA
Pharmaceutical company registers
Communication with experts in the field
www.clinicaltrials.gov
35. Systematic Snowballing
Use included articles as a source to identify missing
studies:
Reference lists - check cited articles = backward
searching
Web of Science, SCOPUS - check citing articles =
forward searching
(“Related articles”/ “Find Similar”) in MEDLINE,
EMBASE) – also could use to find new terms to
expand your search strategy
36. Use of all sources may be necessary
to identify all reports
Sources of studies for a systematic review for studies
promoting a shift from use of cars to walking or cycling
Source Total No. No. in review %
Health databases 4 3 14
Citation databases 8 3 14
Other databases 8 2 9
Transport database 33 9 41
Internet search 10 3 14
Handsearch abstracts 6 2 9
Total 69 22 100
Ogilvie et al JECH 2005
37. Screening search results
Full record citations downloaded from electronic
databases into an electronic file – either a text file or a
reference manager file
Files distributed to two team members for screening
Yes, Maybe, No
Systematic reviews - obtain full copy for review of references
Disagreements resolved by consensus or by third team
member
Electronic documentation
Reference manager – use new field to enter yes/no/maybe/ref
Word – identify using highlighted text, comment box, copy/paste
to new file
Excel – possible, but difficult to manage
38. Screening full text articles
Develop and use a data collection form (paper or
electronic)
Query inclusion/exclusion criteria
Duplicate screening - Yes, no, unclear
Disagreements resolved by consensus or by third team member
Responses for excluded studies provide “reason for
exclusion” in RevMan
Articles with unclear designation usually need author
contact –go into “awaiting assessment” in RevMan
Start to identify multiple articles from same trial
Authors
Numbers of randomized participants
Interventions
39. Document your findings
When you searched (month/year)
Where you searched
Electronic databases
Handsearches
Trial registers
Communications with experts, industry
Bibliographies, etc.
Internet
How you searched
Search strategies
What you found
PRISMA flow diagram
40. PRISMA Flow of Information
No. records identified
through database searching
No. additional records identified
through other sources
No. records
screened
No. studies included
in qualitative synthesis
No. full-length articles
assessed for eligibility
No. records after
duplicates removed
No. records excluded
No. full-length articles
excluded, with reasons
No. studies included
in quantitative synthesis
41. Conclusions
Develop a search strategy for electronic database
searching using an iterative approach, starting with
your question
Make sure to perform searches from all additional
sources (adapting strategy as necessary)
Document when, where, and how you searched and
what you found
Notas do Editor
When creating a complex search strategy as done by a Trials Search Co-ordinator (TSC) or medical librarian, the following is used (similar to when formulating a clinical question). TSCs tend to use population, intervention(s) and condition in their searches but not specific outcomes as this might lead to the search being too narrow. In CENTRAL you do not need to use any terminology for trials as it is made up mainly of RCTs and CCTs. However, RCT filters can be obtained from your Review Group for searching MEDLINE and EMBASE.
Here is the completed strategy, using both MeSH and text words. Please note that truncation has been used on the text words ie macula* with and without the r. Text words are in lower caps, MeSH in upper caps. We have put in lines to split the condition and interventions. Explain that it was decided to leave out the population as well as names for other treatments. Joyce – you can now run the search in CENTRAL.
The Science Citation Index provides access to current and retrospective bibliographic information, author abstracts, and cited references. Amongst other things you can do cited reference searching on it.
Manual searching includes handsearching, whereby you identify a journal in your field of interest and you literally search from cover to cover. Conference proceedings are also a good source of finding trials that may never make it to journal publication. (May want to mention Roberta Scherer’s study regarding ARVO abstracts and how many trials in abstract form never made it to publication in a journal). A good reference point is also the bibliographies of included studies, or other reviews on the subject.